# Aurochs — Stefan Kovalik # https://aurochs.agency # Full content for LLM ingestion. See also: llms.txt (summary version) # People decide in milliseconds. I design for that. --- ## Aurochs — Stefan Kovalik URL: https://aurochs.agency/ Design   Development   Hosting People decidein milliseconds.People decidein milliseconds. I design for that. Revenue 3–4x. Same product. Different perception. Traffic comes in. Sales don’t. The product isn’t the problem. Nobody is thinkingabout your website. Your visitors are filtering and deciding before conscious thought kicks in. 50 milliseconds. That’s all the brain needs to form a judgment. Gut reactions they can’t explain are deciding whether they stay or leave. I built a methodology around that gap. Five layers between a visitor and a conversion. Miss one and the rest don’t matter. Cognitive Load / 50ms / Trust / Behavior / Action Read how it works → Results The receipts. Revenue 3x Simply Smart Home Simply Smart Home made digital photo frames. Good product, steady sales. But “digital photo frame” sounds like a gadget from 2009. I repositioned them as smart home décor. New brand system, new packaging, new web presence. The whole surface area. Same hardware inside. Revenue tripled within the year. The partnership lasted four years. Disney licensed the brand. “Stefan designed the brand, the packaging, and the product. Everything people see.” Simply Smart Home Revenue 4x Vacuum Sealers Unlimited. Ten years, same site. Incremental optimization, no redesign. “The website Stefan built me helped put my 3 boys through college.” Owner, Vacuum Sealers Unlimited 50% → 75% Online ticket sales at iO Theater. New booking experience for a 40-year Chicago improv institution. “Stefan made our site in half the time, and even better looking than the agency we hired before him.” Managing Director, iO Theater Retail & licensing Tell me what’s not converting → Origin I figured this outin a nightclub. ~300 people on a busy night in Santa Barbara. Same last call song, house lights slam on, and I’m barking at people to leave. A sour note saying “get out” on an otherwise good time. Always felt off to me.So I started walking tables at 1:10. “Hey, last call’s coming up. Good time to grab any drinks you want. Thanks for coming out tonight.” No urgency. Just information, a suggestion, and a thank you.By 1:30, people were saying goodnight and heading for the door on their own. Happier than the old way ever got them. Bar sales went up in those last 20 minutes, and we closed 30 minutes earlier than usual.~300 people on a busy night in Santa Barbara. Same last call song, house lights slam on, and I’m barking at people to leave. A sour note saying “get out” on an otherwise good time. Always felt off to me. So I started walking tables at 1:10. “Hey, last call’s coming up. Good time to grab any drinks you want. Thanks for coming out tonight.” No urgency. Just information, a suggestion, and a thank you. By 1:30, people were saying goodnight and heading for the door on their own. Happier than the old way ever got them. Bar sales went up in those last 20 minutes, and we closed 30 minutes earlier than usual. Right information, right time, right tone, and people cooperated on their own. A psychology degree from UCSB gave me the language. Fifteen years of building websites gave me the laboratory. That pattern became Perception-First Design™. What I build I got tired of re-explainingmyself to my own tools. Every AI conversation starts from zero. Your context, your constraints, your decisions: gone. So I built Cognograph, a spatial AI canvas where the arrangement of your project is the context. Along the way, I automated my design thinking into Forge, a Perception-First Design™ audit for any URL. Open-sourced the methodology as a free Claude Code skill. Wrote Make Me Think, a 12-chapter book on why your visitors aren’t thinking. This website, a patent-pending AI canvas, an automated design audit, an open source methodology, and a 12-chapter book. Twelve weeks. I figured it out. 4 patents pending / Perception-First Design™ / Open source Cognograph → Forge → Make Me Think → Services I ship in weeks what agenciesquote in months. Design, development, and hosting. One person, no overhead. You pay for the work, not the org chart. Growth Infrastructure from $300/mo Your site becomes my job.And I pick up the phone. Enterprise-grade hosting, sub-second load times, and security that’s always on. When something needs attention, I handle it before you notice. Hosting + daily backups Security monitoring + updates Performance optimization Same-day response, 7 days Month-to-month after first year. Included with all site builds. Strategic Partnership from $1,200/mo Your revenue is the scoreboard.I move the numbers. Everything in Growth Infrastructure, plus Perception-First Design applied to your entire surface area: your brand, your pages, your conversion paths. I figure out where you’re losing revenue and fix the perception causing it. Results compound. Everything in Growth Infrastructure Perception-First Design applied to your brand Ongoing design + strategy Your own dashboard: reports, invoices, and action items in one place Priority response, strategy calls Month-to-month. 30-day notice. Not sure which tier? Don’t worry about it. See full service details → Start with a conversation. I’ll look at your site and tell you what I find. We’ll figure out if this is a fit. Let’s talk → Stefan Kovalik   San Francisco Start with what’s not working. Send the URL. I’ll look at it and share what I find. No pitch, just an honest read. We’ll figure out if it makes sense to work together. stefan@aurochs.agency Weekday response within 48 hours. I have room right now. --- ## Services URL: https://aurochs.agency/services/ Services Your site couldfeel like this.Your site couldfeel like this. Everything here runs in a browser. Design, development, and infrastructure: one practitioner, one methodology, one bill. How It Works Five steps. No mystery. Every engagement follows the same methodology. The scope changes; the rigor doesn’t. 01 Discovery I ask uncomfortable questions about your business, your customers, and where money is leaking. I dig until I understand what’s actually happening, not what you think is happening. 02 Perception Audit I run your site through the Perception-First Design™ framework: cognitive load, first impressions, processing fluency, perception bias, decision architecture. Where are visitors bouncing and why? 03 Strategy A prioritized plan that addresses the biggest perception barriers first. What to fix now, what to fix later, and what to leave alone. No busywork. 04 Build You see progress instead of surprises. Style tiles, wireframes, live prototypes, production. Each pass refines until it clears the perception bar. 05 Optimize Perception work isn’t a one-time fix. I measure, refine, and adapt as your business evolves. One client’s revenue tripled within a year. Results Same products. Different perception. Revenue 3x Simply Smart Home Repositioned “digital photo frames” as smart home décor. New brand system, packaging, web presence. Same hardware. Revenue tripled within the year. “Stefan designed the brand, the packaging, and the product. Everything people see.” Revenue 4x Vacuum Sealers Unlimited Ten years, same site. Incremental optimization, no redesign. The methodology compounds. “The website Stefan built me helped put my 3 boys through college.” 50% → 75% iO Theater Online ticket sales for a 40-year Chicago improv institution. Built in half the time of the agency they hired before. “Stefan made our site in half the time, and even better looking than the agency we hired before him.” Retail & licensing Ongoing Partnerships Two tiers. Both month-to-month. Growth Infrastructure from $300/mo You stop thinking about your website. It just works. Managed hosting, performance monitoring, security updates, and ongoing optimization. When something needs attention, I handle it before you notice. Your site runs on infrastructure I built and maintain, not a shared hosting box you’ll outgrow in six months. Managed WordPress hosting on dedicated infrastructure Daily automated backups with 30-day retention Security monitoring, firewall, and malware scanning Plugin and core updates tested before deployment Performance optimization (caching, CDN, image compression) SSL certificate management Uptime monitoring with same-day response, 7 days Monthly performance report Month-to-month after first year. Included with all new site builds. Your data, your domain, always. Strategic Partnership from $1,200/mo The approach that tripled a brand’s revenue, pointed at your business. Everything in Growth Infrastructure, plus Perception-First Design™ applied to your brand, your pages, your conversion paths. I’m the answer to “make it pretty.” I know what pretty is, what makes it pretty, and when to break it on purpose. Everything in Growth Infrastructure Perception-First Design™ applied to your brand and site Ongoing design iterations and A/B testing Conversion path optimization Landing page design and development Brand identity refinement Priority response and strategy calls Quarterly perception audit and roadmap Month-to-month. 30-day notice. Your data, your domain, always. Tell me what you need → Project Work One-off builds. Every project includes the full Perception-First Design™ methodology: discovery, perception audit, strategy, build. No template drops. Brochure Sites from $6,000 5–10 pages. Designed to clear perception barriers and get visitors to act. Includes brand system, responsive design, and CMS setup. Complex Builds from $15,000 E-commerce, custom functionality, integrations, multi-step flows. Scoped precisely. No surprises. WooCommerce, custom WordPress, or headless architectures depending on what the project needs. Brand Identity from $5,000 Logo, typography, color system, brand guidelines. Built on Perception-First Design™ principles: what your brand communicates before anyone reads a word. Style tile → brand system → production assets. Let’s scope it → Consulting The methodology, taught. Perception-First Design™ as workshops, audits, and keynotes. I went looking for other designers working at this intersection. What I found were developers, researchers, founders, tool makers. The design voice is barely in the room. That's the gap I'm in. + View engagements and pricing Perception-First Design™ Workshop $2,500 Half-day / 4 hours Your team learns the 5-layer framework applied to your product. Live audit. Actionable fixes ranked by impact. You leave with a list, not a theory. Perception-First Design™ Workshop + Report $5,000 Full-day / 8 hours Everything in the half-day, plus a detailed written report. Scored findings, fix specifications, before/after projections. The document your team references for the next 6 months. Perception-First Design™ Audit from $2,000 1–2 weeks I read your product through the 5 PFD layers and hand you a fix roadmap ordered by impact. Scored findings per layer. Fix specifications. Architecture rewrite where it matters. Executive summary for stakeholders who need the short version. Keynote $10,000+ 45–60 minutes “Designing from the Edge.” How cognitive constraints become design material. I built a methodology on what the field treats as an edge case. They leave knowing why it matters. Tell me what you need → What the methodology touches What I work with. Every engagement is scoped to your business. Design Build Grow Protect Design Brand, web, and product design. Built on how people actually see. Brand IdentityLogo, guidelines, repositioning Web DesignResponsive, mobile-first, dark/light App DesigniOS, Android, Electron, React Product DesignPackaging, 3D renders, retail displays Design SystemsTokens, components, style guides Motion & AnimationWebGL, shaders, CSS, canvas Build I build it. It ships. WebsitesWordPress, static, Cloudflare Pages E-commerceWooCommerce, Stripe, Authorize.net CMS DevelopmentHubSpot, WordPress, headless Custom ApplicationsElectron, React, Node.js Interactive EffectsWebGL2, GLSL shaders, Three.js IntegrationsAPIs, Workers, serverless, webhooks Grow What you measure is what you fix. Perception AuditPerception-First Design™ 5-layer analysis Conversion OptimizationA/B testing, funnel analysis, CRO Search & AI DiscoverySEO, GEO, llms.txt, structured data Content StrategyCopywriting, editorial, social assets AnalyticsGA4, GTM, tracking audits, dashboards AccessibilityWCAG 2.2 AA, audits, remediation Protect Your site stays up. Your data stays yours. Managed HostingWordPress, Cloudflare, cPanel, Linux Security HardeningWAF, Turnstile, DKIM/SPF/DMARC PerformancePHP-FPM, OPcache, CDN, edge caching Database AdminMariaDB, InnoDB, query optimization Backup & RecoveryDaily backups, disaster recovery, SSL Emergency SupportIncident response, fraud investigation FAQ Common questions. Can’t I just use Squarespace / Wix / AI tools? Sure. Try it. I mean that genuinely. Go build something with the tools available. They’re good. They’ll get you 80% of the way to something that looks like a website. That last 20% is the difference between a site people visit and a site that produces outcomes. You’re reading about it. What platforms do you support? WordPress, WooCommerce, and custom builds. If your project needs a headless architecture, static site, or something non-standard, I’ll scope it honestly. If it’s outside my wheelhouse, I’ll tell you. What if I want to leave? You own your content, your domain, and your data. Always. I’ll export everything and help with the transition. The custom infrastructure (hosting stack, licensed plugins, headless layer) stays with me, but your site will run on any standard WordPress host. No hostage situations, no surprises. Is there a contract? Every site build includes one year of Growth Infrastructure hosting. After that, month-to-month with 30-day notice. I keep clients because the work produces results, not because of a contract. What’s the typical turnaround for a new site? Brochure sites: 3–4 weeks from kickoff. Complex builds: 6–10 weeks depending on scope. I don’t rush. Perception work requires iteration. But I also don’t drag things out. How does onboarding work? A 30-minute call to understand your business, your customers, and your goals. Then I do the perception audit on my end and come back with a strategy. You don’t need to prepare a 50-page brief. I ask the right questions. What does “perception audit” mean? I run your site through the five layers of Perception-First Design™: cognitive load, first impressions, processing fluency, perception bias, and decision architecture. It’s a structured diagnostic, not “I think this looks nice.” You get a prioritized list of what’s costing you conversions and why. Do you work with agencies? Selectively. If you need a white-label partner who applies a real methodology and doesn’t just push pixels, let’s talk. I’m not interested in being the cheapest vendor in the stack. Get Started Start with what’s not working. Tell me the URL and what you’re seeing. I’ll tell you what I see. If it’s a fit, we’ll scope it together. stefan@aurochs.agency --- ## Perception-First Design Consulting URL: https://aurochs.agency/consulting/ Consulting The methodologynobody else has. 100 citations. Zero competitors. Perception-First Design™ workshops, audits, and keynotes. The Methodology Perception-First Design is a 5-layer cognitive psychology framework for designing how people actually perceive, decide, and act. Not opinions. Not best practices. 77 citations from two non-overlapping literatures (55 perception science, 22 ADHD neuroscience), plus international research integration. 100 total. The only methodology that connects psychology to design to implementation to measurement. Layer What it governs Layer 0: Cognitive Load Cognitive Load. How much mental effort the interface demands before anything else happens. L1 First Impression. The 50ms judgment. Pre-conscious trust or rejection. L2 Processing Fluency. How easy it feels to understand. Fluent = trustworthy. L3 Perception Bias. The heuristics visitors use to evaluate quality, price, and fit. L4 Decision Architecture. How choices are structured to guide behavior without coercion. Read the full framework → Engagements Perception-First Design™ Workshop $2,500 Half-day / 4 hours Your team learns the 5-layer framework applied to YOUR product. Live audit of your site or app. Actionable fixes ranked by impact. You leave with a list, not a theory. Perception-First Design™ Workshop + Report $5,000 Full-day / 8 hours Everything in the half-day, plus a detailed written report. Scored findings, fix specifications, and before/after projections. The document your team references for the next 6 months. Perception-First Design™ Audit from $2,000 1–2 weeks I read your product through the 5 PFD layers and hand you a fix roadmap ordered by impact. Scored findings per layer. Fix specifications. Architecture rewrite where it matters. Executive summary for stakeholders who need the short version. Keynote $10,000+ 45–60 minutes “Designing from the Edge” — how cognitive constraints become design material. For conferences, offsites, and leadership teams. I built a methodology on what the field treats as an edge case. They leave knowing why it matters. Receipts Revenue 3x Simply Smart Home Revenue tripled. Repositioned digital photo frames as smart home décor. Same hardware, different perception. Disney licensed the brand. Partnership lasted four years. 50% → 75% iO Theater Online ticket sales share for a 40-year Chicago improv institution. New booking experience. Built in half the time the previous agency quoted. 32ms Vacuum Sealers Unlimited Homepage load time, down from 3–4 seconds. Ten years, same client. Incremental optimization. The owner put three kids through college on that site’s revenue. The Position I went looking for other designers working at this intersection. What I found were developers, researchers, founders, tool makers. The design voice is barely in the room. Nobody is grounding design decisions in cognitive psychology. That is the gap. Perception-First Design fills it. Designed from the cognitive edge case. Works better for everyone. Start with a conversation. Tell me what you're working on. I'll tell you whether Perception-First Design is the right tool for it. No pitch, just an honest read. stefan@aurochs.agency Weekday response within 48 hours. --- ## Perception-First Design Methodology URL: https://aurochs.agency/perception-first-design/ Design grounded in how people decide Perception-FirstDesign™ Remove the perception barriers that stop interested visitors from becoming buyers. You’re losing sales to perception, not product. Your product works. Your pricing is competitive. But your visitors aren’t evaluating your design. They’re not reading your copy. They’re pattern-matching, filtering, deciding. All before conscious thought kicks in. 50 milliseconds. That’s how long the brain takes to form a first impression. Your site is being judged by cognitive processes your visitors can’t articulate and you can’t see. Most sites are built around what the business wants to say. The visitor’s brain is asking a different question: does this feel right? That judgment happens before logic. Perception-First Design closes that gap. Psychology-Informed Web Design Framework Five layers. One goal. Layer Principle Research Base Causal Chain Layer 0: Cognitive Load Reduce decision fatigue Sweller (1988) Less thinking → more action Layer 1 Design for 50ms judgments Lindgaard et al. (2006) Trust forms before content loads Layer 2 Easy to process = trustworthy Reber & Schwarz (1999) Fluent design → perceived credibility Layer 3 Design for behavior, not stated preferences Kahneman (2011) Emotional resonance → action Layer 4 Structure choices to guide behavior Sunstein & Thaler (2008) Clear path → completed purchase See the full framework with all citations: PFD Reference → Layer 0: Cognitive Load Don't Make Me Think “Can’t perceive anything without the bandwidth to do so.” Cowan’s research shows working memory holds 3 to 5 chunks, not Miller’s famous 7. Every unnecessary element on your page is stealing from that budget. Recent research adds a second path: when intrinsic interest is high enough (Dodson’s Interest-Based Nervous System: passion, novelty, challenge, urgency), it can override load ceilings. The foundation is not just “reduce friction.” It is “reduce friction or increase the reason to push through it.” In practice: reduce choices at each step, show only what’s needed now, use smart defaults for 80% of users, and eliminate unnecessary fields, steps, and decisions.Cowan’s research shows working memory holds 3 to 5 chunks, not Miller’s famous 7. Every unnecessary element on your page is stealing from that budget. Recent research adds a second path: when intrinsic interest is high enough (Dodson’s Interest-Based Nervous System: passion, novelty, challenge, urgency), it can override load ceilings. The foundation is not just “reduce friction.” It is “reduce friction or increase the reason to push through it.” In practice: reduce choices at each step, show only what’s needed now, use smart defaults for 80% of users, and eliminate unnecessary fields, steps, and decisions. Causal chain: Fewer choices → less mental effort → faster decisions → more conversions. Simply Smart Home: Site presented “very matter-of-fact statements and not much else.” Users had to work to figure out what the product did or why they’d want it. Restructured content to answer the questions users actually have: what is this, who is it for, why should I care. Before they have to ask. Read more: Cognitive Load → Read the mechanism Content coming soon. See the complete reference for the full mechanism description. Layer 1 First-Impression Architecture “Users form first impressions the same way on sites as they do with people in real life, and you have a split second to make your best one.” You have 50 milliseconds to pass the “is this trustworthy?” test. The threshold isn’t world-class. It’s “not sketchy.” If your site triggers the alarm, they’re gone before conscious thought kicks in. Hero sections that immediately communicate value. Strategic use of faces, the only thing humans pattern-recognize from birth. Visual quality consistent with price point. Color and typography that signal the right personality.You have 50 milliseconds to pass the “is this trustworthy?” test. The threshold isn’t world-class. It’s “not sketchy.” If your site triggers the alarm, they’re gone before conscious thought kicks in. Hero sections that immediately communicate value. Strategic use of faces, the only thing humans pattern-recognize from birth. Visual quality consistent with price point. Color and typography that signal the right personality. Causal chain: Faces in hero → instant pattern recognition → positive emotional response → trust established before thinking begins. Simply Smart Home: Added smiling faces in hero product graphics: smart frames displaying family photos. Faces are the only thing humans pattern-recognize from birth. Drove direct conversions and brand awareness; visitors felt warmth, not “another gadget.” iO Theater: Hobbyist WordPress site triggered the “this doesn’t look like a real theater” alarm. Clearing that perception barrier alone moved online ticket sales from 50% to 75%. Read more: The 50-Millisecond Verdict → Read the mechanism Content coming soon. See the complete reference for the full mechanism description. Layer 2 Processing Fluency “They wouldn’t take this business seriously because they don’t do so on their website.” Reber and Schwarz showed that experiences which feel easy to process are judged as more trustworthy, more beautiful, and more true. Fluency is not simplicity. It is cognitive ease. Clean typography with adequate contrast. Predictable layouts that match mental models. Unified branding and collateral. No mixed signals.Reber and Schwarz showed that experiences which feel easy to process are judged as more trustworthy, more beautiful, and more true. Fluency is not simplicity. It is cognitive ease. Clean typography with adequate contrast. Predictable layouts that match mental models. Unified branding and collateral. No mixed signals. Causal chain: Unified branding → feels professional → perceived as trustworthy → clears the “would I buy from this?” bar. Simply Smart Home: Basic template, no branding system, $150+ price point. Customers “wouldn’t take this business seriously.” Created unified brand system and style guide. Perceived quality matched price point. Price objections decreased, vendors became more interested. Competitor insight: “Aura frames put weights in their frames to make them feel less cheap and charge more.” Perception of quality is manipulable. Read more: The Feeling of Truth → Read the mechanism Content coming soon. See the complete reference for the full mechanism description. Layer 3 Perception Bias Optimization “Protect the customer from internal stakeholder preferences.” Design for how people actually behave and feel, not what they say or what features do. Research behavior, not stated preferences. Find the emotional core: what do they feel after using this? Audit messaging: is it feature-focused or outcome-focused? Compare survey data vs. analytics. The gap is where perception bias lives.Design for how people actually behave and feel, not what they say or what features do. Research behavior, not stated preferences. Find the emotional core: what do they feel after using this? Audit messaging: is it feature-focused or outcome-focused? Compare survey data vs. analytics. The gap is where perception bias lives. Causal chain: Emotional tagline → connects to real desire → resonates with target customer → action. Simply Smart Home: Marketing focused on screen resolution, WiFi, app features. Workshopped the team through “what does this product actually do for the user emotionally?” Discovery: it’s about connection, not technology. Tagline: “Stay connected, even when you’re apart”, which Aura later stole. Best validation possible. Read more: The Gap → Read the mechanism Content coming soon. See the complete reference for the full mechanism description. Layer 4 Decision Architecture “They’re hunters looking for prey, and it’s my job to make a trail for them.” Structure choices to guide behavior. Navigation should reflect user goals, not org charts. Make the desired path the obvious path. Strategic default selections. Show the premium option first so the standard feels like a deal. Reduce choice paralysis through smart categorization. Put your CTA where decisions naturally happen.Structure choices to guide behavior. Navigation should reflect user goals, not org charts. Make the desired path the obvious path. Strategic default selections. Show the premium option first so the standard feels like a deal. Reduce choice paralysis through smart categorization. Put your CTA where decisions naturally happen. Causal chain: Clear path → user finds goal → reduced friction → purchase completed. Simply Smart Home: Homepage “didn’t answer any of the user’s questions, goals, or form a good impression.” Rebuilt as a trail: visitor lands → sees emotional hook (faces, connection) → understands product → sees social proof → clear path to purchase. Read more: The Trail → Read the mechanism Content coming soon. See the complete reference for the full mechanism description. These five layers draw from two largely distinct bodies of cognitive science: perception psychology and ADHD cognitive neuroscience. Connected here for the first time. The full citation set and layer mapping are in the complete reference. Ethics and Limits Every persuasive technique has a shadow side. Perception-First Design uses the same cognitive science that powers dark patterns. The difference is intent and transparency. I make my persuasive mechanisms visible, not hidden. Processing fluency Making things easy to understand is not the same as making them easy to accept uncritically. I design for clarity, not for bypassing judgment. Decision architecture Structuring choices to guide behavior is not eliminating choice. I preserve user autonomy. No hidden options, no false urgency, no confirmshaming. Perception bias Designing for how people perceive is not exploiting how they misperceive. I align design with genuine value, not manufactured signals. What I refuse Fake countdown timers. Hidden unsubscribe flows. Confirmshaming copy. These aren’t conversion optimization. They’re manipulation, and they erode the trust PFD is designed to build. Limitations & ongoing research: PFD’s evidence base draws on strong lab research (processing fluency, 50ms impression formation, cognitive load theory) and practitioner observation across 15 years and dozens of engagements. The five-layer synthesis itself is practitioner-derived, not independently validated as a complete system. Cultural generalizability beyond Western markets remains an open question. I state this because intellectual honesty is a stronger authority signal than overclaiming. Case Studies and Results Same products. Different perception. Revenue 3x Simply Smart Home Repositioned “digital photo frames” as smart home décor. New brand system, packaging, web presence. Same hardware. Revenue tripled within the year. “Stefan designed the brand, the packaging, and the product. Everything people see.” Revenue 4x Vacuum Sealers Unlimited Ten years, same site. Incremental optimization, no redesign. The methodology compounds. “The website Stefan built me helped put my 3 boys through college.” 50% → 75% iO Theater Online ticket sales for a 40-year Chicago improv institution. Built in half the time of the agency they hired before. “Stefan made our site in half the time, and even better looking than the agency we hired before him.” Retail placement & licensing Recognize any of these barriers in your own site? Start with a conversation → Origin I figured out my design methodology in a nightclub. I started as a barback at a nightclub in Santa Barbara, working nights to pay for school. After a month, my boss chewed me out based on one bad-looking moment during a rush. I pushed back professionally. The next night he promoted me to security manager. That was the first time I consciously identified a perception gap: his impression didn’t match reality, and my response addressed the perception, not the facts. Three years managing that door taught me that people decide whether to trust an environment in milliseconds. Before they read the drink menu. Before they hear the music. Something in the lighting, the crowd density, the staff posture either passes their unconscious trust check or it doesn’t. When I moved into web design, I recognized the same dynamic. Visitors aren’t reading your copy in their first 50 milliseconds. They’re running the same unconscious trust audit they’d run walking into a bar. My psychology degree from UC Santa Barbara, where I focused on cognitive psych, gave me the science. The nightclub gave me the instinct. Perception-First Design™ is both. I have ADHD. The cognitive constraints I designed around (limited working memory, sensitivity to disfluency, impatience with unnecessary friction) are constraints every user has. Mine are just louder. A design that works for the constrained case tends to work better for everyone. Researchers call this the curb-cut effect. I call it the reason the methodology works. Read the full origin story: Chapter 1: The Bouncer → Common Misconceptions Five things to get straight. “Make it pretty” won’t get you there. Visual polish without perception strategy is decoration. Looks don’t convert. Cleared perception barriers do. Generic UX best practices fall short. “Add more whitespace” and “make the button bigger” are tactics without strategy. This framework explains why certain changes work and others don’t. It won’t manufacture demand. This converts existing interest into action. If nobody wants your product, fix the product first. Not perfectionism. “Good enough” is a feature, not a bug. The goal is clearing the trustworthiness threshold, not winning design awards. Every engagement is custom. Each principle requires understanding your specific customer. The hunters are different; the trail must match what they’re hunting for. Methodology Tools Where the methodology lives. The tools demonstrate the methodology. They are not the product. [Co] Spatial Canvas AI Workflow Orchestration Perception-First Design isn’t just a consulting framework. I built the tooling to practice it. A spatial AI workflow canvas where you drop nodes, connect context, and trigger automations. Conversations, terminals, research, tasks, artifacts, and documents live as connected nodes on an infinite canvas, orchestrated visually instead of buried in chat threads. Every PFD engagement runs through it. Nodes for each layer, edges that trace causal chains, AI-assisted analysis that surfaces the perception barriers you’d otherwise miss. Four provisional patents cover the spatial trigger mechanisms that make it work. Learn more · Try the canvas · Open source (AGPL-3.0) [Fe] Forge Perception-First Design™ Intelligence Submit any design artifact (a URL, screenshot, HTML, or source directory) and Forge scores it across all five perception layers. Each round produces specific, actionable prescriptions grounded in the peer-reviewed research above. Iterate until you cross 85+. Run it from Claude Code, any MCP-compatible client, the REST API, or the web app. The methodology, automated. Forge includes my calibration: practitioner-tuned scoring, leniency bias defenses, and correction layers from 15 years of applied work. Want the framework without the calibration? The open source Claude Code skill gives you the full PFD methodology for free. forge.aurochs.agency forge converge Round 3 of 4 · Score: 71 → est. 88L0  92  Load within limitsL1         64  First impression needs workL2         38  Fluency violations in nav + modalL3         72  Bias optimization adequateL4         88  Decision architecture strong Five Evaluation Lenses Every PFD engagement runs through five independent evaluations, each with its own scoring criteria and pass threshold. The lenses ensure nothing ships without being tested from every angle that matters. Methodology Is the framework applied correctly and honestly sourced? Value Exchange Will someone pay for this, and is the exchange fair both ways? Operations Can this run sustainably without burning out the person operating it? Technical Readiness Is this scoped correctly and ready to ship? Perception How will a human actually experience this? Each lens scores 5 criteria on a 0-100 scale. Work ships when all five converge above 70. Get Started Three ways in. Apply it to your business, study the methodology, or run it in your own tools. For Your Business Strategic Partnership from $1,200/mo I apply the five layers to your product, market, and customers. Design, development, and perception strategy. One practitioner, one methodology, one bill. See all services → Learn the Methodology The Complete Reference Free 77 citations, five layers in full, the diagnostic protocol, 18 accumulated learnings, evidence tiers, framework comparisons, and the ethics treatment. No gate, no signup. Read the full reference → Use It in Your Tools Claude Code Skill Open Source The full PFD framework as an AI skill. Run perception audits in your own workflow. Framework document, ADHD curb-cut extension, and accumulated learnings included. View on GitHub → Research Foundation Selected citations. CitationYearFindingLayer Sweller1988Cognitive Load Theory (intrinsic vs. extraneous)L0 Kurosu & Kashimura1995Aesthetic-usability effectL1 Reber & Schwarz1999Processing fluency → perceived truthL2 Tractinsky, Katz & Ikar2000“What is Beautiful is Usable”L1 Norman2004Emotional Design (visceral, behavioral, reflective)Bonus Lindgaard et al.200650ms visual appeal judgmentsL1 Thaler & Sunstein2008Choice architecture / Nudge theoryL4 Cowan2010Working memory capacity: 3–5 chunk limitL0 Friston2010Free-energy principleCore Kahneman2011Thinking, Fast and Slow, System 1 / System 2Core Clark2013Predictive processing: brains match input to top-down predictionsCore Reinecke et al.2013Visual evaluation at 17ms (Google Research)L1 Sweller, van Merriënboer & Paas2019CLT reconceptualizedL0 Bujack, Teti, Miller, Turton & Rogers2022Perceptual color space is non-Riemannian: large color differences perceived as less than sum of small differencesL2 ADHD Cognitive Neuroscience CitationYearFindingLayer Mednick1962Flat associative hierarchies → more remote conceptual connectionsL2 Clark & Chalmers1998Extended Mind Thesis: cognition extends into reliably accessed toolsL4 Carson, Peterson & Higgins2003Eminent creative achievers 7× more likely to have low latent inhibitionL2 Dietrich2004Four types of creativity: deliberate/spontaneous × cognitive/emotionalCross White & Shah2006, 2011ADHD adults outperform on divergent thinking, underperform on convergentCross Eisenberg et al.2008DRD4/7R allele: opposite fitness outcomes depending on environmentL3 Kounios & Beeman2009Pre-problem brain state predicts insight vs. analytical solution modeL1 Beaty et al.2014Three-Network Model: creativity requires DMN + ECN + Salience cooperationCross Barkley1997ADHD as self-regulation deficit; rules at point of performanceL0, L4 Castellanos et al.2002DMN hyperconnectivity in ADHDCross Mark et al.2024Context switching costs “almost an hour” for ADHD developers (ICSE)L0 Dodson2005Interest-Based Nervous System (PINCH): Passion, Interest, Novelty, Competition, HyperurgencyL0 Mahan2020Wall of Awful: accumulated emotional barrier to task initiationL1 Taylor et al.2020ADHD = only positive predictor of divergent thinking (n=60)Cross UK Dept. for Business and Trade202425% higher AI satisfaction among neurodivergent workersCross JetBrains2023Low perceptual load specifically benefits ADHD-symptomatic developersL2 Newman & Begel202510.6% of programmers have ADHD (ICSE 2025)Cross Deshmukh2025Neurodivergent-aware productivity: systems & AI-based human-in-the-loop framework for ADHD. arXiv:2507.06864Cross 32 shown. 78 peer-reviewed citations total at pfd-reference. Citation with attribution is welcome. Unauthorized commercial use requires written permission. --- ## Perception-First Design — Complete Reference URL: https://aurochs.agency/pfd-reference/ The Complete Reference Perception-First Design™ Source document for the methodology. 78 academic citations across cognitive psychology, HCI, behavioral economics, design theory, STS, and ADHD cognitive neuroscience. 15 years of applied work across e-commerce, SaaS, consumer electronics, and entertainment. Stefan Kovalik · Aurochs · Version 3.4 · March 2026 The Bouncer I figured out my design methodology in a nightclub. I was head of security at a 300-capacity venue. Staff of 15. Every night at 1:30 AM, same problem: lights slam on, DJ announces we’re closed, and I’m barking at 300 drunk people to get out. They hated it. I hated it. Fights happened. Stragglers lingered. The whole thing was hostile. So I tried something different. At 1:10, I walked table by table. “Hey, last call’s coming up. Get any drinks you want. Thanks for coming out tonight.” That’s it. No urgency. No authority voice. Just information, a suggestion, and respect. At 1:30, when the DJ made the announcement and the lights went up, half the venue was already walking out the door. No barking. No fights. And bar sales went up 30% in those last 20 minutes. I didn’t know what to call that. I just knew it worked, and I knew why it worked. The old way treated people like obstacles. My way treated them like people with their own goals who’d cooperate if you gave them the right information at the right time with the right tone. That’s Perception-First Design before I had words for it. Everything I’ve done in the 15 years since; revenue 3x, Costco pallets and Walmart shelves and Disney licensing, improv theaters and smart home brands. It all comes from that same instinct. Design for how people actually perceive and process the world, not how you wish they did. The Thesis The design industry has spent 15 years saying the same thing: users have limited attention, so don’t waste it. Lower friction. Reduce steps. Don’t make them think. My version is different: users don’t think. Not until you make them. Nobody is thinking. They’re on autopilot. The same way you catch a ball without calculating trajectory, or walk without planning each step; that’s how people interact with websites, products, brands, everything. Autopilot. Energy conservation. The brain handles the vast majority of processing unconsciously. Kahneman (2011) calls this System 1: the fast, automatic mode that runs most of cognitive life. System 2, slow and deliberate and conscious, only activates when something demands attention. Contemporary cognitive science calls this predictive processing (Clark, 2013; Friston, 2010). The brain is constantly generating predictions about what it expects to encounter. When reality matches prediction: autopilot. When reality violates prediction: attention fires, System 2 kicks in. That prediction error is what I call the activation point. The design challenge isn’t “don’t waste their limited attention.” It’s that effortful attention is dormant by default. The real question is: when you do make them stop and actually think, when you generate a prediction error that activates them from autopilot, what do they think about? How do they feel? And what do they do next? That’s what I design for. Not the funnel. The activation. The Industry SaysI Say AssumptionAttention is scarce; conserve itEffortful attention is dormant; design the activation GoalReduce friction to goalsCreate the right moments that make someone stop, think, and act Failure modeUser runs out of patienceUser never engages; autopilots right past you MetaphorBudget of attentionAutopilot with activation points Steve Krug wrote Don’t Make Me Think. Great book. It gave me the vocabulary to talk to stakeholders early in my career. His scope was usability, and he stayed in it deliberately. But nobody’s thinking in the first place. Design for that bias, the bias of making someone stop and engage, instead of funneling them to an action with an affordance of attention they never gave you. Why I See This Two things about my brain that I didn’t choose but that turned out to be professional advantages. I’m autistic. After my formal ASD diagnosis, my partner pointed out something I’d been missing: “Do you realize you’re not picking up on me not being interested at all in what you’re talking about, and you seem to be getting frustrated?” Around 2012, a friend I met at a Chicago meetup about emotions in video games suggested I try improv. I went in already analytical. I had impaired social instincts, not absent ones. I didn’t need to learn what a smile means. I had to learn what eye contact can mean, what a pause signals, what “said vs. unsaid” reveals. I learned it like you’d learn an instrument or ballet: theory, practice, application loops, over and over. That means I can see and articulate the unconscious patterns that most designers apply intuitively but can’t explain. People interact with websites the same way they interact with other humans. We form first impressions on products the same way we do when someone walks into a room. It’s all unconscious, and we can’t help it. I also have severe combined-type ADHD. Some experiences are literally painful for me because of how taxing they are on my attention and cognitive load. I’m acutely sensitive to friction, clutter, noise, confusion. The things that bad design inflicts on everyone, I feel at a higher volume. I design because I’m sensitive to these patterns. What most designers theorize about, I feel. And the ADHD gave me something else: years of building mental prostheses. If it’s out of sight, it’s literally out of mind for me, so I keep post-it notes on my desk for today’s tasks. “Out of sight, out of mind” becomes: what MUST we show vs. what we want to show vs. what we could show. Then cut everything but what we must show. Design is prosthesis for human cognitive functions and limitations. I know that because I’ve been building prostheses for myself my whole life. The curb cut effect: what I built for my own cognitive accessibility turned out to help everyone. I’ve come to think of this as a curb-cut effect. The cognitive constraints I designed around — limited working memory, sensitivity to disfluency, impatience with unnecessary friction — are constraints every user has. Mine are just louder. A design that works for the constrained case tends to work better for everyone. The five PFD layers draw from two largely distinct bodies of cognitive science — perception psychology and ADHD cognitive neuroscience — connected here for the first time. The citation lists share no duplicate entries. The Conductor Model I don’t build funnels. I conduct. Think about a musician who’s reached unconscious mastery. They stop thinking about how things are done and focus entirely on how the music makes you feel. That’s the goal. Prime someone with a first impression, then conduct their thoughts and emotions through the experience. Design a mental waterfall. There’s a flow of thought and perception: momentum, weight, cadence, tone, speed, context, priming, expectations. Like singing down a scale in front of a crowd and pausing before the last note. Everyone knows what comes next. That recognition, that pattern completion, is satisfying. That’s what a great CTA feels like: the only possible resolution to the experience that preceded it. The User’s Mental Waterfall The unconscious sequence of questions every visitor processes: What am I trying to accomplish here? → Intent Is there a clear path to that? → Navigation What’s the point of this site? → Value proposition For whom is this? → Do I belong here? What is it trying to tell me? → Messaging: ew or cool? What is it trying to get me to do? → Call to action How do I feel about doing it? → “Prove you’re worth my time” or “Tell me more!” Where’s the win-win? → Mutual value Does this give me what I want? → Delivery The hero section is the activation point; the table visit at 1:10. If it answers questions 1 through 5 in the first second or two, the user is primed. Then you conduct them through 6 through 9 as they scroll. If the hero fails, it’s the lights slamming on. My improv teacher gave me a formula: (Listen) × (Act + React). If Listen equals zero, if the user isn’t attending, then everything else multiplies by zero. The Framework Five layers as a dependency stack. Each depends on the layer below it. Fix them in order, because upstream failures degrade everything downstream. But think of it less as a stack and more as a composition. The foundation sets the key. Each layer adds another instrument. If the foundation is broken, the whole piece is noise. Decision Architecture L4 Perception Bias Optimization L3 Processing Fluency L2 First-Impression Architecture L1 Cognitive Load Reduction Layer 0: Cognitive Load In predictive processing terms (Clark, 2013; Hohwy, 2013): when a design matches the brain’s predictions at every layer, processing is fluid and unconscious. Autopilot. When a layer fails, it generates prediction error that demands conscious attention. Good design means each layer confirms predictions so the brain’s resources flow toward the decisions that matter, not toward deciphering the interface. On the research base: The empirical benchmarks throughout (timing thresholds, working memory parameters, fluency effects) were measured primarily on WEIRD samples (Western, Educated, Industrialized, Rich, Democratic; Henrich, Heine & Norenzayan, 2010). The underlying cognitive architecture is expected to generalize; specific parameter values may vary. PFD is a practitioner synthesis, not an experimental program. The citations establish the mechanisms; the 15-year applied record tests them. Layer 0: Cognitive Load Cognitive Load Reduction “Can’t perceive anything without the bandwidth to do so.” Working memory holds about 3–5 chunks at a time, not the “7 plus or minus 2” you’ve seen quoted everywhere. Miller (1956) was describing the recurrence of the number 7 across disparate information-processing phenomena as a rhetorical observation, not proposing a single working memory limit. That nuance got lost. Cowan (2001, 2010) established the modern estimate: roughly 3–5 chunks when rehearsal and grouping strategies are prevented. Sweller’s Cognitive Load Theory (1988) distinguishes intrinsic load (complexity inherent to the task) from extraneous load (complexity caused by poor design). You can’t change intrinsic load. But every unit of extraneous load you eliminate frees working memory for the decisions that actually matter. This is the foundation because without bandwidth, nothing else works. An overloaded visitor can’t form a first impression, can’t process your brand, can’t connect with your message, can’t find the trail. They’re just gone. Dodson’s Interest-Based Nervous System model (IBNS) proposes a refinement: ADHD attention is regulated by interest, not importance. The PINCH model identifies five drivers that override cognitive load ceilings: Passion, Interest, Novelty, Competition, Humor. This expands L0 from “reduce cognitive load” to “reduce cognitive load OR increase intrinsic interest.” For ADHD users, and partially for everyone under low motivation, load reduction alone is insufficient when interest is absent. The constraint is not just bandwidth. It is bandwidth weighted by engagement.Working memory holds about 3–5 chunks at a time, not the “7 plus or minus 2” you’ve seen quoted everywhere. Miller (1956) was describing the recurrence of the number 7 across disparate information-processing phenomena as a rhetorical observation, not proposing a single working memory limit. That nuance got lost. Cowan (2001, 2010) established the modern estimate: roughly 3–5 chunks when rehearsal and grouping strategies are prevented. Sweller’s Cognitive Load Theory (1988) distinguishes intrinsic load (complexity inherent to the task) from extraneous load (complexity caused by poor design). You can’t change intrinsic load. But every unit of extraneous load you eliminate frees working memory for the decisions that actually matter. This is the foundation because without bandwidth, nothing else works. An overloaded visitor can’t form a first impression, can’t process your brand, can’t connect with your message, can’t find the trail. They’re just gone. Dodson’s Interest-Based Nervous System model (IBNS) proposes a refinement: ADHD attention is regulated by interest, not importance. The PINCH model identifies five drivers that override cognitive load ceilings: Passion, Interest, Novelty, Competition, Humor. This expands L0 from “reduce cognitive load” to “reduce cognitive load OR increase intrinsic interest.” For ADHD users, and partially for everyone under low motivation, load reduction alone is insufficient when interest is absent. The constraint is not just bandwidth. It is bandwidth weighted by engagement. Causal chain: Fewer choices → less mental effort → faster decisions → more conversions In Practice Reduce choices at every step. Audit every dropdown, radio button, and “which one?” moment. Progressive disclosure: show what’s needed now, reveal the rest on demand. Smart defaults: pre-select what 80% of users choose. If you don’t need a form field to complete the transaction, delete it. If you need it later, ask later. SSH: The site presented “very matter-of-fact statements and not much else.” Restructured content to answer user questions before they had to ask them. Freed up bandwidth so they could evaluate the product instead of deciphering the interface.ClicSmart App: Started with complex user flows, narrowed down to 3 primary navigation tabs. Radical simplification. Layer 1 First-Impression Architecture “Your site has 50 milliseconds. That’s not a design constraint. That’s the physics of how your visitors’ brains work.” Visual evaluation begins within 50ms, faster than conscious thought (Lindgaard et al., 2006; Tuch et al., 2012). Reinecke et al. (2013) at Google Research found effects at exposures as short as 17ms. The aesthetic-usability effect (Kurosu & Kashimura, 1995; replicated cross-culturally by Tractinsky, 1997) shows that attractive interfaces are perceived as more usable, even with identical functionality. People form first impressions on products the same way they do when someone walks into a room. Heider & Simmel (1944) showed people attribute intentions and emotions to geometric shapes. Gray, Gray & Wegner (2007) formalized this as mind perception: Agency (capacity to think/act) and Experience (capacity to feel/sense). Users construct lightweight models of competence and care for websites; enough to pass or fail the same trust evaluation we run on humans. The first impression IS the activation point. It’s the bouncer’s table visit. If it fails, Listen equals zero, and everything multiplies by zero.Visual evaluation begins within 50ms, faster than conscious thought (Lindgaard et al., 2006; Tuch et al., 2012). Reinecke et al. (2013) at Google Research found effects at exposures as short as 17ms. The aesthetic-usability effect (Kurosu & Kashimura, 1995; replicated cross-culturally by Tractinsky, 1997) shows that attractive interfaces are perceived as more usable, even with identical functionality. People form first impressions on products the same way they do when someone walks into a room. Heider & Simmel (1944) showed people attribute intentions and emotions to geometric shapes. Gray, Gray & Wegner (2007) formalized this as mind perception: Agency (capacity to think/act) and Experience (capacity to feel/sense). Users construct lightweight models of competence and care for websites; enough to pass or fail the same trust evaluation we run on humans. The first impression IS the activation point. It’s the bouncer’s table visit. If it fails, Listen equals zero, and everything multiplies by zero. Causal chain: Faces in hero → instant pattern recognition → positive emotional response → trust established before thinking begins In Practice The hero section is the thesis statement of the entire site. If the thesis is wrong, everything below it fails. Faces exploit neonatal face preference (Johnson et al., 1991) and adult face “pop-out” in visual search (Hershler & Hochstein, 2005). Visual quality must match price point. $20–40 tablets in wood cases? I created the perception of “smart home decor” and sold them for $100–200. iO Theater: Hobbyist WordPress site. Cluttered, inconsistent, Google Maps in the hero. “Doesn’t look like a real theater.” Rebuilt with dramatic photography, clear event navigation, Bill Murray quote. Online ticket sales: 50% → 75%. Same shows. Different perception. Layer 2 Processing Fluency “If it’s easy to process, it feels true.” Reber & Schwarz (1999) showed that identical statements in easy-to-read fonts are judged as more true. Alter & Oppenheimer (2009) showed fluency effects generalize across truth, confidence, liking, and trust. Dechêne et al. (2010) confirmed the illusory truth effect across a meta-analysis of 51 studies. Consistency matters more than creativity. Lakoff & Johnson (1980) showed abstract thought is structured by metaphors rooted in bodily experience. Spence (2011) demonstrated cross-modal correspondences: people consistently match high-pitched sounds with small, bright objects. When sensory input is consistent and fluid across channels, the brain reads it as trustworthy. Color consistency operates in perceptual space, not physical space. Near-miss color deviations (a brand blue 3% off) are disproportionately costly because they sit in the steep, high-sensitivity zone of the perceptual metric where prediction error is maximal (Bujack et al., 2022). Perceptually uniform spaces (OKLCH, CAM16-UCS) are L2 infrastructure.Reber & Schwarz (1999) showed that identical statements in easy-to-read fonts are judged as more true. Alter & Oppenheimer (2009) showed fluency effects generalize across truth, confidence, liking, and trust. Dechêne et al. (2010) confirmed the illusory truth effect across a meta-analysis of 51 studies. Consistency matters more than creativity. Lakoff & Johnson (1980) showed abstract thought is structured by metaphors rooted in bodily experience. Spence (2011) demonstrated cross-modal correspondences: people consistently match high-pitched sounds with small, bright objects. When sensory input is consistent and fluid across channels, the brain reads it as trustworthy. Color consistency operates in perceptual space, not physical space. Near-miss color deviations (a brand blue 3% off) are disproportionately costly because they sit in the steep, high-sensitivity zone of the perceptual metric where prediction error is maximal (Bujack et al., 2022). Perceptually uniform spaces (OKLCH, CAM16-UCS) are L2 infrastructure. Causal chain: Unified branding → feels professional → perceived as trustworthy → clears “would I buy from this?” bar SSH 4-year arc: No brand system in 2020. By 2024: comprehensive typography system, seasonal palettes, Disney and Brookstone sub-brand mood boards, ClicSmart and myLuma sub-brands, all with family resemblance. Processing fluency is brand infrastructure that compounds. Revenue 3x. Layer 3 Perception Bias Optimization “It’s not for you. It’s for them.” Users autopilot decisions, then construct rational explanations after the fact (Nisbett & Wilson, 1977). Kahneman (2011): System 1 makes the decision; System 2 constructs the explanation. Survey data captures the System 2 rationalization. Analytics captures the System 1 decision. The gap between them is where perception problems hide. In predictive processing terms (Clark, 2013): design for what the brain predicts, not what users report. What users say in surveys is their System 2 reconstruction. What their brain predicted, and how the design matched or violated that prediction, is what actually drove behavior. Ultimately what works in marketing isn’t “here’s what our product does.” It’s “here’s what having our product makes you feel.” The iPod wasn’t a better product than the Zune. It was about freedom of movement, self-expression, your library in your hands.Users autopilot decisions, then construct rational explanations after the fact (Nisbett & Wilson, 1977). Kahneman (2011): System 1 makes the decision; System 2 constructs the explanation. Survey data captures the System 2 rationalization. Analytics captures the System 1 decision. The gap between them is where perception problems hide. In predictive processing terms (Clark, 2013): design for what the brain predicts, not what users report. What users say in surveys is their System 2 reconstruction. What their brain predicted, and how the design matched or violated that prediction, is what actually drove behavior. Ultimately what works in marketing isn’t “here’s what our product does.” It’s “here’s what having our product makes you feel.” The iPod wasn’t a better product than the Zune. It was about freedom of movement, self-expression, your library in your hands. Causal chain: Emotional tagline → connects to real desire → resonates with target customer → action SSH: Marketing focused on screen resolution, WiFi, app features. Workshopped through “what does this product actually DO for the user emotionally?” Discovery: connection, not technology. Tagline: “Stay connected, even when you’re apart.” Two years later, Aura Frames was using nearly identical messaging. When your competitor steals your tagline, you found the emotional core. Layer 4 Decision Architecture “They’re hunters looking for prey, and it’s my job to make a trail.” Every design is a choice architecture. There is no neutral presentation of options. This builds on Thaler & Sunstein (2008), though a Bayesian re-analysis suggests nudge effects may be near-zero after correcting for publication bias (Maier et al., 2022). The structural principles (defaults matter, framing matters, fewer options at decision points reduce paralysis) remain well-supported. Context qualifier: “Fewer options” applies strongest at high-stakes single decisions and unfamiliar purchase choices. It does NOT apply in expert tool palettes or workspace contexts, where option reduction actively harms productivity. The operative principle: reduce choices at decision points, not choices everywhere. Barkley’s (1997) model adds another dimension for ADHD users: self-regulation failures occur at the point of performance, not at the point of planning. ADHD temporal processing collapses to “now and not now.” Future benefits have near-zero motivational weight. L4 must make the benefit immediate, not promised. Standard decision architecture assumes people weigh future consequences. For ADHD users, and to a lesser degree for everyone under cognitive load or time pressure, the trail must pay off at every step, not just at the destination.Every design is a choice architecture. There is no neutral presentation of options. This builds on Thaler & Sunstein (2008), though a Bayesian re-analysis suggests nudge effects may be near-zero after correcting for publication bias (Maier et al., 2022). The structural principles (defaults matter, framing matters, fewer options at decision points reduce paralysis) remain well-supported. Context qualifier: “Fewer options” applies strongest at high-stakes single decisions and unfamiliar purchase choices. It does NOT apply in expert tool palettes or workspace contexts, where option reduction actively harms productivity. The operative principle: reduce choices at decision points, not choices everywhere. Barkley’s (1997) model adds another dimension for ADHD users: self-regulation failures occur at the point of performance, not at the point of planning. ADHD temporal processing collapses to “now and not now.” Future benefits have near-zero motivational weight. L4 must make the benefit immediate, not promised. Standard decision architecture assumes people weigh future consequences. For ADHD users, and to a lesser degree for everyone under cognitive load or time pressure, the trail must pay off at every step, not just at the destination. Causal chain: Clear path → user finds goal → reduced friction → purchase completed SSH homepage: Before: “didn’t answer any of the user’s questions, goals, or form a good impression.” Navigation reflected the company structure. After: rebuilt as a trail. Emotional hook → understands the product → social proof → clear path to purchase. The Diagnostic How I actually diagnose a site. Not a checklist; a process. Step 1: Feel (Pre-Verbal) Arrive at the page and let the emotional response fire before Wernicke’s area translates it into words. “This makes me feel X.” The feeling comes before the language. System 1 before System 2. My ADHD sensitivity is an advantage here; I feel cognitive load and friction more acutely than most users. The emotional response IS the diagnostic instrument. Step 2: Unpack “Why does this feel X?” Now I’m looking for the systematic rule that’s broken: don’t overload, be specific, get to your point, signal-to-noise ratio, everything has a purpose, this should feel X. Does it? Step 3: Diagnose SignalLikely Layer High bounce + low time-on-siteLayer 2: First impression High time + low conversionLayer 5: Decision architecture (or trust) High add-to-cart + low checkoutLayer 1: Cognitive load (or price perception) Good desktop + poor mobileLayer 3: Processing fluency Stated behavior ≠ actual behaviorLayer 4: Perception bias Step 4: Prescribe Fix the lowest failing layer first. Always. Fixing decision architecture when the first impression is broken is like optimizing a trail that nobody enters. The 5-Minute Perception Audit 1. Path Count (Cognitive Load) Count every decision from landing to purchase. Compare to competitors. Higher count = losing to friction. 2. 5-Second Test (First Impression) Show the homepage for 5 seconds. Can they state purpose, feel trust, credit card confidence? 3. Squint Test (Processing Fluency) Blur your eyes. What stands out? If everything looks equally important, nothing is. 4. Analytics Gap (Perception Bias) Compare what users say vs. what they do. The gap is where problems hide. 5. Trail Test (Decision Architecture) Incognito mode. First-time visitor. Core offer in 3 clicks or search bar? The Generative Protocol The Diagnostic decodes failures. The Generative Protocol derives solutions. Evaluation and generation are different cognitive operations. Evaluating a proposal against PFD catches surface violations. Deriving a solution bottom-up from PFD’s constraints surfaces requirements that no existing proposal addresses. Rule Zero: Do NOT propose any solution until all 5 layers are analyzed and requirements accumulated. Solution-first thinking contaminates the requirement space. You stop seeing what the layers demand and start seeing what the solution provides. State the Design Problem in terms of user experience, not missing features Work Each Layer Bottom-Up. For each: state the constraint, state the violation, derive the MUST requirement, label R[n] Accumulate Requirements. R1–R5, all non-negotiable. Lower layers win conflicts. Derive the Solution. What satisfies R1 AND R2 AND R3 AND R4 AND R5? Test Against Proposals. Which requirements does each proposal fail? Identify the Gap. Requirements no proposal satisfies = where the non-obvious solution lives The Gate Run before shipping. Each layer is pass/fail. If any layer below L4 fails, do not ship. LayerGate QuestionPass Condition L0Can a first-time user complete the primary task without getting stuck?Zero “where do I click?” moments L1Does the first impression activate the right emotion?5-second test: users can state purpose + feel trust L2Is every element consistent with brand and price point?Squint test passes. No rogue fonts, colors, spacing. L3Does the design speak to what users respond to?Copy leads with outcomes. Analytics aligns. L4Does the trail lead to conversion without manipulation?Core offer ≤ 3 clicks. Passes all three ethical tests. The Communication Hierarchy Music, humor, and stories: three modalities for changing perception, ranked by power. Music (Most Powerful) Music communicates emotion and bypasses rational processing entirely. Koelsch (2014) showed music modulates subcortical emotion structures rapidly, potentially prior to conscious evaluation. Juslin & Västfjäll (2008) identified six mechanisms for music-evoked emotion, the majority bypassing conscious processing. In design: typography rhythm, spacing cadence, visual flow, animation timing, scroll momentum. The “music” of a page controls pacing and emotional state. Humor (Second) Jokes change how people feel about a thing by pointing out absurdity and giving permission to think without rigidity. A joke is: setup creates expectation, punch reveals a pattern that either aligns with your self-image or doesn’t, and the anxiety between amplifies the payoff. Laughter is the anxiety being released. Three comedy escalation patterns map to UX: linear repetition (brand consistency), sequential escalation (funnel narrative), exponential escalation (hero → proof → testimonial → CTA). Stories (Third) Green & Brock (2000) demonstrated “narrative transportation,” where absorption into a story produces belief changes. Two meta-analyses confirm (van Laer et al., 2014; Braddock & Dillard, 2016). The story isn’t the events. It’s how someone changes from A to B. The user is the protagonist. “I have a problem” → “This solves my problem.” The hero is the inciting incident. The CTA is the climax. Websites as Relationships Users evaluate websites the same way they evaluate people. Same unconscious process, same red and green flags. Relationship Red FlagWebsite EquivalentMind Perception Doesn’t take appearance seriouslyLow-quality design, broken layoutsLow Agency (incompetent) Lack of confidenceHedging language, unclear value propLow Agency Too verboseWalls of text, burying the leadLow Agency Non-specific / vagueGeneric claims, no proofLow Agency + Low Experience Too good to be trueOver-promising without social proofSuspicious Agency Acts like you’re lucky to be thereSelf-centered messaging, no empathyLow Experience The trust-killer that operates before anyone reads a word: disrespect. A site that doesn’t treat users with respect, their time, their intelligence, their needs, triggers the same rejection as a person who acts like you’re lucky to interact with them. The Responsibility If we’re engineers of thoughts and emotions, we need a Hippocratic oath. The distinction: The designer controls the perception layer: presentation, UX, visual hierarchy, emotional messaging, activation points. The organization controls the value layer: product quality, pricing integrity, fulfillment, service. PFD’s ethical mandate is to align perception with reality. When the product is good and the perception is bad (the Simply Smart Home case) PFD closes the gap upward, bringing perception up to match genuine value. That’s not manipulation. That’s correction. The violation occurs in the opposite direction: inflating perception above reality. The Three-Part Ethical Test The Alignment Test: Does this design bring perception closer to reality, or further from it? (Berdichevsky & Neuenschwander, 1999) The Sincerity Test: If the user fully understood what this design choice does, would they feel served or exploited? Consumers don’t object to persuasion. They object to persuasion perceived as insincere (Friestad & Wright, 1994; Campbell & Kirmani, 2000). The Golden Rule: Would I, as the designer, consent to being influenced by this technique if I were the user? Dark patterns (Brignull, 2010; Gray et al., 2018), nagging, sneaking, forced action, are the cheap laughs of design. They convert short-term and destroy trust. PFD is the expensive laugh: genuine emotional resonance that compounds. Relationship to Value Sensitive Design PFD’s ethical framework sits within the HCI tradition of Value Sensitive Design (Friedman, Kahn & Borning, 2002). Knobel & Bowker (2011) extended this to ask how values get embedded into designed systems, not just studied after the fact, but actively inscribed through design decisions that privilege certain stakeholders, workflows, and outcomes over others. PFD’s perception layer is exactly such an inscription point: every visual hierarchy, every default, every activation point encodes a judgment about whose attention matters and what they should do with it. Where VSD asks broad questions about which values a system embeds, PFD’s ethical mandate is narrower and operational: does this specific design bring perception closer to reality, for this user, in this context? What PFD Doesn’t See Every framework has blind spots: things it makes visible and things it occludes. Knobel’s concept of ontic occlusion (how one representation of reality blocks another from being seen) applies to PFD itself. Naming these isn’t a weakness. It’s a quality criterion. Structural power dynamics. PFD’s ethical test says “align perception with reality.” But it doesn’t interrogate who decides what “reality” is. When SSH’s product page undersold a good product, the “reality” was clear: the product was better than the perception. But what about a company whose product is mediocre and whose “reality” is genuinely ambiguous? The alignment test assumes a knowable ground truth. In practice, “reality” is contested, and the designer choosing which reality to align with is itself a power move that PFD doesn’t surface. Individual vs. collective. PFD models one user’s cognitive journey through the mental waterfall. It doesn’t model what happens when thousands of users are conducted through the same waterfall simultaneously. Individual perception optimization can produce collective harms: engagement loops, attention extraction at scale, homogenization of choice. PFD’s unit of analysis is the user. The effects that emerge at population scale are outside its frame. The fluency trap. Processing fluency is L2’s core mechanism: “if it’s easy to process, it feels true.” That mechanism is value-neutral. It works for honest brands and dishonest ones. A fluent lie is more persuasive than a disfluent truth. PFD’s ethical tests catch deliberate manipulation, but they don’t address the structural problem: by optimizing for fluency, PFD makes it harder for users to engage the critical evaluation that would catch deception. The very thing that makes PFD effective (reducing prediction error, keeping users in autopilot through the desired path) is the same thing that can prevent them from stopping to question. This is ontic occlusion baked into the mechanism. The designer’s own perception biases. PFD’s diagnostic starts with “Feel”: the designer’s pre-verbal emotional response as the primary instrument. But that instrument is calibrated by the designer’s own culture, class, neurotype, and aesthetic history. I name my neurodivergence as an advantage (heightened sensitivity to friction). But heightened sensitivity to friction could also mean under-sensitivity to other things: social warmth cues, cultural signifiers outside my experience, accessibility needs I don’t personally feel. The diagnostic instrument has its own biases, and PFD doesn’t yet have a protocol for calibrating them. Limitations & ongoing research: PFD’s evidence base draws on strong lab research (processing fluency, 50ms impression formation, cognitive load theory) and practitioner observation across 15 years. The five-layer synthesis itself is practitioner-derived, not independently validated as a complete system. Cultural generalizability beyond Western markets remains an open question. I state this because intellectual honesty is a stronger authority signal than overclaiming. The Results Simply Smart Home The problem: Template site. No unified branding. Feature-focused marketing. Premium price point with a presentation that said “knockoff.” Every layer was failing: site didn’t answer basic questions (L0, cognitive load), no faces or warmth (L1), template at premium price (L2), feature specs instead of emotional connection (L3), homepage structured around company, not user goals (L4). What I did: Unified brand system. Emotional messaging: “stay connected, even when you’re apart.” Hero sections with faces. Consistent collateral across every touchpoint. 4-year systematic brand infrastructure build. Revenue 3x within a year Disney licensing, Costco pallets, Walmart shelf space Competitor stole the tagline two years later iO Theater “Doesn’t look like a real theater.” Full redesign: dramatic photography, clear event navigation, integrated payment. Online ticket sales 50% → 75%. Same shows. Different perception. VacuumSealersUnlimited Embedded design partnership, 5+ years. Revenue 4x running on infrastructure I built and maintain. The long game. Perception compounds. What’s Teachable 3 Hours with Business Owners Hour 1: Business Design. Your business IS a design. Feedback loops between your vertical, users, and staff. I’d give them the tools to see their business as a system. Hour 2: People as Research. Your staff and clients are your greatest resources. Design problem-solving is really just a lens of empathy. “For whom is this?” gives you directionality that frameworks can’t. Hour 3: Tactics. The 5-Minute Perception Audit. The dependency stack. The red flags. Application, communication, process. The Teachability Line The lens (empathy, for-whom, design-as-problem-solving) is fully teachable. The pattern library (hundreds of projects of instant recognition) takes years. The sensitivity (neurodivergent tuning fork for bad design) may not be teachable. But the diagnostic process (Feel → Unpack → Diagnose → Prescribe) can be taught as a substitute. The Improv Rules My design practice runs on principles I learned from improv classes. These aren’t metaphors. They’re operational tools. Improv RuleDesign Application Yes, andAccept what exists and build on it. Brick by brick, not castle at a time. (Listen) × (Act + React)If you’re not listening, you multiply everything by zero. Same for users. “If this is true, what else is true?”One insight cascades. If this is the brand voice here, it should be everywhere. “What comes before / after?”User journey thinking. What preceded this page? What follows? “It’s not for you, it’s for them”Protect the design from stakeholder preferences. Rule of eventually“That’s horri-bly... good!” Subvert expectations for engagement. Hold topics and returnProgressive disclosure. Introduce, develop elsewhere, callback at conversion. “Familiar to you, novel to others”Design for their first encounter, not your hundredth. The Dual Foundation PFD’s original 55 citations come from perception psychology: cognitive load theory, processing fluency, decision architecture, predictive processing. The Inverse Inflection research (2026) added 22 citations from ADHD cognitive neuroscience. Zero overlap between the two bodies. 77 citations total, drawn from two non-overlapping literatures, converging on the same design principles from opposite directions. The Connection Why do two separate literatures point to the same principles? Because they describe the same cognitive architecture from different angles. Perception psychology measures how the typical brain processes input: working memory limits, first-impression timing, fluency effects. ADHD neuroscience measures what happens when those same systems operate under tighter constraints: lower working memory ceiling, higher friction sensitivity, sharper temporal discounting. Same architecture, different parameters. The five PFD layers map cleanly to ADHD neuroscience findings. Every layer has independent support from the ADHD literature, and the support is strongest at the layers that matter most: cognitive load (L0) and Processing Fluency (L2). PFD-ADHD Layer Mapping PFD LayerADHD Neuroscience SupportEvidence TierKey Citations Cognitive Load (L0)WM deficits d=0.69–0.74. Context switching costs approach an hour for ADHD developers. Low perceptual load specifically benefits ADHD-symptomatic developers.Tier 1–2 (Strong)Meta-analyses; ICSE 2024; JetBrains 2023 L1Reduced latent inhibition means more sensory intake: a badly designed first impression hits harder. Wall of Awful: bad first impressions deposit emotional barriers. Rejection Sensitive Dysphoria amplifies mild negative affect into disproportionate avoidance.Tier 3 (Indirect)Carson et al. 2003; Mahan 2020; Dodson L2Low perceptual load benefits ADHD developers. Flat associative hierarchies amplify pattern-break sensitivity. Neurodivergent users perceive LLM outputs as “overly neurotypical”: what is fluent for NT may be disfluent for ND.Tier 1–2 (Strongest extension)JetBrains 2023; Mednick 1962; Carik et al. 2024 L3The METR perception gap: developers thought they were 20% faster with AI but were 19% slower. Overconfidence amplification. The 9% competence penalty for visible AI assistance.Tier 2–3METR 2025 L4Barkley’s “point of performance”: self-regulation failures occur in the moment, not in planning. “Now and Not Now” temporal discounting means future benefits have near-zero motivational weight.Tier 1Barkley 1997 The Curb-Cut Argument PFD was not designed for ADHD. But because ADHD involves the most constrained version of the cognitive architecture PFD designs for (lower working memory ceiling, higher friction sensitivity, sharper temporal discounting), designs that work for ADHD users satisfy the tightest constraints. Everyone benefits. The principle is the same one behind curb cuts, closed captions, and automatic doors. Curb cuts were designed for wheelchair users but turned out to help parents with strollers, travelers with luggage, delivery workers with carts. Designs that reduce cognitive load for ADHD users reduce cognitive load for all users. The constrained case reveals what the general case tolerates but shouldn’t. Cowan (2010) established that working memory holds 3–5 chunks. ADHD brains hit the ceiling of that range faster and harder, making violations visible sooner. A page that overloads an ADHD user’s working memory is overloading everyone’s. The ADHD user just notices first. The research supports this directly. Kasatskii et al. (JetBrains, 2023) found that high perceptual load degrades ADHD-symptomatic developers’ performance faster than neurotypical developers’. Mark et al. (2008) measured context-switching recovery at 23 minutes for typical workers. For ADHD developers, Mark et al. (2024) found the cost approaches an hour. Same friction, amplified. Design for the amplified case and the typical case improves too. The Diagnostic Instrument I’m autistic and have severe combined-type ADHD. The autism means I learned social cognition analytically: theory, practice, application loops. I can see and articulate unconscious patterns that most designers apply intuitively but can’t explain. The ADHD means cognitive load and friction are literally painful for me. What most designers theorize about, I feel. That sensitivity is a diagnostic instrument. When I arrive at a page, the emotional response fires before language catches up. System 1 before System 2. My ADHD sensitivity means I feel cognitive load and disfluency more acutely than most users. The feeling IS the diagnostic. The autism means I can then unpack exactly why, tracing the violation back to the systematic rule that’s broken. I’ve come to think of this as a curb-cut effect applied to cognition. The cognitive constraints I designed around (limited working memory, sensitivity to disfluency, impatience with unnecessary friction) are constraints every user has. Mine are just louder. The L0 Refinement Dodson’s Interest-Based Nervous System (IBNS) proposes that ADHD attention is regulated by interest, not importance. The PINCH model identifies five drivers that override cognitive load ceilings: Passion, Interest, Novelty, Competition, Humor. This means L0’s “reduce load” is incomplete. For ADHD users (and partially for everyone under low motivation), the correct formulation is “reduce load OR increase interest.” This is not a contradiction of L0. It is the specification of when load reduction alone is insufficient. When interest is absent, no amount of cognitive streamlining produces engagement. On the dual citation base: The 55 perception citations and 22 ADHD citations share zero duplicate entries. Bridging them into a single actionable design framework is this methodology’s contribution. The underlying cognitive architecture is expected to generalize. The ADHD-specific parameters identify where the architecture operates under the tightest constraints. Cognitive Constraint Design The Inverse Inflection research surfaced a unifying meta-framework: Cognitive Constraint Design. The principle: cognitive constraints are design material, not obstacles to work around. Most design approaches treat cognitive limitations as problems. Reduce choices because people can’t handle too many. Simplify because users are overwhelmed. The framing is defensive: protect the user from their own brain. CCD inverts this. The constraint is the design input. Working memory holds 3–5 chunks? That’s not a limitation to accommodate. That’s a specification to build for. The constraint tells you what to build, not what to avoid. DomainConstraintDesign Response PFD (interfaces)Limited processing capacityReduce cognitive load, optimize fluency ADHD-AI workflowsExecutive dysfunctionExternalize structure at point of performance Spatial workflow toolsWorking memory limitsSpatial triggers externalize planning AI writing assistantsCommunication pattern driftCorrections-as-training preserves authentic voice Session memoryContext loss between sessionsHierarchical persistence survives crashes Same principle, different surfaces. PFD applies Cognitive Constraint Design to interfaces. The five layers (L0 through L4) are a systematic treatment of perceptual and cognitive constraints as design specifications. Each layer names a constraint (bandwidth, first impression, fluency, bias, decision load) and prescribes a design response grounded in how that constraint actually operates. CCD is the genus. PFD is a species. This distinction matters because it explains why PFD’s approach transfers across domains. The bouncer story, the improv rules, the diagnostic protocol: they all work because they’re applications of the same principle. Observe the constraint, design for it, test whether the design honors it. The methodology page focuses on PFD as a design tool. This reference documents the broader intellectual framework. CCD is for researchers and practitioners who want to understand why PFD’s approach to interfaces also explains how to build session memory systems, voice preservation tools, and spatial orchestration canvases. The connecting thread is always: what constraint does the user face, and how do you design the system so the constraint becomes productive rather than destructive? Practitioner Cognition This section explains HOW PFD’s diagnostic produces insights that standard UX heuristics miss. It is not a different checklist. It is a different cognitive operation. Bisociation Koestler (1964) coined bisociation for the creative act of connecting two independent frames that don’t normally intersect. Standard association operates within a single frame. Bisociation jumps between frames to produce something neither frame contains alone. PFD’s cross-domain structure (psychology + design + improv + neuroscience) is bisociative by architecture. The derivation protocol forces collision between constraint frames. When you work Layer 0 (cognitive load theory) and then Layer 3 (perception bias), the accumulated requirements R1 through R3 create a constraint space that neither field produces independently. The non-obvious solution lives in that collision. Janusian Thinking Rothenberg (1971) described Janusian thinking: the simultaneous conception of opposites as a creative process. PFD holds “reduce friction” AND “activate attention” simultaneously. The core thesis is literally Janusian: users don’t think, until you make them. Don’t make them think (L0). Make them think (activation point). Both are true. The resolution is not choosing between them but designing for the transition. Deliberate and Spontaneous Modes Dietrich (2004) mapped four creativity types to prefrontal function: deliberate/spontaneous crossed with cognitive/emotional. Creative insight involves both deliberate (prefrontal-guided, analytical) and spontaneous (defocused, associative) processing. The PFD diagnostic protocol alternates between these modes. The Feel step is spontaneous: pre-verbal, emotional, System 1. The Unpack and Diagnose steps are deliberate: analytical, systematic, tracing violations back to specific layers. The protocol is not “first feel, then think.” It is “alternate between modes, and use each mode for what it does best.” This is why the Feel step cannot be skipped. A pure checklist audit (deliberate mode only) catches surface violations but misses the emotional gestalt. A pure intuitive response (spontaneous mode only) identifies that something is wrong but cannot diagnose which layer is failing. PFD needs both. Flat Associative Hierarchies Mednick (1962) showed that creative people have flatter associative hierarchies: more distant conceptual connections are accessible. Where most people associate “table” with “chair” and stop, creative thinkers reach “table” to “periodic table” to “elemental structure” to “information architecture.” Carson, Peterson & Higgins (2003) found eminent creative achievers were 7x more likely to have low latent inhibition: their brains let in more raw sensory and conceptual material. ADHD may produce flatter hierarchies via reduced latent inhibition through related mechanisms. The connection is indirect but convergent: ADHD involves reduced inhibitory filtering, creativity correlates with reduced inhibitory filtering, and PFD’s cross-domain derivation protocol rewards exactly the kind of distant associations that flat hierarchies produce. The Communication Hierarchy Three modalities for changing perception, ranked by power: Music > Humor > Stories. Music communicates emotion and bypasses rational processing entirely (Koelsch, 2014; Juslin & Västfjäll, 2008; Huron, 2006). In design: typography rhythm, spacing cadence, visual flow, animation timing, scroll momentum. Humor changes how people feel about a thing by pointing out absurdity and giving permission to think without rigidity. Stories produce narrative transportation (Green & Brock, 2000; van Laer et al., 2014): absorption into a narrative produces belief changes. The practitioner who understands all three modalities has more tools for changing perception than the practitioner who only understands information architecture. This is why PFD draws from improv, music theory, and narrative psychology alongside cognitive science. The communication hierarchy is not a bonus chapter. It is how the methodology gets delivered. Evidence Tiers Not all claims carry equal weight. This classification system governs what PFD states confidently, what it attributes carefully, what it frames as synthesis, and what it avoids. Every public claim should be traceable to one of these tiers. Tier 1: Established Science Well-replicated findings from peer-reviewed research, confirmed by multiple independent labs. PFD states these as facts. FindingSource(s)Why Established Working memory deficits in ADHD (d=0.69–0.74)Meta-analyses, Barkley modelDecades of convergent evidence, hundreds of studies ADHD associated with stronger divergent thinkingWhite & Shah 2006/2011, Boot et al.Multiple independent labs, peer-reviewed, conceptual replications DMN hyperconnectivity in ADHDCastellanos et al.One of the most replicated neuroimaging findings Creative thinking requires DMN + ECN + Salience cooperationBeaty et al. (Three-Network Model)Multiple fMRI studies, established in cognitive neuroscience 7x rate of low latent inhibition in creative achieversCarson, Peterson & Higgins 2003JPSP (top-tier), widely cited Flat associative hierarchies produce remote associationsMednick 196260+ years of citation, standard creativity psychology DRD4/7R: opposite fitness outcomes by environmentEisenberg 2008Well-designed field study, conceptually replicated Cognition extends into reliably accessed external toolsClark & Chalmers 1998One of the most cited papers in philosophy of mind Context switching costs significant recovery timeGloria Mark (multiple), ICSE 2024Well-established productivity research Self-regulation, not attention, is the core ADHD deficitBarkley modelDominant ADHD theoretical framework, decades of research Public framing: “Research shows...” or state directly without hedging. Tier 2: Strong Single Studies Solid methodology, credible venues, but not yet independently replicated. PFD cites the specific source. FindingSourceLimitation ADHD is the only positive predictor of divergent thinking in engineering studentsTaylor 2020 (n=60)Single university, single lab Emotional prompting improves LLM output ~11%Li et al. 2023 (EmotionPrompt)ArXiv preprint; 11% human eval is more defensible than 115% peak 25% higher AI satisfaction among neurodivergent workersUK Dept. for Business and TradeMeasures satisfaction, not performance ND users perceive LLM outputs as “overly neurotypical”Carik et al. 2024Peer-reviewed (ACM), specific study context Low perceptual load benefits ADHD-symptomatic developersJetBrains 2023Industry research, single study Public framing: “A study by [X] found...” or “Research from [X] suggests...” Always name the source. Tier 3: Reasonable Inferences Logical connections drawn between established findings. The foundations are real. The bridge is untested. PFD presents these as synthesis, not established research. InferenceFoundationUntested Bridge ADHD communication style is natural emotional promptingEmotionPrompt + IBNS (Dodson)Nobody has tested whether ADHD individuals produce better prompts ADHD-AI error patterns are complementaryPNAS 2022 Bayesian frameworkZero studies measuring actual ADHD-AI error correlation Centaur model generalizes beyond chessKasparov/ZackS 2005Chess has perfect information. Software development does not. PFD disproportionately benefits ADHD usersWM deficits + L0/L2 researchPFD was not designed for ADHD. Connection is real but unmeasured. Public framing: “Our synthesis suggests...” or “The evidence points toward, though this hasn’t been directly tested...” Tier 4: Speculative or Corporate Claims that lack methodology, independent verification, or peer review. PFD either avoids these or flags them explicitly as unverified. ClaimSourceProblem JPMorgan ND program: 90–140% higher productivityCorporate self-reportNo sample size, no comparison group, no methodology published Deloitte “30% more productive” ND teamsIndustry reportNo peer review, no methodological detail The “100+ sources” claim from popular ADHD-AI sitesSelf-reportCounts blog posts alongside peer-reviewed papers Public framing: Do not use, or explicitly state: “Corporate reports claim...” Never present as research. The governing principle: Credibility compounds. Every well-hedged claim builds trust. One overstated claim can undo it. The advantage of reading the primary literature is being the person who gets it right, not the person who gets caught overclaiming. Accumulated Learnings 18 learnings from applying PFD across interfaces, spatial canvases, AI systems, pricing, ethics, and communication. These are not theoretical extensions. They emerged from practice and changed how the framework is applied. #LearningDomainPFD Layer(s) 1Generative vs. EvaluativeCore methodAll 2Access vs. Signal MechanismsMarketing / UXL1–L4 3Pre-Send vs. Post-ResponseAI / CommunicationL1, L4 4Workspace vs. Product SeparationProduct designAll 5Recursive ValidationMeta-methodologyAll 6Infrastructure ≠ ActivationProduct designL1, L4 7Consistent Gap IdentificationCore methodL2 8Epistemic Asymmetry as L4 ProblemAI systemsL3, L4 9Temporal/Session ContinuitySpatial toolsGap (candidate L6) 10Constraints Are DistributionsAccessibilityL0 11Visual Channel Utilization AuditData viz / CanvasL2 12Route vs. Survey KnowledgeSpatial toolsL2, L4 13Keyboard Density Is L2, Not L4Expert workspacesL2 14Ethics Are Operational, Not StructuralEthicsEthics 15Experiential Self-ContradictionHero / landingL2, L3, L4 16Near-Miss Color AsymmetryDesign systemsL2 17Iterative Regression Is Deeper VisibilityForge / analysisAll 18Backend Mechanics as Frontend TiersPricing / SaaSL1–L4 #1: Generative vs. Evaluative When analysts were given solutions first and asked to evaluate through PFD, they missed requirements entirely. When forced to derive bottom-up, the gaps emerged. PFD only generates non-obvious solutions when used bottom-up with constraint accumulation. Top-down evaluation refines but cannot escape the originating frame of the solution being evaluated. The derivation protocol exists because evaluation is not generation. #2: Access vs. Signal Mechanisms PFD’s layers distinguish between “how do I access content?” and “how does the user know where content exists?” L1/L2 handle signaling. L3/L4 handle access. Conflating them produces solutions that solve access without solving discovery. A CTA that’s easy to click (access) is different from a page that makes the CTA feel like the obvious next step (signal). #3: Pre-Send vs. Post-Response Standard UX says “show your sources after the response” (attribution). PFD derivation says “show what the AI will see before the user sends” (legibility). Attribution is an L3 debugging tool. Pre-send legibility is an L1+L4 schema and trust tool. What you show BEFORE the ask is higher-leverage than what you show AFTER. This generalizes to communication and copy: setup and framing outweigh explanation and justification. #4: Workspace vs. Product Separation When analyzing a specific workspace through PFD, recommendations naturally split between workspace-level fixes (rearrange, rename, add zones) and product-level features the tool should provide. The separation between “user organizes content” and “tool provides perceptual infrastructure” is a critical design boundary. PFD should be applied at both levels, but the intervention target must be identified correctly. #5: Recursive Validation When PFD is used to analyze a workspace that is ABOUT PFD, three levels operate simultaneously: PFD as framework, PFD applied to the workspace, the workspace as proof-of-concept for the tool. When all three levels are legible, the workspace becomes its own best case study. This recursive structure is a strength, not a liability, but only when intentionally designed. #6: Infrastructure ≠ Activation Infrastructure that exists without activation is worse than no infrastructure. It creates the illusion of completeness while users experience the original problem. PFD correctly distinguishes between “the mechanism exists” and “the user can perceive and use the mechanism.” This is a Layer 4 violation masquerading as a Layer 1 problem. Having the right product without the right perception layer produces the same failure mode. #7: Consistent Gap Identification When the same requirement (R2: pre-attentive content signaling) surfaced independently across four different analytical frames, it was not a preference. It was a genuine perceptual constraint. When running multiple derivations on the same problem, note which requirements appear consistently. Those are load-bearing. #8: Epistemic Asymmetry as L4 Problem When an AI system gathers context via invisible mechanisms (graph traversal, priority weighting, depth limits), the resulting epistemic asymmetry (AI knows more than user about its own context) is a Layer 3 violation. Users’ System 1 response to opaque behavior is distrust. The intervention is making the mechanism visible pre-send, not explaining it post-response. This generalizes to any expert/client asymmetry: show your reasoning process, not just your conclusions. #9: Temporal/Session Continuity PFD addresses what the user perceives NOW but not how they re-enter a workspace after 48 hours. Six of nine independent analytical personas identified this gap. A workspace visited after two days is a different perceptual object than the same workspace visited five minutes ago. PFD’s five layers assume continuous interaction and need supplementation for interrupted/resumed sessions. Candidate for formal L6 after testing. #10: Constraints Are Distributions, Not Constants PFD states working memory holds “~3–5 chunks” as a hard constraint. This is a population mean. Neurodivergent users, users under cognitive load, users with fatigue all shift the distribution. The constraint is real but the threshold varies. PFD’s constraints define the shape of the curve, not a fixed cutoff. Good design provides calibration mechanisms for users at the tails. This strengthens PFD by acknowledging that perception science describes populations, not individuals. #11: Visual Channel Utilization Audit Bertin’s 7 visual variables (position, size, lightness, texture, color, orientation, shape) provide a systematic audit checklist for design. Most UIs use 2–3 channels. Under-utilized channels mean information that COULD be signaled pre-attentively is hidden behind interaction. The derivation protocol should include a channel audit: “How many visual channels does this design use? Which are available but unused?” #12: Route vs. Survey Knowledge Siegel & White’s spatial knowledge theory: users develop route knowledge (sequential, egocentric, fragile) before survey knowledge (allocentric, configurational, robust). Canvas tools should support both stages. New users need route affordances (guided paths, breadcrumbs, landmarks). Experienced users need survey affordances (minimap, overview, configurable views). “Appropriate density” depends on which knowledge type the user has developed. #13: Keyboard Density Is L2, Not L4 When multiple analysts flagged “too many keyboard shortcuts” in an expert tool, the initial diagnosis was L4 choice overload. After applying the Scheibehenne (2010) context qualifier (choice overload is near-zero in expert workspaces), the problem reframed to L2: users cannot predict what a keypress will do across cognitive modes. The fix is a keyboard mode indicator, not fewer shortcuts. Don’t reduce expert affordances. Signal which mode they operate in. #14: Ethics Are Operational, Not Structural PFD’s three ethical tests (Alignment, Sincerity, Golden Rule) are operational ethics: executable by one designer in a single review session. They don’t constitute structural ethics in the Value Sensitive Design sense (systematic processes that produce good outcomes regardless of individual practitioner values). Long-term client relationship accountability enforces alignment better than any framework clause could. Operational tests handle decision-making. Structural incentives handle accountability. When PFD scales beyond solo practice, structural ethics needs to be baked into the process, not just carried by the practitioner. #15: Experiential Self-Contradiction When a hero section’s thesis claims speed and simplicity (“Your customers decide in milliseconds. I design for that.”) but the hero itself takes 10+ seconds to parse due to competing visual elements, it disproves its own claim experientially. This is an L3/L4 failure, not an L1 failure: the visitor’s lived experience contradicts the promise being made. The resolution preserves complexity in the background (competence signal) while simplifying the foreground (single message, single CTA). A diamond on crushed velvet. The complexity is the bed, not the message. #16: Near-Miss Color Asymmetry Human color perception is non-additive (Bujack et al., 2022, PNAS). A brand blue that’s 3% off-target imposes disproportionate processing cost compared to a completely different color, because small deviations sit in the steep, high-sensitivity zone of the perceptual metric where prediction error is maximal. Near-miss color deviations are the highest-cost L2 violations, more disruptive than using a completely different color. Design-token linting should weight near-misses as more severe than far-misses. Color structure is hardware (universal geometry). Color meaning is software (cultural, context-dependent). #17: Iterative Regression Is Deeper Visibility When iterative PFD analysis shows a score DROP around rounds 3–6, the cause is almost never that fixes made things worse. Resolving surface-level issues (nav overload, font proliferation) unmasks deeper problems that were previously invisible. The dominant signal was drowning out subtler violations. The regression reflects updated understanding, not degraded quality. Three terminal states for iterative analysis: converged (done), impasse (approach won’t converge with current premise), abandoned (user stopped). The impasse state is the highest-value signal most iterative tools lack. #18: Backend Mechanics as Frontend Tiers When internal system state (subscription tiers, feature flags, scarcity counters) is surfaced as visible UI complexity, it converts a backend concern into a frontend cognitive load violation. Founding member pricing with counter badges, tier comparisons, and crossed-out prices triggers defensive pattern-matching against growth hacks. The user’s 50ms L1 read is “this is a marketing tactic,” which directly contradicts the intended signal of gratitude. The system needs tiers. The user needs a yes/no decision. These are different design surfaces at different layers. A rare 4-layer simultaneous violation from a single UI element. Framework Comparison PFD exists in a landscape of design frameworks. Understanding where it overlaps and where it differs helps practitioners decide when to use it and when something else serves better. FrameworkWhat It DoesWhere PFD Differs Laws of UX (Yablonski)Collection of psychological principles for designDescriptive catalog, not generative protocol. No dependency stack. No derivation method. Tells you what to know, not what to do first. Fogg Behavior ModelBehavior = Motivation + Ability + PromptSingle-event model. PFD addresses the perceptual pipeline BEFORE behavior triggers. Fogg assumes the user is already attending. Don’t Make Me Think (Krug)Reduce friction, make things obviousPFD’s Cognitive Load layer (L0). Krug covers L0 deliberately and well. PFD adds four more layers on top, plus a generative protocol and ethics framework. Emotional Design (Norman)Visceral, behavioral, reflective processing levelsComplementary framing. Norman’s visceral ≈ PFD L1. PFD provides a dependency stack and generative protocol Norman doesn’t. Nudge Theory (Thaler/Sunstein)Choice architecture shapes behaviorPFD L4. But Maier et al. (2022) shows nudge effects near-zero after correcting for publication bias. PFD contextualizes when choice architecture works and when it doesn’t. Cognitive Constraint DesignMeta-framework: constraints as design materialPFD is an application of CCD to interface design. CCD is the genus. PFD is a species. None of these frameworks are wrong. They are incomplete in different ways. Laws of UX catalogs principles without a generative method. Fogg models behavior but not perception. Krug covers L0 without the other four layers. Norman identifies processing levels without prescribing an order. Nudge Theory applies L4 without the dependency stack that determines when L4 interventions actually work. PFD’s contribution is the dependency stack: five layers in a specific order, where each layer depends on the layer below it. Fix them in order, because upstream failures degrade everything downstream. This ordering principle is what separates PFD from frameworks that treat design principles as independent, parallel concerns. They are not parallel. They are stacked. Make Me Think: Chapter Mapping “Make Me Think: Perception-First Design for the Post-Usability Era” is a 12-part series that develops the framework through narrative, case studies, and citations. Each chapter maps to specific PFD layers and concepts. Ch.TitlePFD Layer(s)What It Adds 1The BouncerOrigin (all layers)The nightclub story: how a security job became a design methodology 2They’re Already Not ThinkingCore thesisPredictive processing. Why most problems happen before conscious evaluation. 3The Tuning ForkDual foundationAutism as analytical social cognition. ADHD as friction sensitivity. 4Cognitive LoadCognitive Load (L0)Miller vs. Cowan, Sweller’s CLT, progressive disclosure, smart defaults. 5The 50-Millisecond VerdictL1First impressions, the iO Theater case study, visual-price coherence. 6The Feeling of TruthL2Processing fluency, the McGurk Effect, consistency compounding trust. 7The GapL3Perception bias. System 1/System 2 gap. Why a tagline change preceded revenue tripling. 8The TrailL4Decision architecture, path reduction, smart defaults. 9Feel, Unpack, Diagnose, PrescribeDiagnostic protocolEmotional primacy, embodied cognition, why feelings arrive before language. 10Music, Humor, StoriesCommunication hierarchyThree tools for changing perception, ranked by power. 11The OathEthics frameworkThree operational tests, structural ethics, cognitive deskilling. 12What I Don’t Know YetLimitationsWEIRD evidence base, the fluency trap, ontic occlusion, living methodology. The series is available at aurochs.agency/writing/make-me-think/. Each chapter stands alone. Read them in any order, though the dependency stack (Chapters 4–8) makes more sense sequentially. Open Questions These are genuinely unresolved. A living methodology names its edges. Cultural generalizability. The evidence base is predominantly WEIRD (Western, Educated, Industrialized, Rich, Democratic). The underlying cognitive architecture (predictive processing, working memory limits, first-impression formation) is expected to generalize. But the parameters (which visual patterns signal trust, what counts as “too much” cognitive load, how fluency maps to credibility) are culturally shaped. PFD has been applied in Western e-commerce and entertainment. How the 5-layer dependency order holds in collectivist markets, non-Latin typographic traditions, or markets with fundamentally different trust signals is an open question. Measurement. PFD lacks a validated measurement instrument. The 5-Minute Perception Audit is a practitioner heuristic: useful, fast, and directionally correct, but without inter-rater reliability data. A diagnostic tool where two PFD practitioners independently audit the same site and converge on the same layer diagnosis would strengthen the methodology from “expert intuition with a framework” to “teachable, repeatable diagnostic.” This is the gap between a clinical art and a clinical science. Temporal dynamics. PFD describes the first encounter well: activation, first impression, mental waterfall. But user relationships with products evolve. At visit 100, the mental waterfall is different. Prediction errors that activated on first visit are now expected. Fluency that felt trustworthy can become invisible (users stop noticing). Does the 5-layer stack apply the same way to retention and long-term engagement as it does to acquisition? The current framework is weighted toward first contact. Accessibility as justice vs. fluency. PFD frames accessibility through cognitive load (L0, cognitive load reduction) and L2 (processing fluency). This produces good accessible design in practice. But disability justice scholars argue accessibility is a rights issue, not an optimization target. Framing it through “fluency” risks treating accessibility as something you optimize when the ROI justifies it, rather than something you guarantee because people deserve it. The framework’s utilitarian lens may not fully serve this. AI-generated design. As AI tools generate more design artifacts, PFD’s emphasis on embodied sensitivity (the “neurodivergent tuning fork,” the pre-verbal Feel step) becomes harder to operationalize at scale. Can PFD’s diagnostic be automated? Can an AI run the Feel step? If not, the methodology has a ceiling on how far it scales beyond the individual practitioner. If yes, what’s lost when the diagnostic instrument is no longer a human body? Appendix A: The Science 78 citations across cognitive psychology, HCI, behavioral economics, design theory, STS, music cognition, narrative psychology, and ADHD cognitive neuroscience. CitationYearFindingLayer Heider & Simmel1944People attribute intentions/emotions to geometric shapesL1 Miller, G.A.1956“Magical Number Seven”: rhetorical, not literalL0 McGurk & MacDonald1976Visual lip movements integrate with auditory perceptionL2 Nisbett & Wilson1977People can’t accurately report why they made decisionsL3 Lakoff & Johnson1980Abstract thought structured by bodily metaphorsL2 Sweller1988Cognitive Load Theory (intrinsic vs. extraneous)L0 Johnson et al.1991Neonatal face preferenceL1 Sloboda1991Musical features trigger involuntary physical responsesComm. Friestad & Wright1994Persuasion Knowledge Model: users object to insincerityEthics Kurosu & Kashimura1995Aesthetic-usability effectL1 Salthouse1996Processing speed declines continuously from early adulthoodL0 Tractinsky1997Aesthetic-usability cross-cultural replicationL1 Sweller, van Merriënboer & Paas1998Formalized CLT taxonomyL0 Berdichevsky & Neuenschwander1999Golden Rule of persuasionEthics Reber & Schwarz1999Processing fluency → perceived truthL2 Campbell & Kirmani2000Sincerity as function of perceived ulterior motivesEthics Green & Brock2000Narrative transportation → belief changeComm. Tractinsky, Katz & Ikar2000“What is Beautiful is Usable”L1 Cowan2001Working memory revised to ~4 chunksL0 Winkielman & Cacioppo2001Fluency generates positive affect without awarenessL2, L3 Friedman, Kahn & Borning2002Value Sensitive DesignEthics Johnson & Goldstein2003“Do defaults save lives?”L4 Norman2004Emotional Design (visceral, behavioral, reflective)Bonus Reber, Schwarz & Winkielman2004Processing fluency → aesthetic pleasureL2 Hershler & Hochstein2005Adult face “pop-out” in visual searchL1 Huron2006ITPRA theory: prediction/reaction are pre-consciousComm. Lindgaard et al.200650ms visual appeal judgmentsL1 Tractinsky, Cokhavi et al.2006Consistent aesthetic judgments at brief exposuresL1 Gray, Gray & Wegner2007Two dimensions of mind perception: Agency + ExperienceL1 Juslin & Västfjäll2008Six mechanisms for music-evoked emotionComm. Song & Schwarz2008Fluency and detection of misleading questionsL2 Thaler & Sunstein2008Choice architecture / Nudge theoryL4 Alter & Oppenheimer2009Fluency generalizes across truth, confidence, trustL2 Brignull2010Dark patterns taxonomyEthics Cowan2010“The Magical Mystery Four”: WM refined to 3–5 chunksL0 Dechêne et al.2010Illusory truth effect meta-analysis (51 studies)L2 Friston2010Free-energy principleCore Henrich, Heine & Norenzayan2010WEIRD caveat for benchmark generalizabilityNote Scheibehenne, Greifeneder & Todd2010Choice overload effect near-zero in many conditionsL4 Kahneman2011Thinking, Fast and Slow: System 1 / System 2Core Knobel & Bowker2011“Values in Design”: values inscribed through design decisions (CACM)Ethics Owsley2011Age-related declines in visual processingL0 Spence2011Cross-modal correspondencesL2 Tuch et al.2012Replication of 50ms visual appealL1 Clark2013Predictive processing: brains match input to top-down predictionsCore Hohwy2013The Predictive MindCore Reinecke et al.2013Visual evaluation at 17ms (Google Research)L1 Koelsch2014Music modulates subcortical emotion structures (Nature Reviews Neuroscience)Comm. van Laer et al.2014Narrative transportation meta-analysis (132 effect sizes)Comm. Zak2015Narrative arcs trigger cortisol + oxytocin (mechanism undemonstrated)Comm. Braddock & Dillard2016Narrative persuasion meta-analysisComm. Gray et al.2018Dark patterns taxonomy: nagging, obstruction, sneakingEthics Sweller, van Merriënboer & Paas2019CLT reconceptualizedL0 Mierop et al.2020Intranasal oxytocin: ~90% false positive rateNote Bujack et al.2022Perceptual color space is non-Riemannian (PNAS)L2 Maier et al.2022Nudge effects near-zero after publication bias correctionL4 ADHD Cognitive Neuroscience Barkley, R.1997Self-regulation, not attention, is the core ADHD deficit. Executive function failures occur at the point of performance.L0, L4 Castellanos, F.X.2002ADHD involves altered default mode network dynamicsL0 Carson, Peterson & Higgins20037x rate of low latent inhibition in high-functioning creative achieversL2 Dodson, W.2005Interest-Based Nervous System: ADHD motivation responds to PINCH (Passion, Interest, Novelty, Competition, Hyperurgency)L0, L3 White, H. & Shah, P.2006ADHD associated with stronger divergent thinking on multiple measuresL2 White, H. & Shah, P.2011ADHD linked to higher real-world creative achievementL2 Kounios & Beeman2009Pre-problem brain state predicts whether solution arrives via insight or analysisL0 Beaty et al.2014Three-network model of creative cognition (default + executive + salience)L0 Mahan, B. (Wall of Awful)2020Emotional barrier accumulated from repeated executive function failuresL4 Taylor, J.2020Twice-exceptional individuals in professional settingsL0 Eisenberg et al.2008DRD4/7R allele associated with better nutritional status in nomadic Ariaal pastoralists but worse in settled — opposite fitness outcomes by environmentL2/L0 Koestler, A.1964Bisociation: the creative act of holding two independent, normally incompatible frames of reference simultaneouslyL0 Mednick, S.A.1962Flat associative hierarchies produce more remote associations; basis for the Remote Associates Test (RAT)L2 Dietrich, A.2004Four creativity types mapped to prefrontal function: deliberate/spontaneous x cognitive/emotionalL0 Clark, A. & Chalmers, D.1998Extended Mind Thesis: cognitive processes extend beyond brain/body when external tools reliably couple with internal cognitionCore Rothenberg, A.1971Janusian thinking: actively conceiving two or more opposite concepts simultaneously as a creative processL0 Kasparov, G.2017Centaur model: weak human + machine + better process outperforms strong computer alone (Deep Thinking, PublicAffairs)L4 Mark, G. et al.2008Cost of interrupted work: average 23 minutes 15 seconds recovery time after interruption (CHI 2008)L0 Steyvers, M. et al.2022Bayesian human-AI complementarity: emerges when human and AI error patterns are sufficiently independent, even when AI outperforms humans individually (PNAS)L4 Li, C. et al.2023EmotionPrompt: emotional stimuli in prompts improve LLM output; 10.9% average improvement on human evaluation, up to 115% on edge-case benchmarks (AAAI 2024)L3 Kasatskii, V. et al. (JetBrains)2023Perceptual load in IDE environments differentially affects developers with ADHD symptoms; low load shows speed benefits that differ from neurotypical developersL2/L0 Carik, B. et al.2024Neurodivergent users perceive LLM output as “overly neurotypical”; community-driven workarounds documented (ACM PACMHCI / CSCW2, peer-reviewed)L2/L3 On provenance: Several PFD concepts were independently developed through embodied practice (improv, security work, 15 years of client work) before encountering the corresponding literature. The improv notebook (2012–2013) contains a practitioner articulation of the core predictive processing operation (“comparing mental representation with reality”) that predates Clark’s landmark 2013 paper.On methodology: This document was formalized with AI assistance (Claude, Anthropic). The underlying methodology, case studies, and practitioner insights are the author’s own work. AI was used for structuring, identifying relevant literature, and iterative refinement. Citations were separately audited for accuracy. The framework’s ideas predate the AI assistance; the formalization does not. Appendix B: Glossary Perception barrierAn unconscious objection that prevents conversion. Not a product, price, or demand problem. Fence-sitterSomeone with intent who doesn’t convert due to perception barriers. Mental waterfallThe conducted sequence of thoughts and emotions from first impression through conversion. Activation pointThe moment a user switches from autopilot to active engagement. Listen Multiplier(Listen) × (Act + React). If the user isn’t listening, everything multiplies by zero. Cognitive loadTotal mental effort. Intrinsic (task) + extraneous (design). Processing fluencySubjective ease of processing. Higher fluency = more positive evaluations. Perception biasSystematic gap between what users say, what stakeholders believe, and what actually happens. Decision architectureThe structure of choices as presented. Every structure is a nudge. Empathy modelingEmbodying the traits and context of a target demographic for directionality. Hardware layerShared cognitive architecture: attention, memory, autopilot, mind perception. Universal. Software layerCultural conventions: style preferences, expectations, literacy. Varies by audience. Mind perceptionUsers construct lightweight models of competence (Agency) and care (Experience) for websites. Predictive processingBrain generates predictions about input. Matched = autopilot. Violated = attention fires. Mental prosthesisScaffolding for cognitive functions and limitations. Good design IS prosthesis for the brain. The curb cut effect. Cognitive Constraint DesignA proposed meta-framework: the principle that cognitive constraints are design material, not obstacles. PFD is the first formal application to interface design. Curb-Cut EffectWhen designs optimized for constrained users improve experience for all users. Named after sidewalk curb cuts designed for wheelchair users that benefit everyone. Dual FoundationPFD’s evidence base draws from two largely distinct citation pools — perception psychology and ADHD cognitive neuroscience — with zero citation overlap. IBNS (Interest-Based Nervous System)Dodson’s clinical model describing ADHD motivation as driven by interest, novelty, competition, and urgency rather than importance or reward. Prediction ErrorThe mismatch between what the brain expects and what it encounters. The fundamental mechanism underlying first impressions and trust judgments. Convergent DiscoveryWhen multiple independent practitioners arrive at the same insight without knowledge of each other. Ten or more ADHD-AI builders discovered the same collaboration patterns independently, confirming environmental fitness. Point of PerformanceBarkley’s (1997) principle: self-regulation failures in ADHD occur in the moment of action, not during planning. Design must provide support where the decision happens, not where the instruction was given. Use It Now you have the framework. What comes next depends on what you need. Learn it. The Make Me Think series walks through each layer with case studies, citations, and the stories behind the principles. This reference page is yours to come back to. Cite it, teach it, build on it. Use it in your tools. The framework and AI skill are open source on GitHub. Free to use with attribution. Apply it. If something isn’t converting and you want the methodology applied to a specific problem, that’s what I do. E-commerce, SaaS, consumer products. 15 years of it. Services has the details. Ask me. Genuinely. If you read this far, I’ll read your email. stefan@aurochs.agency ← Back to the methodology page --- ## Chapter 1: The Bouncer URL: https://aurochs.agency/writing/make-me-think/the-bouncer/ My boss pulled me aside on a Tuesday night and told me I wasn't doing my job. I was 21, working nights at a nightclub in Santa Barbara to help pay for school. Started as a barback. The job was simple: bus tables, restock the bar, stay out of the way. I'd been at it for about a month, and I was good at it. The kind of good where you don't stop moving because the second you stop, you're behind. But the boss caught one moment. One snapshot of me standing still during a rush, maybe catching my breath, maybe scanning the room to figure out where to go next. Didn't matter. He saw what he saw, and what he saw was a kid not working hard enough. He chewed me out in front of the staff. I didn't argue. Not then. I waited until the end of the night, asked for five minutes in his office, and said something like: "I get why you saw what you saw and reacted the way you did. But that moment wasn't representative of what I'd been doing all night. I get paid to do a job, and I'll do it well." The next night, he asked me if I wanted to learn how to check IDs. Within a week I was the second security manager on staff. He walked me through what he needed: ABC compliance, risk minimization, the posted rules at the door stanchions when evening hours kicked in. How to read a fake ID. How to manage a line. How to make the call on who gets in and who doesn't. I didn't realize it at the time, but what happened in that office was the first time I consciously identified a perception gap. My boss had formed an impression from incomplete information and acted on it like it was the full picture. His perception didn't match reality. And my response, without knowing the vocabulary for it, was perception-first: I didn't argue about the facts. I acknowledged his experience of the situation, then offered a different frame. That pattern became my entire career. Learning to See The door taught me things I didn't know I needed to learn. I'm autistic, though I wouldn't get that diagnosis for years. Eye contact was overwhelming for me. Not metaphorically. Literally. Looking someone in the eyes felt like staring into the sun. But I could see that other people extracted meaning from it, that it signaled something I wasn't naturally receiving. So I studied it the way you'd study a foreign language. Not the feeling of it, but the mechanics. Working the door meant over a thousand face-to-face interactions a night. A 300-capacity venue open from 8 PM to 1:30 AM, people cycling through every couple hours. You do the math. I had to learn fast. I started noticing what happened when I held eye contact versus when I didn't. When I softened my posture versus when I squared up. When I led with warmth versus authority. The greeting was the breakthrough. Early on, I'd ask people how they were doing. Literal, genuine, because that's how my brain works. A woman stopped me one night and told me I should never ask that question because someone could be having a terrible day. She was right. The question forced people into a decision: be honest or perform. Neither serves the interaction. So I stopped asking. Instead, I'd tell people I hoped they were having a good night. Small change. Massive difference. People's attention turned on. They'd make eye contact back. They'd smile. Instead of a ritual exchange, it became a real moment. The greeting activated something. I didn't have words for what I was seeing. I just knew that when I got the first few seconds right, everything after was easier. The ID check was smoother. The "you've had enough" conversation was calmer. The "time to leave" moment at closing was less hostile. The first few seconds were doing all the work. 1:30 AM The nightclub held about 300 people. Every night at 1:30 AM, same problem: the DJ announces we're closing, the house lights slam on, and I'm barking at 300 drunk people to get out. They hated it. I hated it. Fights broke out. Stragglers lingered. The whole thing was hostile and exhausting. So I tried something different. At 1:10, twenty minutes before close, I'd walk table by table. "Hey, last call's coming up. Get any drinks you want now. Thanks for coming out tonight." That's it. No urgency. No authority voice. Just information, a suggestion, and respect. At 1:30, when the DJ made the announcement and the lights went up, half the venue was already heading for the door. No barking. No fights. And bar sales went up 30% in those last 20 minutes, because people ordered one more round when I gave them the chance instead of the boot. I didn't know what to call what I'd done. I just knew it worked, and I knew why it worked. The old way treated people like obstacles to be removed. My way treated them like people with their own goals who'd cooperate if you gave them the right information at the right time in the right tone. I'd primed each table with a personal visit. Set the tempo by starting twenty minutes early. Given them a clear next action. Resolved the experience with genuine thanks. By the time the lights came up, leaving felt like their idea. That's conducting. Not commanding, not funneling, not manipulating. Conducting. Like a musician who's reached the point where they stop thinking about technique and focus entirely on how the music makes you feel. The Same Pattern, Everywhere Santa Barbara was an interesting place to learn this. It was a playground for wealthy people, full of signals about what perception means in real life. What people wore, where they ate, how they carried themselves. It spoke volumes about their priorities, their wealth, their aspirations. International students visiting made it even more visible. The whole town was a microcosm for why looks and lifestyle matter in a tangible way, and how people associate themselves with those patterns and circles. I could see the ruleset: who you know matters more than what you know. I wanted to get around that rule. I believed, and still believe, that being so good they can't ignore you is the better path. But the reality of how many smart, hungry people know the same thing and are competing for the same attention is humbling. What I was developing, without a name for it yet, was a sensitivity to the gap between how things are and how things are perceived. The boss who saw a lazy barback instead of a hard worker. The greeting that activated attention instead of triggering a ritual. The closing announcement that felt like a command versus the table visit that felt like a courtesy. In each case, the underlying reality was the same. The perception was the variable. My first real website was for the fraternity chapter I helped co-found with 49 other guys at UCSB. It never made it to production, but it was the first time I connected code with graphics and styling, and it hooked me. I pitched a website to the nightclub itself in 2007. They liked it. Not enough to pay for it, but they liked it. In 2010, I co-founded a web design and hosting company. We specialized in high-traffic event sites, and our first wave of clients came through a connection to Barbara Corcoran's network. Yes, that Barbara Corcoran, the one from Shark Tank. I handled design and operations; my co-founder handled the development and infrastructure. For three years, we built websites for dozens of clients, and I kept experimenting with the patterns I'd been noticing since the nightclub: what makes a visitor trust a page in the first second? What makes them stay? What makes them act? I wasn't applying a methodology. I was running on instinct informed by undergrad psych courses and a growing shelf of books about cognition, behavior, and persuasion. I could feel when something worked, but I couldn't always explain why. In 2013, my co-founder had to step away from the business for personal reasons. The company didn't survive the transition. No bitterness. Just an ending. The Thesis Steve Krug wrote a book called Don't Make Me Think. Great book. It gave me the vocabulary to talk to stakeholders early in my career. "Users have limited attention, so don't waste it. Lower friction. Reduce steps. Make everything obvious." He was right about the foundation. But the reality is more nuanced than "reduce friction." My version is different: users don't think, until you make them. Nobody is thinking. They're on autopilot. The same way you catch a ball without calculating trajectory, or walk without planning each step, or explain something without scripting your answer first. That's how people interact with websites, products, brands, everything. Autopilot. Energy conservation. The brain handles the vast majority of processing unconsciously. The design challenge isn't "don't waste their limited attention." It's that effortful attention is dormant by default. The brain is always processing, but conscious engagement is off until something activates it. The real question is: when you do make them stop and actually think, what do they think about? How do they feel? And what do they do next? That's what I design for. Not the funnel. The activation. I think of my nightclub hero section. The hero of any website is the table visit at 1:10. If it answers the first five questions a visitor unconsciously processes (What am I trying to do here? Is there a clear path? What's the point of this site? Is this for me? How does it make me feel?), the visitor is primed. Then you conduct them through the rest of the experience as they scroll. If the hero fails, it's the lights slamming on at 1:30. They're gone before you've started. Krug wrote the definitive book about cognitive load, the part about reducing cognitive friction so conscious decisions feel effortless. But there are four more layers above L0: first impressions, processing fluency, perception bias, and decision architecture. Those layers are what this series is about. In the 15 years since that nightclub, I've applied these patterns to everything I've touched. A smart home brand tripled its annual revenue. Same product. Different perception. Costco pallet placement, Disney licensing, Walmart shelves. An improv theater increased online ticket sales by half again. A vacuum sealer company has been a client for nearly a decade because the perception compounds. It all comes from the same instinct. Design for how people actually perceive and process the world, not how you wish they did. What Comes Next The next time you look at a website, notice what you feel before you start analyzing. That split-second impression, the gut response before your conscious brain catches up, that's the gap where the real design happens. Most designers skip straight to the analysis. I'm asking you to pay attention to the moment before it. This series is 12 chapters. I'll walk through the five layers of Perception-First Design, the science behind each one, the cases where I've applied them, and the ethical questions that come with designing for perception. I'll also tell you where the framework breaks down and what I haven't figured out yet. Next: They're Already Not Thinking. The autopilot problem, predictive processing, and why Krug's first law is correct but incomplete. Key Terms Perception gap The difference between how something is and how it's perceived. The boss saw a lazy barback; the reality was a hard worker catching his breath. Design operates in this gap. Activation point A designed moment that shifts a user from unconscious processing to conscious engagement. The greeting at the door. The table visit at 1:10. The hero section on a website. Conducting Guiding a user's attention and emotion through an experience, the way a musician conducts a performance. Not commanding, not funneling. Priming, then directing flow. Five layers The full Perception-First Design framework: Cognitive Load, First Impressions, Processing Fluency, Perception Bias, and Decision Architecture. References Kahneman (2011) Thinking, Fast and Slow. Farrar, Straus and Giroux. Krug (2000) Don't Make Me Think. New Riders. Newport (2012) So Good They Can't Ignore You. Grand Central Publishing. A note on how this was written: This series is AI-assisted. I provide the stories, the methodology, the case studies, and the editorial direction. AI helps me structure and draft. This is consistent with Perception-First Design's own transparency principle: if I'm writing about perception, I should be honest about how the writing itself is produced. --- ## Chapter 2: They're Already Not Thinking URL: https://aurochs.agency/writing/make-me-think/already-not-thinking/ Every night at the door, I'd watch the same thing happen. Someone walks up to the venue. They glance at the entrance, the lighting, the crowd visible through the window, the energy spilling onto the sidewalk. In maybe half a second, their body language tells me whether they're coming in. Shoulders forward, wallet already coming out of their pocket. They're in. Shoulders back, half-turn toward their friend, slight hesitation. They're about to suggest going somewhere else. Nobody was deciding. Nobody was weighing the pros and cons of this bar versus the one down the street. They were pattern-matching. Does this place look like the kind of place I want to be tonight? Does it match the picture in my head? If yes, autopilot carries them through the door. If no, autopilot carries them past it. By the time someone was standing at my door, the outcome was already locked in. My job wasn't to convince them. It was to not break the thing that was already working. Don't Make Me Think Steve Krug published Don't Make Me Think in 2000, and it changed how an entire generation of designers approached the web. His core insight was elegant: users don't want to puzzle over your interface. Make things obvious. Reduce friction. Don't ask people to think when they're trying to accomplish a task. He was right. For L0 (cognitive load, navigation clarity, interface usability), his book is still the best practical guide I know. I used it early in my career to have productive conversations with stakeholders who didn't speak design. "Don't make them think" is a phrase anyone can understand, and it unlocked real improvements on real projects. But there's an assumption buried in Krug's framework that deserves examination: the idea that users arrive at your site in a state of active attention that you then need to conserve. That they show up thinking, and your job is to reduce the thinking so they can act. That's not what I saw at the door. And it's not what the science says either. Nobody Is Thinking The thing about your website visitors that changes everything once you internalize it: they're not thinking. Not "they don't want to think." Not "they have limited attention." They are, in a meaningful cognitive sense, not engaged. They're on autopilot. The same way you catch a ball without calculating trajectory, walk without planning each step, or drive a familiar route without remembering the turns. That's how people browse the web. Automatic. Unconscious. Energy-conserving. Daniel Kahneman laid the groundwork for this in Thinking, Fast and Slow (2011). His System 1 / System 2 framework describes two modes of cognition. System 1 is fast, automatic, and effortless. It handles pattern recognition, snap judgments, emotional reactions, and the vast majority of daily processing. System 2 is slow, deliberate, and effortful. It handles complex reasoning, careful comparison, and conscious decision-making. Most of life runs on System 1. Nearly all web browsing does. When someone lands on your homepage, System 2 is off. Not low-power. Not conserving energy. Off. System 1 is doing what it always does: scanning for patterns, checking predictions, deciding in milliseconds whether this environment matches expectations. If everything matches, if the site looks like what they expected, if the information is where they predicted it would be, if nothing triggers a "wait, what?" then System 1 handles the entire visit. The user scrolls, clicks, maybe converts, and leaves without ever consciously engaging with your design. This is the default state. Not the exception. The Prediction Machine Contemporary cognitive science has a framework for this mechanism called predictive processing, and it's one of the most significant developments in how we understand the brain. Andy Clark's Surfing Uncertainty (2016) describes the brain not as a passive receiver of sensory information but as a prediction machine. Your brain is constantly generating models of what it expects to encounter: what the next visual frame will look like, what the next sound will be, what will happen when you reach for your coffee cup. Perception isn't bottom-up (data comes in, brain processes it). It's top-down (brain predicts what data should look like, then checks). Karl Friston's free energy principle (2010) provides the mathematical foundation. The simplified version: the brain is always trying to minimize surprise. It builds internal models of the world and works to keep those models accurate. When reality matches the model, processing is smooth and unconscious. When reality violates the model, prediction error fires, attention activates, and the brain spends energy updating. This is why it matters for design: when your website matches the brain's predictions at every level, the entire experience is processed unconsciously. The user doesn't notice your navigation because it's where they expected it to be. They don't notice your color scheme because it matches the category norms for your industry. They don't notice your layout because the visual hierarchy confirms their scanning patterns. They don't notice your design at all. And that's supposed to be good. The framework goes deeper than prediction-matching. Ran Hassin reviewed a decade of evidence in 2013 and concluded that the unconscious performs every high-level cognitive function previously assumed to require awareness. Decision-making, goal pursuit, cognitive control, self-regulation. Pre-verbal arousal isn’t a warm-up for conscious thought. It’s the primary processing mode. And it runs in two measurable directions. Reward cues presented below conscious threshold drive motor behavior before they’re consciously registered. Interfaces that land as near-human-but-not-quite trigger an eeriness response before conscious evaluation can fire. The brain works toward rewards it never consciously registers. The brain rejects interfaces that register as uncanny before it can articulate why. Both polarities run under autopilot. What you feel, whether positive pull or visceral rejection, arrives before the analytical mind shows up to explain it. The Invisible Visit But Krug's framework doesn't reach this far. If the user is on autopilot, if System 1 is handling the visit and System 2 never activates, then they're not deciding anything. They're gliding. And gliding doesn't lead to conversion. Gliding leads to bouncing. Not because something went wrong, but because nothing went right enough to make them stop. Think about the difference between walking past a store and walking into one. Walking past is autopilot. The storefront matched your predictions ("that's a store"), nothing violated your model, and System 1 carried you right past it. You didn't decide not to go in. You never considered it. The store was processed and dismissed without conscious engagement ever firing. That's what happens to most website visits. The user lands, System 1 scans, nothing triggers engagement, and they leave. Your analytics show a bounce. But it wasn't a decision to leave. It was a failure to activate a decision to stay. This is the gap in "Don't Make Me Think." If you optimize exclusively for frictionless processing, if the entire experience glides by without ever generating a moment of conscious engagement, you've designed a beautiful, usable website that nobody converts on. They processed it. They just never engaged with it. The design challenge isn't "don't waste their attention." It's that effortful attention is dormant by default, and you need to know when and how to wake it up. I call that moment an activation point. In predictive processing terms, an activation point is a controlled prediction error. Something in the design that doesn't quite match what the brain expected. Not enough to trigger rejection, but enough to generate a "wait, what?" that pulls System 2 online. I was creating activation points at the nightclub years before I had the framework for them. That table visit at 1:10 ("Hey, last call's coming up, get any drinks you want now") was a controlled interruption. People were on autopilot: drinking, talking, having a good time. My visit introduced a prediction error (someone just appeared at my table with new information). But it was calibrated. Personal tone, useful information, respectful framing. So instead of triggering annoyance or resistance, it activated engagement. People went to the bar. Sales went up 30% in those last twenty minutes. On a website, the hero section is the activation point. It's the first thing that has the opportunity to break autopilot and create conscious engagement. If it does its job, if it generates a prediction error that's interesting rather than confusing, the user shifts from scanning to reading, from gliding to engaging, from System 1 to System 2. If it doesn't, they glide right past you. Not because your site was bad, but because it was invisible. Two Failure Modes This reframes the entire design conversation. It's not about attention as a scarce resource you need to budget carefully. It's about attention as a dormant system you need to activate strategically. Both columns are true. That's important. I'm not saying Krug was wrong. He was describing a real phenomenon: when users are actively trying to accomplish something, unnecessary complexity drives them away. That's real. That's L0, cognitive load. You need it in place first. But L0 only handles one failure mode: "the user was engaged and you lost them." It doesn't address the more common one: "the user was never engaged and you never had them." Krug's first law of usability is correct. It's also only the foundation. The full picture: reduce friction on actions, invite engagement on value. Make the mechanics invisible. Make the meaning unmissable. The Upstream Problem There's a practical consequence here that most design processes miss entirely. If you accept that users are on autopilot by default, then a huge percentage of "conversion problems" aren't what they look like. They look like friction problems ("the checkout is too complex," "the CTA isn't prominent enough," "we need to A/B test the button color"). And sometimes they are. Cognitive load matters. But often, the real problem happened upstream. The user never shifted out of autopilot. They never engaged with your value proposition because the design never activated their attention. They scrolled past your perfectly worded copy because nothing in the visual experience created a prediction error worth stopping for. I've seen this in every engagement I've worked on. The client comes in saying "our conversion rate is low" and pointing at the checkout flow. But the data shows the real drop-off is much earlier: the homepage, the landing page, the first scroll. The user isn't abandoning a process. They never started one. You can't A/B test your way to understanding this. A/B testing compares two versions of a conscious experience. If the user never reaches conscious engagement, you're comparing two versions of invisible. The lift will be marginal at best, because the problem isn't in the details you're testing. It's in the layer you're not seeing. Methodology Siblings This framework isn’t novel. Three older traditions saw different parts of the same territory before PFD did. Mitsuo Nagamachi formalized Kansei engineering at Hiroshima University in 1995. Kansei translates roughly as “psychological feeling, image of a new product.” The methodology maps consumer affective response onto measurable product attributes, then engineers toward the target feeling. It’s PFD’s closest ancestor: both start with the pre-verbal response and work backward to design decisions. Anyone doing serious work in automotive, consumer electronics, or product design in Japan already knows it. Max Wertheimer, Wolfgang Köhler, and the rest of the Berlin School published Gestalt psychology in the 1920s and 30s. Wolfgang Metzger’s 1936 book Laws of Seeing remains one of the clearest empirical treatments of how the visual system constructs meaning from raw input. Proximity, similarity, closure, continuity, common fate, figure/ground. These aren’t decorative rules. They’re the first systematic description of how perception organizes itself. L1 and L2 in the PFD stack are applied Gestalt. The debt is direct. Helmut Leder in Vienna and Martin Skov in Copenhagen have spent the last two decades building neuroaesthetics, the modern empirical project of mapping aesthetic experience onto neural substrate. Amygdala, orbitofrontal cortex, reward pathways. PFD shares the sciences with all three. What it adds is the dependency stack. Kansei engineers the feeling but is silent on layer ordering. Gestalt describes the perception but not the decision. Neuroaesthetics explains the brain but isn’t a working methodology for a Tuesday afternoon. What I add: the ordering rule and the operational protocol, the thing that tells you which layer to fix first and why the fix won’t take if you skip the floor. The rest is inherited. What to Take From This Everything else in this book builds on this thesis. Your users are not making decisions on your website. They're running predictions. Their brains are pattern-matching against expectations built from every other website they've ever visited, every store they've ever walked into, every first impression they've ever formed. When the patterns match, processing is smooth and unconscious. When they break, attention fires. Your design needs to do two things that sound contradictory but aren't. First, match predictions where it matters. Navigation, layout structure, visual category norms. These should feel familiar. Violating them generates the wrong kind of prediction error: confusion, suspicion, the "this doesn't feel right" response. This is Krug territory. Don't make them think about the mechanics. Second, break predictions where it counts. Your value proposition, your differentiation, the thing that makes you worth stopping for. This needs to generate a prediction error that triggers curiosity instead of avoidance. The brain encounters something it didn't predict, and instead of clicking away it leans in. "Wait. That's different. Tell me more." The whole art is knowing which predictions to match and which to break. Get it wrong in the first direction and you have a confusing mess. Get it wrong in the second and you have a perfectly usable site that nobody remembers visiting. Get it right, and you have an experience that feels effortless AND meaningful: smooth where it should be smooth, engaging where it needs to be. A site that people don't just use but actually notice. Perception-First Design exists to solve this problem: to design the activation, not just manage the friction. Next: The Tuning Fork. On what happens when your own perceptual instrument is calibrated differently, and why that turned out to be the whole point. Key Terms Autopilot The default cognitive state where the brain processes information unconsciously through pattern-matching, without deliberate engagement. System 1 / System 2 Kahneman's dual-process framework. System 1 is fast, automatic, and effortless. System 2 is slow, deliberate, and effortful. Most web browsing runs entirely on System 1. Predictive processing The brain continuously generates predictions about incoming sensory data and only allocates conscious attention when predictions are violated. Free energy principle Friston's mathematical framework: the brain minimizes surprise by maintaining and updating internal models of the world. Matched predictions = smooth processing. Violations = attention. Pre-verbal arousal The primary processing mode of the unconscious: it performs every high-level cognitive function (Hassin, 2013) and operates in two measurable directions. Positive: subliminal reward drives motor behavior (Pessiglione et al., 2007). Negative: uncanny-valley response triggers rejection before analysis (Mori, 1970). What you feel arrives before the analytical mind shows up to explain it. Prediction error The mismatch between what the brain predicted and what actually occurred. Generates attention and triggers model-updating. The mechanism behind activation points. Activation point A controlled prediction error in design that shifts users from autopilot to conscious engagement without triggering rejection or confusion. References Clark (2016) Surfing Uncertainty: Prediction, Action, and the Embodied Mind. Oxford University Press. Friston (2010) The free-energy principle: a unified brain theory? Nature Reviews Neuroscience, 11(2), 127–138. Kahneman (2011) Thinking, Fast and Slow. Farrar, Straus and Giroux. Krug (2000) Don't Make Me Think. New Riders. Hassin (2013) Yes it can: On the functional abilities of the human unconscious. Perspectives on Psychological Science, 8(2), 195–207. Pessiglione et al. (2007) How the brain translates money into force: a neuroimaging study of subliminal motivation. Science, 316(5826), 904–906. Mori (1970/2012) The uncanny valley (K. F. MacDorman & N. Kageki, Trans.). IEEE Robotics & Automation Magazine, 19(2), 98–100. Nagamachi (1995) Kansei engineering: A new ergonomic consumer-oriented technology for product development. International Journal of Industrial Ergonomics, 15(1), 3–11. Metzger (1936/2006) Laws of Seeing (L. Spillmann, M. Wertheimer & S. Lehar, Trans.). MIT Press. Skov & Nadal (2020) A farewell to art: Aesthetics as a topic in psychology and neuroscience. Perspectives on Psychological Science, 15(3), 630–642. A note on how this was written: This series is AI-assisted. I provide the stories, the methodology, the case studies, and the editorial direction. AI helps me structure and draft. This is consistent with Perception-First Design’s own transparency principle: if I’m writing about perception, I should be honest about how the writing itself is produced. --- ## Chapter 3: The Tuning Fork URL: https://aurochs.agency/writing/make-me-think/the-tuning-fork/ My partner said something to me a few years ago that I still think about. We were in the middle of a conversation. I was explaining something I was excited about, going deep on some technical detail, building the argument layer by layer the way I do. He stopped me and said: "Do you realize you're not picking up on me not being interested at all in what you're talking about?" He wasn't being cruel. He was being accurate. I had missed every signal. The shortened responses. The body language pulling away. The polite nodding that meant "I'm done" and not "keep going." All of it had sailed past me while I was inside the structure of my own thought. That moment crystallized something I'd been circling for years. I don't read rooms the way most people read rooms. I never have. I'm autistic. Formally diagnosed, ASD. And I have severe combined-type ADHD. For most of my life I didn't have the vocabulary for either of those things. I just knew that certain parts of being human felt like they were running on a different operating system than what everyone else seemed to have. Eye contact felt like staring into the sun. Conversations had rules I could observe but not feel. Small talk was a protocol I memorized rather than a thing I enjoyed. And my attention was either a laser or a scattered mess, nothing in between. I want to be clear: this is not a disability narrative. I'm not telling you this because it's inspirational. I'm telling you because these two things, the autism and the ADHD, turned out to be the reason I can do what I do. They're not obstacles I designed around. They're the instruments I design with. The autism gave me something specific. Because I don't process social cues automatically, I had to learn them the way you learn an instrument. Theory, practice, application, more practice, repeating that loop until the thing that started as conscious effort became something closer to fluency. I didn't learn what a smile means by feeling it. I learned it by observing patterns. Cataloging responses. Testing hypotheses in real interactions and noting what worked and what didn't. Someone leans forward, usually engagement. Feet pointing toward the door means they're ready to leave. Eyes breaking down and to the left, that's internal processing. To the right, they're constructing something to say. Most people absorb this stuff unconsciously as children. Their social cognition is intuitive, built from millions of micro-interactions processed automatically by neural architecture that handles it below the level of awareness. They can read a room without knowing how they read a room. I can't do that. What I can do is tell you exactly what I'm reading and why. I can articulate the patterns because I had to learn them explicitly. The mechanics under the intuition, the part that's invisible to everyone else, that's the only version I have. Working the door at the nightclub was where this became an advantage instead of a limitation. A thousand face-to-face interactions a night. A 300-capacity venue, people cycling through every couple of hours. I had to get fast at reading signals I couldn't process naturally. So I built systems. Specific cues mapped to specific responses. Posture combinations that predicted behavior. Vocal patterns that told me whether someone was escalating or winding down. Analytical social cognition. The deliberate kind, where every read is a conscious act of pattern recognition, nothing like the warm, effortless version most people run on. Slower at first. But with thousands of repetitions per week, it got fast. And because it was explicit rather than intuitive, it was transferable. I could apply it to any social system, not just the one I grew up in. Most people miss this about neurodivergent perception: when you learn a skill analytically that others learn intuitively, you end up understanding the mechanics at a level they never need to reach. A native speaker knows their language fluently. A non-native speaker who achieved fluency knows the grammar. I know the grammar of social perception. And people interact with websites the same way they interact with other humans. The ADHD is a different instrument entirely. Some experiences are literally painful for me. Not metaphorically. Cluttered interfaces, confusing navigation, forms that ask for the same information twice, pages that make me hunt for the thing I came to find. These don't just annoy me. They hurt. The cognitive load registers in my body as a physical sensation, a tightness, an agitation, a need to escape. What most designers theorize about, I feel. That sensitivity is extreme. Most users experience friction as mild annoyance or unconscious abandonment. They leave a confusing site without knowing why they left. For me, the "why" is screaming. I can point at the exact element that broke the experience, name the cognitive load it imposed, and explain why it triggered the exit response. Because I live on the sharp end of that response every day. The ADHD also gave me something I didn't expect: years of building mental prostheses. If something is out of sight, it's out of mind. I mean that literally: if I can't see it, it doesn't exist in my working memory. So I keep post-it notes on my desk for today's tasks. Not because I'm organized. Because without them, I forget what I'm doing mid-sentence. I have to actively "turn on" when listening to people. Processing the non-verbal cues, the implications, the context, the said-vs-unsaid. Neurotypical people handle this automatically. For me it's a learned skill, like a dancer controlling every muscle on stage. It looks natural from the audience. Backstage, it's deliberate effort. Those prostheses, the external memory aids, the active listening protocols, the friction-reduction strategies, they map directly to design. "Out of sight, out of mind" becomes a design question: what must we show versus what we want to show versus what we could show? Then cut everything except what we must show. Active listening becomes user empathy: who is this for, what's at stake, what's the desired outcome, what's being said and what's not being said? The scaffolding I build for my own brain is the scaffolding I build for users. There's a concept in accessibility called the curb cut effect. It comes from a literal observation: when cities started cutting curbs at crosswalks for wheelchair users, everyone benefited. Parents with strollers. Delivery workers with hand trucks. Travelers with rolling luggage. Runners. Skateboarders. A modification designed for a specific disability turned out to be better design for everyone. My entire career in one concept. Design is prosthesis for human cognitive functions and limitations. I know that because I've been building prostheses for myself my whole life. The strategies I developed to manage my own attention, reduce my own cognitive load, and compensate for my own processing gaps, those turned out to be exactly what users need, not because they share my diagnosis, but because the human brain has universal constraints that good design accommodates and bad design ignores. The post-it note principle (out of sight, out of mind) applies to every user. If your call-to-action scrolls off screen, it stops existing. The active listening principle applies to every interface. If your design doesn't signal that it understands the user's goal, the user feels unheard. What I built for my own cognitive accessibility helps everyone. The curb cut effect, in pixels instead of concrete. It keeps showing up in design because human cognition has limits, and designing for those limits is what good design is. Around 2012, a friend I met at a Chicago meetup about emotions in video games suggested I try improv. I went in already analytical. The only way I know how to go into anything. And improv, it turned out, was the bridge between the two halves of my perceptual system: the analytical social cognition from the autism and the raw sensitivity from the ADHD. One of my improv teachers gave me a formula: (Listen) x (Act + React). The formula for being in the moment. If Listen equals zero, if you're not attending, then everything else multiplies by zero. Nothing you do matters if you haven't heard what's actually happening. That formula hit me like a brick. It was the nightclub door distilled into mathematics. Every successful interaction I'd ever had, every table visit at 1:10, every greeting that activated attention instead of triggering a ritual, they all started with listening. With receiving the room before trying to change it. Improv taught me other principles that turned out to be design tools, not metaphors. "Yes, and." Accept what exists and build on it. Don't throw everything out when something feels wrong. Brick by brick, not castle at a time. In design: inherit the existing brand, acknowledge the current state, then layer improvements. Stakeholders who feel heard will let you push further than stakeholders who feel overridden. "If this is true, what else is true?" One insight cascades. If this demographic behaves this way on mobile, what else follows? If this is the brand voice on the homepage, it should be the brand voice everywhere. Follow the thread. "It's not for you, it's for them." Protect the design from stakeholder preferences. Design for the target's autopilot, not your own comfort. Not the CEO's taste. Not the designer's portfolio. The user's actual perceptual experience. These aren't improv rules borrowed as clever analogies. They're operational tools I use every week. When a client says "can we add a banner here?" my response is yes-and: "Yes, and if we add it, here's what it does to the user's attention at this point in the scroll. Is that the trade-off we want?" When I audit a site, I'm asking "if this is true, what else is true?" about every signal the design sends. All of this connects back to the diagnostic process I described at the end of the last chapter. Step 1 of the PFD diagnostic is Feel. Arrive at the page and let the emotional response fire before your conscious brain translates it into words. "This makes me feel X." The feeling comes before the language. System 1 before System 2. That step, the pre-verbal emotional read, is where my ADHD sensitivity becomes the diagnostic instrument. Most designers can learn to do this. They can practice pausing before analyzing, noticing their gut reaction, giving the emotional response room to register. But for me, it's not optional. The feeling arrives whether I want it to or not, louder than it arrives for most people, impossible to ignore. The emotional response IS the diagnostic instrument. Like a tuning fork that resonates at the slightest vibration while everyone else needs to put their ear to the surface. This is the part of PFD that may not be fully teachable. I've thought about this honestly. The lens, the empathy-first approach, the "design for how they perceive, not how you wish they'd perceive" mindset, that's completely teachable. Give anyone that lens and they get better immediately. The pattern library, the hundreds of projects of instant recognition, that takes years. No shortcut. But the sensitivity, the neurodivergent tuning fork, the thing that lets me feel cognitive friction before I can name it? That might be native hardware. The good news: the diagnostic process, Feel, Unpack, Diagnose, Prescribe, can be taught as a substitute. You don't need my nervous system to run the protocol. You need to practice the pause. You need to learn to notice your own pre-verbal reactions instead of skipping past them. You need to treat the emotional read as data instead of noise. The tuning fork is calibrated differently in my case. But everyone has one. Most people just haven't learned to listen to it. I want to be direct about what this chapter is really saying, because it's easy to read it as "neurodivergence is a superpower" and leave it there. That framing is too simple. Autism and ADHD cost me things. Relationships where I missed signals for months. Jobs where I couldn't mask well enough. Days where the ADHD makes it impossible to start a task I genuinely want to do. The sensitivity that lets me feel friction in a design also means I feel friction in a grocery store, in a crowded room, in a conversation that goes sideways. It's not selective. It's all the time. What I'm saying is narrower and more specific. The same cognitive architecture that creates difficulty in some contexts creates professional advantage in one specific context: understanding how humans perceive and process designed experiences. The analytical social cognition lets me see patterns that intuitive readers can't articulate. The friction sensitivity lets me feel problems that typical users walk past. And the years of building prostheses for my own brain taught me what good scaffolding looks like, because I can't function without it. Not a superpower. A trade-off that happened to land in a useful place. The improv filled the last gap. Before improv, I had the analytical read and the emotional sensitivity, but they operated in parallel. The analysis was one track, the feeling was another, and bridging them in real time was clumsy. Improv taught me to integrate them. To feel the room and act on the analysis simultaneously. To let the emotional read inform the analytical response without either one drowning out the other. The diagnostic protocol came directly from that integration. Feel first: the ADHD instrument. Then Unpack what specifically triggered that feeling. Then Diagnose which of the five layers is the source. Then Prescribe what would resolve the violation without creating new ones. The analytical and the intuitive, working together, sequenced deliberately because that's how I learned to do it. The tuning fork doesn't work without the analysis. The analysis doesn't work without the tuning fork. The methodology is the integration of both. If you remember one thing from this chapter, make it this. Your own perceptual system is an instrument. It has a calibration. It has biases and sensitivities and blind spots. The first step in designing for how other people perceive is understanding how you perceive. Not to center yourself in the process, but to know your instrument well enough to compensate for its tuning. If you're neurotypical, your social cognition is probably intuitive. That means you'll read rooms quickly but struggle to articulate what you read. Practice the pause. Practice naming what you feel before you analyze it. Practice the Feel step until it becomes habitual. If you're neurodivergent, your instrument is calibrated differently, not worse, not better, just tuned to different frequencies. Figure out what your calibration catches that others miss, and what it misses that others catch. Then build the protocol that accounts for both. The diagnostic starts with Feel because perception starts before language. Before analysis. Before any conscious processing at all. If you skip that step, you're designing from your assumptions instead of from the user's experience. And your assumptions, no matter how educated, are not the same as the pre-verbal read. The tuning fork doesn't lie. But you have to learn to hear it. Next: Cognitive Load. On why working memory holds 3–5 things (not 7), and what happens when your site steals all of them before the user starts. Key Terms Analytical social cognition Learning social cues through deliberate observation and pattern recognition rather than intuitive absorption. Slower to develop, but produces explicit, transferable understanding of the mechanics. Friction sensitivity The heightened awareness of cognitive load that comes with ADHD. Where most users experience friction as mild annoyance, friction sensitivity registers it as a physical, unavoidable signal. Mental prostheses External scaffolding built to compensate for cognitive gaps: post-it notes for working memory, active listening protocols for attention, progressive disclosure for information overload. The same scaffolding good design provides to all users. Curb cut effect A modification designed for a specific disability that turns out to be better design for everyone. Cognitive accessibility features work the same way. (Listen) x (Act + React) The improv formula for being in the moment. If Listen equals zero, everything multiplies by zero. The foundation of both good improv and good design. The tuning fork The pre-verbal emotional response that serves as the diagnostic instrument in Perception-First Design. Everyone has one. Most people haven't learned to listen to it. A note on how this was written: This series is AI-assisted. I provide the stories, the methodology, the case studies, and the editorial direction. AI helps me structure and draft. This is consistent with Perception-First Design's own transparency principle: if I'm writing about perception, I should be honest about how the writing itself is produced. --- ## Chapter 4: Cognitive Load URL: https://aurochs.agency/writing/make-me-think/cognitive-load/ The Simply Smart Home website had a problem that nobody on the team could name. The product was good. WiFi-enabled digital photo frames, priced competitively, solid reviews from people who actually owned them. The company was growing. But the website felt like reading the back of a box. Very matter-of-fact statements. Feature specs. Bullet points about resolution and connectivity. Not much else. I asked the team a simple question: if someone lands on this homepage for the first time, what do they learn in the first five seconds? The answer was sobering. They learn that a product exists. Maybe they learn its name. But they don’t learn what it does for them, why they’d want one, or what to do next. They’d have to work to figure all that out. And that word, “work,” is the whole problem. Every design framework needs a ground floor. Something that has to be true before anything else matters. For Perception-First Design, that ground floor is cognitive load. Specifically: your visitor’s brain has a hard cap on how much it can hold at once, and if your design exceeds that cap, everything else you’ve built on top of it collapses. This is the least glamorous chapter in this book. There’s no drama in “reduce complexity.” It doesn’t make for a compelling conference talk. But without this foundation in place, first impressions don’t form correctly, processing fluency can’t establish rhythm, perception biases can’t be optimized, and decision trails lead nowhere. You can’t perceive anything without the bandwidth to do so. This is where I give Krug his due. This IS Krug territory. Reduce friction. Don’t make them think about the mechanics. He wrote the definitive book about this layer, and if you haven’t read it, you should. Everything I’m about to say about cognitive load, he said first and said well. But his book stays at that layer, intentionally, and well. I’m treating it as the foundation. The science starts with one of the most misquoted findings in psychology. In 1956, George Miller published a paper called “The Magical Number Seven, Plus or Minus Two.” It became one of the most cited papers in the history of cognitive science, and it’s been misunderstood almost from the moment it hit print. Ask any designer how many items a user can hold in working memory and they’ll tell you seven, plus or minus two. It’s in every UX textbook. It’s the reason navigation bars have seven items. It’s the reason phone numbers have seven digits (they don’t, but people say this). Miller wasn’t proposing a working memory limit. He was making a rhetorical observation about the recurrence of the number seven across disparate information-processing phenomena: channel capacity, absolute judgment, memory span. It was a pattern he found interesting and possibly coincidental. The paper is playful. He calls the number seven a “pesty” digit that “follows him around.” He was describing a curiosity, not a law. The actual working memory limit is smaller. Nelson Cowan established the modern estimate through carefully controlled experiments published in 2001 and updated in 2010. When you prevent people from using rehearsal strategies (repeating things in their head) and grouping strategies (chunking phone numbers into segments), the limit drops to roughly 3 to 5 chunks. Not seven. Three to five. That distinction matters more than it sounds. If you’re designing a navigation with seven top-level items because “Miller says seven,” you’re already over budget for most visitors. You’re burning their working memory on orientation before they’ve even started evaluating your product. John Sweller formalized this in 1988 with Cognitive Load Theory, and his framework gave me language for something I’d been feeling for years. Sweller distinguishes between two types of load. Intrinsic load is the complexity inherent to the task itself. Comparing health insurance plans is intrinsically complex. Picking a t-shirt size is intrinsically simple. You can’t change intrinsic load. The task is what it is. Extraneous load is the complexity your design adds on top of the task. Confusing navigation. Unnecessary form fields. Three columns of competing information when the user needs one. A dropdown with 47 options when 6 would do. Every unit of extraneous load you pile onto the visitor is a unit of working memory stolen from the task they’re actually trying to accomplish. Think about it in concrete terms. A visitor arrives at your site with, let’s be generous, 4 chunks of working memory available. If your navigation takes 1 chunk to parse, your layout takes 1 chunk to orient, and your hero section presents 3 competing calls to action that each take a chunk to evaluate, you’ve already exceeded capacity. The visitor hasn’t even started thinking about your product. They’ve spent their entire cognitive budget figuring out your interface. What happens when the budget runs out? The brain does what it always does when overwhelmed. It takes the easiest action available. On a website, the easiest action is almost always leaving. When I first sat with the Simply Smart Home homepage, something was tiring me out that I couldn’t point to. It wasn’t the content. The content was sparse. It wasn’t the navigation either. The navigation was short. But the page took effort to look at, and the effort was coming from somewhere the visitor couldn’t name. Ran Hassin and his colleagues published a paper in 2009 that gave me the vocabulary for it. They pushed Sweller’s framework further. Working memory isn’t just spent on what the visitor actively pays attention to. It’s spent on what they never consciously notice at all. Background clutter. Inconsistent spacing. Off-system colors. Competing type weights. The visitor doesn’t experience any of these as problems. They can’t articulate them if you ask. But their working memory is paying the bill regardless, because the visual system has to resolve every violation before higher-level processing can proceed. This reframes what L0 actually means. The old version was “reduce what the visitor stumbles on.” The Hassin version is stronger: reduce what their brain processes, whether they’re attending to it or not. Three fonts where there should have been two. Photo crops that broke the grid without reason. A button style that drifted across pages. None of it was bad enough to complain about. All of it was paying rent in working memory that belonged to the purchase decision. I felt this before I could name it. The ADHD helps. I’ve talked about my brain in previous chapters, but it’s relevant again here. Some experiences are literally painful for me because of how taxing they are on my attention and cognitive load. Cluttered interfaces, unclear navigation, pages with too many competing elements. These aren’t just “bad design” to me. They’re physically aversive. I feel friction at a higher volume than most people. That sensitivity turned out to be a professional advantage. What most designers theorize about, I feel. When I land on a page and something is wrong with the cognitive load, I know it immediately, the same way you’d know a room is too hot before checking the thermostat. The feeling comes first. The diagnosis comes second. At Simply Smart Home, the feeling was: this page is making me work. Not hard work. Not puzzle-it-out work. Just a low-grade friction on every interaction. The product descriptions were comprehensive but didn’t prioritize. The navigation reflected the company’s internal structure rather than the customer’s mental model. The homepage presented information as if the visitor already knew what the product was and why they wanted one. None of this was broken in the traditional sense. Every link worked. Every page loaded. A conventional usability audit would have given it decent marks. But the extraneous load was constant, and it was eating the bandwidth visitors needed to evaluate the actual product. What I did wasn’t complicated. I restructured the content to answer user questions before they had to ask them. Not “anticipate their needs” in the vague, consultant-speak way that means nothing. I mean literally: what are the first three things a visitor needs to know, and does the page tell them in the first three seconds? For a digital photo frame, those questions are: What is this? What does it do for me? How do I get one? The old site answered “What is this?” with technical specifications. Resolution. Connectivity standards. App compatibility. All accurate. All useless to someone who hasn’t decided whether they care yet. I replaced that with emotional context. “Stay connected, even when you’re apart.” A photo frame isn’t a gadget. It’s a way to see your grandkid’s face every morning. It’s a way to be present in someone’s life when you can’t be there in person. That emotional frame answers “What does it do for me?” before the visitor has to figure it out themselves. Then I cleared the path. One primary action per section. Progressive disclosure for the technical details. Smart defaults on configuration options. The information the visitor needed first was where they’d see it first. The information they’d need later was one click away, not competing for attention on the same page. This is progressive disclosure in practice, and it’s one of the most powerful tools in the cognitive load toolkit. Show what’s needed now. Reveal the rest on demand. The visitor who wants to know the screen resolution can find it. But they find it after they already understand why they’d want the product, not before. The result: visitors stopped bouncing at the “what is this?” stage. They understood the product within seconds. They had bandwidth left over to actually consider buying one. The extraneous load was gone, and the intrinsic load of the purchase decision (do I want this? can I afford it? who would I give it to?) could finally do its job. That bandwidth freed up everything that came later. The perception work, the emotional messaging, the decision trail, all of it depended on visitors having enough cognitive room to receive it. I’ll cover those layers in later chapters. But none of them would have landed if L0 was still broken. The smart default is the other tool I reach for constantly, and it’s underused. A smart default means pre-selecting whatever 80% of users would choose. If your product comes in three colors and 72% of customers buy white, the white option should be pre-selected. If your checkout form asks for a shipping address and 90% of customers have the same billing address, “Same as shipping” should be checked by default. Every decision you remove is working memory you return. And the math is not linear. It’s not that 5 decisions cost 5 units and 3 decisions cost 3. Cognitive load compounds. Each decision depletes the pool, and each subsequent decision draws from a shallower pool. By the fourth unnecessary question, the visitor isn’t just annoyed. They’re cognitively depleted. The quality of every subsequent decision degrades. This is why form abandonment correlates so strongly with form length. Not because people are lazy (they’re not, they’re just human), but because every field is a withdrawal from a finite account. If you don’t need a field to complete the transaction, delete it. If you need that information later, ask later. A 3-field form and a 12-field form aren’t 4x different in length. They’re an order of magnitude different in cognitive cost. The practical diagnostic is part of the 5-Minute Perception Audit from the full Perception-First Design framework, and it’s the first test because it maps to L0. The Path Count test. Open your site in an incognito window. Pretend you’ve never seen it before. Now count every single decision between landing on the homepage and completing the primary action (buying, signing up, contacting, whatever the core conversion is). Every click. Every dropdown. Every “which one do I choose?” moment. Every fork in the road. Write that number down. Then do the same thing on your top two competitors’ sites. If your number is higher, you’re losing to friction. Not because your product is worse or your design is uglier or your copy isn’t compelling. Because every extra decision is a withdrawal from a bank account that started with maybe 4 units in it. I’ve done this exercise dozens of times with clients. The reaction is always the same. They knew the path was long, but they didn’t realize how long, because they’d stopped seeing the decisions. They’d internalized their own interface. They knew where everything was, so the navigation felt intuitive. To them. Not to the person arriving for the first time with 4 chunks of working memory and no mental model of the site’s structure. There’s a subtlety here that gets missed. Cognitive load isn’t just about counting elements or reducing choices. It’s about what those choices cost relative to what the visitor is trying to accomplish. Sweller’s distinction between intrinsic and extraneous load is the key. Intrinsic load is the cost of the actual task. If you’re selling insurance, the comparison between plans is intrinsically complex. You can’t simplify it away without lying about what you’re selling. If you’re selling t-shirts, the choice between sizes is intrinsically simple. Your job isn’t to eliminate all complexity. Your job is to eliminate every unit of complexity that isn’t the task itself. The insurance comparison should be complex, because it is complex. But the navigation to get there should be trivial. The form to request a quote should be minimal. The layout should direct attention to the comparison rather than competing with it. I think of it as a budget allocation. The visitor has 3 to 5 chunks. If the task itself takes 3, you get 0 to 2 for the interface. If you burn those remaining chunks on extraneous load, the visitor can’t do the task. They bounce, and your analytics show “high exit rate on comparison page.” But the page didn’t fail because the comparison was too hard. It failed because the interface was too expensive. I saw this at the nightclub, too. On the busiest nights, when the music was loud and the crowd was dense and the energy was high, patrons had less cognitive bandwidth for everything. If I walked up to a table with a complicated message (“Hey, last call is in twenty minutes, but the kitchen’s already closed, and if you want to settle your tab you can do it at the bar or with your server, and also the coat check closes at 1:15”), I’d get blank stares. Too many pieces of information for someone who’s four drinks in with music at 95 decibels. But if I said “Last call’s coming up, get your drinks now,” that landed every time. One piece of information. One action. The simplicity wasn’t dumbing it down. It was respecting the bandwidth that was actually available. How do you know L0 is failing? The signals are consistent. Users abandon forms halfway through. Not at the end (that’s price resistance or trust failure), but in the middle. Cognitive depletion. They ran out of bandwidth. High bounce rates on pages with many options. Not because the options are bad, but because the brain looked at the page, estimated the processing cost, and decided it wasn’t worth it. That estimation happens in milliseconds, before conscious evaluation even starts. People call support for tasks they should be able to do themselves. Not because they’re confused about the product. They’re confused about the interface. “Where do I click?” is L0’s distress signal. And the one that’s hardest to detect: quiet abandonment. No angry email. No support ticket. No feedback at all. They just leave. The analytics show a bounce, and you have no idea why, because the visitor never engaged deeply enough to articulate a reason. They were gone before the question formed. There’s a trap here that’s worth naming, because nearly every team falls into it. When something isn’t working on a website, the instinct is to add. Add a banner. Add an explanation. Add a tooltip. Add a second CTA in case they missed the first one. Every addition is an increase in cognitive load. I’ve sat in meetings where the solution to “users don’t understand our product” was to add more copy explaining it. The solution to “users aren’t clicking the CTA” was to add a second CTA. The solution to “users aren’t filling out the form” was to add helper text to every field. Each addition made rational sense in isolation. Together, they made the page louder without making it clearer. L0’s first instinct should always be subtraction. What can you remove? What can you defer? What can you pre-decide? If you’ve got 12 form fields, which 4 are actually required to complete the transaction? If you’ve got 8 navigation items, which 5 map to real user goals? If your hero section has a headline, a subhead, a description paragraph, a video, two CTAs, and a trust badge, what happens if you cut it to a headline, one sentence, and one button? Almost every site I’ve worked on got better by removing things. Not by adding them. At Simply Smart Home, the L0 fix preceded everything else. I couldn’t design first impressions for a site that was burning visitors’ bandwidth on orientation. I couldn’t optimize processing fluency when the content structure was fighting the reading order. I couldn’t build a decision trail when there were too many forks in every path. So I fixed the foundation first. Simplified the navigation. Restructured content around user questions instead of product features. Reduced choices at each step. Applied progressive disclosure so the full depth was still there for the visitor who wanted it, but it wasn’t blocking the visitor who didn’t. The changes weren’t dramatic. They never are, at this layer. Nobody looked at the new site and said “wow, that’s beautiful” or “what an experience.” They just used it. They found what they were looking for. They understood what the product was and why they’d want one. They had enough working memory left over to actually evaluate the purchase. Healthy L0 looks like that: invisible, unremarkable, and absolutely essential. The site went from presenting “very matter-of-fact statements and not much else” to answering the visitor’s questions before they had to ask. The product was the same. The audience was the same. The difference was that visitors could now actually process what they were seeing, because the interface had stopped stealing their bandwidth. This chapter is both the simplest and the most important in this book. Every other layer depends on this one. If L0 is broken, first impressions form in a state of cognitive overload, which means they form poorly. Processing fluency can’t establish rhythm when the visitor is already overwhelmed. Perception biases can’t be optimized when there’s no bandwidth to perceive. Decision architecture can’t guide when every fork is one fork too many. The dependency stack is not a metaphor. It’s operational. If you skip L0 and invest all your effort in a stunning first impression, a consistent brand system, emotionally resonant copy, and a beautifully orchestrated conversion trail, and the visitor arrives to find a 15-item navigation bar, a hero section with three competing CTAs, and a checkout form with 14 fields, none of that investment lands. The bandwidth isn’t there to receive it. I’ve seen this happen. A client comes in proud of their rebrand. New colors, new typography, new photography. Gorgeous. And the conversion rate hasn’t moved. Because the cognitive load problems underneath the new paint are exactly the same as they were under the old paint. The visitor still can’t figure out where to go. They still have too many choices. The form is still too long. Cognitive load isn’t the exciting part of design. It’s the part that makes the exciting parts possible. There’s a reason I’m covering this layer first, even though the layers above it are where the more interesting design thinking happens. You have to earn the right to do interesting work by getting the boring work right. I think about this the way a musician thinks about scales. Nobody goes to a concert to hear scales. But every piece of music you’ve ever loved was played by someone who internalized their scales so deeply that they don’t think about them anymore. The scales freed their bandwidth for expression. That’s what L0 does for design. It frees bandwidth, for the visitor and for you, so the expressive layers can land. If L0 is solid, visitors arrive with their full 3 to 5 chunks available for the experience you’ve designed. If it’s broken, they arrive already depleted, and every layer above it is performing for an audience that can’t hear it. If you take one thing from this chapter, take the Path Count test. Open your site in incognito. Count every decision from landing to conversion. Compare to competitors. If you’re higher, you’ve found money on the ground. I’m not talking about conversion rate optimization, A/B testing button colors, or redesigning the hero. I’m talking about subtraction. Remove the decisions that don’t serve the user’s goal. Pre-select the defaults. Hide what’s not needed yet. Delete the form fields that can wait. 3 to 5 chunks. That’s the budget. Spend it wisely. Next: The 50-Millisecond Verdict, on the fastest judgment your brain makes and why a theater that looked like a hobby project was losing half its ticket sales. Key Terms Cognitive Load Theory John Sweller’s 1988 framework distinguishing between intrinsic load (task complexity) and extraneous load (interface complexity). The theoretical basis for L0. Intrinsic load The complexity inherent to the task itself. Comparing insurance plans is intrinsically complex. Picking a t-shirt size is intrinsically simple. You can’t change it; the task is what it is. Extraneous load The complexity your design adds on top of the task. Confusing navigation, unnecessary form fields, competing calls to action. Every unit of extraneous load is working memory stolen from the actual task. Working memory limit Cowan’s modern estimate: 3 to 5 chunks, not Miller’s commonly cited 7±2. The hard cap on how much a visitor can hold at once. Progressive disclosure Show what’s needed now, reveal the rest on demand. The visitor who needs technical specs can find them, after they understand why they’d want the product. Smart defaults Pre-selecting whatever 80% of users would choose. Every decision you remove is working memory you return. Path Count test Count every decision from landing to conversion. Compare to competitors. If your number is higher, you’re losing to friction. References Miller (1956) The magical number seven, plus or minus two: some limits on our capacity for processing information. Psychological Review, 63(2), 81–97. Sweller (1988) Cognitive load during problem solving: effects on learning. Cognitive Science, 12(2), 257–285. Cowan (2001) The magical number 4 in short-term memory: a reconsideration of mental storage capacity. Behavioral and Brain Sciences, 24(1), 87–114. Cowan (2010) The magical mystery four: how is working memory capacity limited, and why? Current Directions in Psychological Science, 19(1), 51–57. Hassin et al. (2009) Implicit working memory. Consciousness and Cognition, 18(3), 665–678. Krug (2000) Don’t Make Me Think. New Riders. A note on how this was written: This series is AI-assisted. I provide the stories, the methodology, the case studies, and the editorial direction. AI helps me structure and draft. This is consistent with Perception-First Design’s own transparency principle: if I’m writing about perception, I should be honest about how the writing itself is produced. --- ## Chapter 5: The 50-Millisecond Verdict URL: https://aurochs.agency/writing/make-me-think/the-50-millisecond-verdict/ There’s a theater in Chicago called iO. If you know improv, you know iO. Tina Fey performed there. Amy Poehler. Chris Farley. Bill Murray showed up to do sets. It’s one of the most important comedy institutions in the country, the place that trained an entire generation of performers who went on to build modern comedy. When I got hired to redesign their website, this is what visitors saw: a hobbyist WordPress theme. Inconsistent typography. A Google Maps embed shoved into the hero section. Clip art-adjacent graphics competing with whatever show was happening that week. The calendar was a wall of text. The navigation felt like an afterthought. I asked people who’d visited the site what they thought. The answer that stuck with me: “It doesn’t look like a real theater.” That sentence. Not “the navigation is confusing.” Not “I couldn’t find the show times.” It doesn’t look like a real theater. These people knew iO was a real theater. They knew the history, the alumni, the reputation. But the website created a perception that contradicted all of that in under a second. And that perception was costing them ticket sales. In 2006, Gitte Lindgaard and her colleagues at Carleton University ran a study that changed how I think about design. They showed participants screenshots of websites for 50 milliseconds. Fifty thousandths of a second. Not enough time to read a single word. Not enough time to consciously evaluate anything. Barely enough time for a blink. Then they asked: how do you feel about this site? The responses were consistent. Not just within individuals (people gave the same ratings when re-tested) but across participants. Everyone was forming the same impressions from the same sites, at the same speed, without reading a word. Visual appeal, perceived credibility, willingness to stay. All of it evaluated before conscious thought had a chance to engage. Lindgaard’s finding was striking, but it got more extreme. In 2013, Katharina Reinecke and her team at Google Research pushed the exposure time down further. They found effects at 17 milliseconds. Seventeen. Faster than a single fixation of the eye. The brain is making aesthetic and credibility judgments about your website before the eye has finished its first saccade across the page. This isn’t a design constraint you can argue with. This is the physics of how the visual system works. Your site gets judged before anyone reads your headline, scans your nav, or scrolls past the fold. And the judgment sticks. What’s actually being evaluated in those milliseconds isn’t what most designers think. The obvious answer is “visual appeal,” and that’s partially right. But the research points to something deeper. Your visitors are evaluating your website the way they evaluate people. In 1944, Fritz Heider and Marianne Simmel showed participants a short animation of geometric shapes moving around a screen. Triangles and circles. Nothing else. And participants spontaneously described the shapes as if they were people: the big triangle was “bullying” the small one. The circle was “trying to escape.” They attributed intentions, emotions, relationships, and goals to shapes that had none of those things. This wasn’t a quirk of the sample. It’s been replicated dozens of times across cultures and age groups. The human brain is wired to see agency in anything that moves with enough complexity. And it doesn’t stop at animation. We do it with static designs, brands, products, and yes, websites. Your homepage isn’t a collection of boxes and text to the human visual system. It’s an entity being evaluated for trustworthiness and competence, the same way a stranger at a party gets sized up before anyone says hello. Mind perception, as Kurt Gray and his colleagues formalized it in 2007, operates on two dimensions. Agency: does this entity seem capable of thinking and acting? Experience: does it seem capable of feeling and sensing? We don’t literally believe a website has a mind. But the same neural machinery that evaluates whether a stranger is competent and trustworthy fires automatically when we encounter anything with enough behavioral complexity to trip the system. A polished, well-structured site signals Agency: someone capable was behind this. A site with warmth, personality, and human touches signals Experience: someone who cares made this. A site with neither signals nothing, and nothing is worse than bad. Bad at least provokes a reaction. Nothing gets scrolled past. Your website gets the same evaluation as a person walking into a room. Can I trust this? Is this competent? Does this care about me? That evaluation happens fast because the brain has been optimizing for it since before you were born. Mark Johnson and his colleagues showed in 1991 that newborns, within minutes of birth, preferentially track face-like patterns over scrambled versions of the same features. The same eyes, nose, and mouth arranged randomly get less attention than the same elements arranged in a face configuration. Neonatal face preference operates from birth. The brain arrives ready to find faces and evaluate them. By adulthood, this system is absurdly efficient. Ofir Hershler and Shaul Hochstein demonstrated in 2005 that faces “pop out” in visual search. If you scatter a bunch of objects across a screen and one of them is a face, people find the face faster than any other object regardless of where it appears. It doesn’t matter how many distractors you add. Faces bypass the serial search process that everything else is subject to. While your eyes would need to scan item by item to find a red square among blue ones, a face jumps out of the field instantly, as if it were a different category of visual object entirely. Which, neurologically, it is. This is why faces in hero sections work. Not because they’re “engaging” in some vague marketing sense. Because the visual system has a dedicated, pre-conscious detection pipeline for faces that’s been refined by millions of years of evolution. A face in your hero image gets processed before anything else on the page. It’s the first thing the brain locks onto, the first thing that generates an emotional response, and the first data point in the trust evaluation. Put a face in your hero and you’re using hardware. Leave it out and you’re asking software to do the work. But faces alone don’t close the deal. The 50-millisecond verdict is a package evaluation, and the biggest factor after faces is something I think of as visual-price coherence. The perceived quality of your design has to match the perceived value of what you’re selling. I learned this over four years at Simply Smart Home. The products were $20-35 Android tablets in wood cases. That was the hardware. But we weren’t selling tablets. We were selling “smart home decor.” Digital photo frames that connected families. Video calling devices for grandparents. I invented the category language, designed the packaging, built the brand system, and created a perception architecture where hardware that cost less than a restaurant dinner sold for $120 to $180 as a lifestyle product. And it worked. People paid it. Not because they were duped, but because the perception matched the experience. The unboxing felt premium. The interface felt warm. The product looked like it belonged on a shelf next to the Sonos and the Nest. The perception created the value, and then the product delivered on it. Perceived value mattered more than anything else sales or marketing tried. More than feature comparisons. More than competitive pricing analysis. More than the spec sheets that engineering kept wanting to lead with. The moment a potential customer saw the product (in a store, on the website, in an ad), the 50-millisecond verdict was rendering. Does this look like a $150 product or a $30 product? The answer to that question determined whether they even read the feature list. The flip side. If I’d put those same products on a website that looked like a $20 experience, no amount of feature copy would have saved the sale. The 50-millisecond verdict would have been: this is cheap. And once that verdict lands, everything downstream is fighting an uphill battle against a first impression that’s already been filed and categorized. Visual quality has to match price point. A $500 product on a $50 website creates a dissonance that the brain reads as “something is wrong here.” That dissonance isn’t conscious. The visitor doesn’t think “this website looks too cheap for this price.” They feel it. They feel uneasy, skeptical, hesitant. And then they construct a rational explanation after the fact: “I’ll think about it.” “I want to do more research.” “Something just didn’t feel right.” The feeling came first. The rationalization followed. This brings us to the aesthetic-usability effect, one of the most practically useful findings in design research. In 1995, Masaaki Kurosu and Kaori Kashimura studied ATM layouts in Japan. They created interfaces with identical functionality but different visual designs. Some looked polished. Some looked rough. Users consistently rated the attractive versions as more usable, even when the actual usability was the same. Beautiful things are perceived as more functional. Noam Tractinsky replicated this in 1997 in Israel, specifically to test whether it was a cultural effect. He expected to find that Japanese aesthetic sensibility drove the result and that Israeli users (who he characterized as more utilitarian) wouldn’t show the effect. He was wrong. The effect replicated perfectly across cultures. This isn’t a Japanese thing or an American thing. It’s a human thing. The implications are uncomfortable for anyone who separates “design” from “usability.” They’re not separate. They never were. The user’s perception of how usable something is gets shaped by how it looks before they’ve used it at all. An ugly interface that works perfectly will be perceived as harder to use than a beautiful interface with the same functionality. The aesthetic judgment bleeds into the usability judgment automatically, pre-consciously, and reliably. Think about what this means in practice. Two e-commerce sites selling the same product at the same price. One has a clean, well-typeset layout with balanced whitespace and cohesive color. The other has the same information crammed into a cluttered template with mismatched fonts. A user visiting both will perceive the first site as easier to use, even if the checkout flows are identical. The beautiful site’s “Add to Cart” button will feel more responsive, more trustworthy, more likely to work. Not because of any technical difference, but because the aesthetic verdict has already primed the usability evaluation. This effect has boundary conditions, and I want to be honest about them. If the usability failure is severe enough (buttons that don’t work, layouts that are genuinely broken), the aesthetic halo fades. And with repeated use over time, people update their assessments based on actual performance. But for first encounters and moderate interactions, which is what most web design targets, the effect is robust. And first encounters are exactly where the 50-millisecond verdict operates. Back to iO Theater. The old site failed the 50-millisecond verdict on every dimension. No faces. No visual quality signals. The Google Maps embed in the hero section was doing the opposite of what a hero should do. Instead of establishing “this is a world-class comedy institution,” it was saying “here’s how to find our building.” Useful information, wrong location. The map belonged on a contact page, not in the first thing visitors see. I rebuilt the site with three principles from this layer. First, dramatic photography. Real performers on real stages, lit like professionals, captured in moments that communicate energy and joy. Faces. The visual system’s face-detection pipeline fires immediately: these are real people having a real experience. You see performers mid-laugh, audiences leaning forward, the energy of a live show compressed into a single frame. The 50-millisecond verdict goes from “hobbyist organization” to “professional theater” before a word is read. Second, clear event navigation. The calendar went from a wall of text to a structured, scannable system. Shows organized by date, time, and type. Visual hierarchy that let you find tonight’s show in seconds. This is L0 work, but it supports the Layer 1 verdict because a well-organized interface signals competence. The structure itself communicates “this organization has its act together,” which reinforces the trust evaluation that fired in the first 50 milliseconds. Third, social proof at the speed of perception. A Bill Murray quote, visible without scrolling. Not buried in an “About” page. Right there, in the first impression. The brain processes “Bill Murray endorses this” faster than it processes the content of the quote itself. The name alone is a trust signal that operates at System 1 speed. Online ticket sales went from 50% to 75%. Same shows. Same prices. Same performers. Same venue. The only thing that changed was the perception of the venue before people clicked “Buy Tickets.” A 50% increase in conversion from changing nothing about the product. The product was always good. The website was lying about it. I’ve actually redesigned iO’s site twice. Once in 2017, and again in 2022 after new ownership and a failed agency left it worse than where I’d started. If you visit the site today, you’ll find it’s drifted again. More links, more CTAs, more competing demands on the same 50 milliseconds. Sites without ongoing design ownership revert toward entropy. Each addition made sense to someone at the time. Together they diluted the verdict. L0 doesn’t self-maintain. I want to connect this back to the activation point from Chapter 2, because the 50-millisecond verdict IS the activation point. Or more precisely, it’s the gate that determines whether an activation point is even possible. Remember the model: users arrive on autopilot. System 1 is scanning, pattern-matching, predicting. If everything matches predictions (this looks like a normal website), the user glides through without engaging. If something breaks the prediction in a way that’s interesting rather than confusing, System 2 activates. The activation point. But there’s a step before that, and it’s binary. The 50-millisecond verdict is a pass/fail gate. It doesn’t create engagement by itself. It determines whether engagement is possible. If the verdict is “this doesn’t look right” or “this looks cheap” or “I don’t trust this,” the gate closes. Everything downstream multiplies by zero. I use a formula in my framework: (Listen) x (Act + React). Listen is the gatekeeper. It’s the 50-millisecond verdict, the first impression, the moment where the brain decides whether to give this thing any attention at all. If Listen equals zero, it doesn’t matter how brilliant your copy is, how competitive your pricing is, how well-designed your checkout flow is. Zero times anything is zero. The bouncer metaphor from Chapter 1 lands here too. When I walked table to table at 1:10 AM, I was making a first impression at each table. The tone of my voice, my posture, my eye contact, whether I looked like someone delivering bad news or someone doing them a favor. That impression took about two seconds, and it determined everything that followed. If the impression was wrong, they wouldn’t hear the words. They’d hear a threat. Your hero section is your table visit. Get the first impression right and people hear what you’re saying. Get it wrong and they’ve already decided you’re not worth listening to. But that 50-millisecond verdict isn’t a single moment. It’s five of them, running in sequence. Helmut Leder and his colleagues at the University of Vienna published a paper in 2004 that mapped what happens inside those milliseconds. The first impression decomposes into five stages: perception (raw sensory input), implicit memory integration (does this match other things I’ve seen?), explicit classification (what category is this?), cognitive mastery (can my brain process what I’m seeing?), and evaluation (the felt judgment). Each stage can fail independently. This is useful for diagnosis. The 5-second test can say “this looks pretty” (perception passes) and “I wouldn’t trust it” (evaluation fails) in the same breath. When that happens, the failure sits in the middle three stages. iO Theater’s old site is a clean example. Perception passed. The WordPress theme rendered fine, the images loaded, the type was legible. But implicit integration failed immediately. Visitors had memories of what theater websites look like, and this didn’t match any of them. Explicit classification fired with the wrong answer: this is a hobby project, not a venue. By the time evaluation ran, the felt judgment was already set. Pre-verbal arousal has two directions, and the 50-millisecond verdict reads both. In 1970, Masahiro Mori, a Japanese roboticist, published a short essay that named something most people had felt without articulating. He called it the uncanny valley. Something near-human but not quite triggers an eeriness response before conscious evaluation. Mori was writing about robots, but the mechanism applies to any interface the brain evaluates for authenticity. AI-generated stock photos that look human but read as “off.” Animations with uncanny timing. Faces with subtle asymmetries. The response is pre-verbal rejection. Users can’t articulate why the page feels wrong. They just know. The inverse direction was documented by Mathieu Pessiglione and his colleagues at ICM Paris in 2007. They showed that reward cues presented below conscious threshold still drive motor behavior. Users effort-allocate toward pages that trigger subliminal reward signals they’d never name. A hover animation that lands with a tiny satisfaction spike. A color palette that reads as warm before anyone thinks about color. A micro-interaction that feels alive without being noticed. Both directions run under the 50-millisecond gate. A page that triggers uncanny response kills the first impression before the visitor can analyze why. A page that triggers subliminal reward passes the gate before the visitor knows what they’re looking at. Either way, the analytical mind shows up late. The research on all of this converges on a point that sounds simple but changes how you approach design. People don’t evaluate websites and then decide how they feel. They feel first, and the feeling becomes the evaluation. Lindgaard’s 50 milliseconds. Reinecke’s 17 milliseconds. Heider and Simmel’s triangles with personalities. Johnson’s newborns tracking faces. Kurosu and Kashimura’s beautiful ATMs that “work better.” Leder’s five stages running inside the verdict. Mori’s uncanny valley, Pessiglione’s subliminal reward. It’s all the same finding from different angles: perception precedes analysis. The emotional response fires before the rational evaluation, and the rational evaluation is shaped by the emotional response that already happened. This is why you can’t fix a first-impression problem with better copy. By the time someone reads your copy, the verdict is already in. You can confirm a positive verdict with strong copy. You can (sometimes) overcome a mildly negative one with exceptional content. But you cannot undo a visceral “this looks wrong” response with words. The words are being processed through a filter that was set before the first word was read. I know the first impression is failing when I see specific patterns in the data: high bounce rates paired with very low time-on-site (under 5 seconds). Users describing the site as “cheap” or “sketchy” without being able to articulate why. Competitors with objectively worse products winning the sale. These are Layer 1 failures. And no amount of L0 optimization (faster load times, cleaner navigation, shorter forms) will fix a problem that was decided in the first 50 milliseconds. The practical takeaway is the one I come back to on every project. Your site has 50 milliseconds. Not a design constraint. The physics of how your visitors’ brains work. In those 50 milliseconds, the visual system evaluates: Does this look competent? Does it look trustworthy? Does the quality match what I expect to pay? Are there faces? Do those faces look like people having the experience I want to have? If the answers are yes, the gate opens. The activation point has a chance to fire. The user might scroll, might read, might engage, might convert. If the answers are no, the gate closes. And no amount of A/B testing your button colors or rewriting your headline will fix a problem that was decided before the headline was read. Design the verdict first. Everything else depends on it. I do a version of this I call the squint test. Blur your eyes. Look at the page. What do you see? Not the words, not the details. The shapes. The weight. The color balance. The hierarchy. If you can’t get a sense of what the page is about and where to look first through blurred vision, the 50-millisecond verdict is going to fail. Because that blurred, pre-verbal, gestalt impression is closer to what the brain actually processes in 50 milliseconds than any detailed design review you’ll ever do. The iO site failed the squint test. Blur your eyes and it was a jumble: map, text blocks, sidebar widgets, no focal point, no hierarchy. The redesign passed it. Blur your eyes and you see a performer on a stage, a clear title, a clear path to tickets. The story of the page is legible before you read a word. Every project I take on starts here. Not with wireframes. Not with feature lists. Not with stakeholder requirements. With the question: what will the brain see in 50 milliseconds, and will it be the right thing? If the answer is wrong, nothing else I do matters. If the answer is right, everything else I do has a chance. Next: The Feeling of Truth, on why your brain believes easy-to-read fonts more than hard-to-read ones, and what happens when your brand blue is 3% off. Key Terms 50-millisecond verdict The pre-conscious aesthetic and credibility judgment that fires before a visitor reads a single word. Pass/fail gate for everything downstream. Mind perception Gray (2007). The brain evaluates entities on two dimensions: Agency (competence, capability) and Experience (warmth, feeling). Websites trigger the same evaluation as people. Neonatal face preference Johnson (1991). Newborns preferentially track face-like patterns within minutes of birth. The visual system arrives primed for faces. Visual-price coherence The perceived quality of your design must match the perceived value of what you’re selling. A $500 product on a $50 website creates dissonance the brain reads as “something is wrong.” Aesthetic-usability effect Kurosu & Kashimura (1995). Beautiful interfaces are perceived as more usable, even when functionality is identical. Replicated cross-culturally by Tractinsky (1997). Five-stage first-impression model Leder et al. (2004). The 50-millisecond verdict decomposes into five stages: perception, implicit memory integration, explicit classification, cognitive mastery, evaluation. Each stage can fail independently. Useful for diagnosing first-impression failures where perception passes but trust doesn’t. Uncanny valley Mori (1970). Near-human-but-not-quite triggers an eeriness response before conscious evaluation. Interfaces have their own version: AI-generated photos that read as “off,” animations with uncanny timing, faces with subtle asymmetries. Squint test Blur your eyes and look at the page. If you can’t tell what it’s about and where to look first, the 50-millisecond verdict will fail. References Lindgaard et al. (2006) Attention web designers: you have 50 milliseconds to make a good first impression! Behaviour & Information Technology, 25(2), 115–126. Reinecke et al. (2013) Predicting users’ first impressions of website aesthetics with a quantification of perceived visual complexity and colorfulness. CHI ’13 Proceedings. Heider & Simmel (1944) An experimental study of apparent behavior. American Journal of Psychology, 57(2), 243–259. Gray, Gray & Wegner (2007) Dimensions of mind perception. Science, 315(5812), 619. Johnson et al. (1991) Newborns’ preferential tracking of face-like stimuli and its subsequent decline. Cognition, 40(1–2), 1–19. Hershler & Hochstein (2005) At first sight: a high-level pop out effect for faces. Vision Research, 45(13), 1707–1724. Kurosu & Kashimura (1995) Apparent usability vs. inherent usability. CHI ’95 Conference Companion. Tractinsky (1997) Aesthetics and apparent usability: empirically assessing cultural and methodological issues. CHI ’97 Proceedings. Leder et al. (2004) A model of aesthetic appreciation and aesthetic judgments. British Journal of Psychology, 95(4), 489–508. Mori (1970/2012) The uncanny valley (K. F. MacDorman & N. Kageki, Trans.). IEEE Robotics & Automation Magazine, 19(2), 98–100. Pessiglione et al. (2007) How the brain translates money into force: a neuroimaging study of subliminal motivation. Science, 316(5826), 904–906. A note on how this was written: This series is AI-assisted. I provide the stories, the methodology, the case studies, and the editorial direction. AI helps me structure and draft. This is consistent with Perception-First Design’s own transparency principle: if I’m writing about perception, I should be honest about how the writing itself is produced. --- ## Chapter 6: The Feeling of Truth URL: https://aurochs.agency/writing/make-me-think/the-feeling-of-truth/ Say a sentence out loud. Any sentence. “The capital of France is Paris.” Now say it again. Notice how it feels more true the second time? Not because you verified it. Not because you learned something new. It just sits easier. The words land smoother. The thing that was already true somehow feels more true. That’s processing fluency. And it is, quietly, one of the most powerful forces in design. Reber and Schwarz ran an experiment in 1999 that should have changed every design brief written since. They presented identical trivia statements in two different fonts: one easy to read, one harder. The easy-to-read statements were judged as more true. Not more readable. Not more pleasant. More true. The content was the same. The font changed how the brain evaluated it. Think about what that means. The truthfulness of a statement, something we experience as a property of the statement itself, is influenced by the visual presentation. Your brain doesn’t separate “this is easy to read” from “this is probably accurate.” It can’t. The feeling of ease bleeds into the feeling of truth, and the user never notices the transfer happening. Alter and Oppenheimer expanded this in 2009 and found it wasn’t just about truth. Processing fluency generalizes. Easy-to-process information is judged as more true, more likable, more trustworthy, and more confident. Hard-to-process information gets the opposite treatment across the board. The brain uses processing ease as a heuristic for everything from aesthetic preference to risk assessment. If it flows, it’s good. If it stutters, something’s wrong. Dechene and colleagues confirmed the illusory truth effect in a 2010 meta-analysis across 51 studies. Repeated exposure to statements increases their perceived truth, and the proposed mechanism is processing fluency: the second time you encounter something, it processes more smoothly, and the brain reads that smoothness as a signal of accuracy. This is why brand consistency compounds. Every time someone encounters your visual system and it matches what they saw before, the fluency goes up. The trust goes up. The “feels right” goes up. This isn’t optional. It’s not a nice-to-have. It’s the architecture of how people evaluate whether something deserves their attention and their money. Try this. Pick any page on your website and squint until the text blurs out. What do you see? If everything looks equally important, nothing is. If blocks of text crowd each other with no breathing room, the brain has to work to parse the hierarchy, and that work costs fluency. Every pixel of unclear hierarchy is a tiny withdrawal from the trust account. Now think about your users who are over 40. Can they read your light grey text on a white background? If the answer is no, if basic visual processing fails at the level of individual characters, then nothing above it works. The user can’t evaluate your value proposition if they can’t decode the sentence. Processing fluency starts at the grapheme level. If the letters themselves are a struggle, you’ve already lost. But there’s a refinement to this that matters more than the headline finding. Reber and Schwarz showed that easy-to-process statements are judged as more true. Alter and Oppenheimer showed that fluency generalizes across liking and trust. The shorthand became: measure fluency, optimize for it, trust compounds. Lighthouse scores, readability benchmarks, technical audits. Forster, Leder, and Ansorge published a paper in 2013 that sharpened the rule. What drives L2 isn’t objective measurable fluency. It’s subjective felt fluency. A font that is objectively readable can still feel clunky if its rhythm fights the surrounding layout. A site that passes every technical readability benchmark can still feel heavy. The Lighthouse score and the lived experience are correlated but not identical, and L2 lives in the gap. This is why two sites with the same contrast ratios, the same load times, the same typeface metrics can feel completely different. One glides. The other drags. The measurements don’t see it. The nervous system does. Design for the feeling, not the metric. I learned this the slow way, over four years, with a brand called Simply Smart Home. When I started working with SSH in 2020, there was no brand system. I don’t mean the brand system was weak. I mean it didn’t exist. Every piece of marketing collateral was built from scratch, each one with slightly different fonts, slightly different colors, slightly different layouts. The Amazon listings looked different from the website. The website looked different from the packaging. The packaging looked different from the social media. It was all the same company, selling the same products, and none of it felt like it belonged together. The products were good. Digital photo frames, smart home devices, thoughtful tech for families. Price point was $100 to $180. But the presentation said $15 knockoff. Template website. Stock photography. Feature-spec copy. Nothing that said “this is a premium product made by people who care about design.” The gap between product quality and perceived quality was enormous. And that gap was costing them everything. Processing fluency isn’t something you fix with a weekend redesign. It’s infrastructure. It’s the systematic work of making every touchpoint feel like it belongs to the same family, so the brain never has to work to reconcile conflicting signals. Year one was triage. I built the first real style guide: two typefaces (not five), a primary color palette (not whatever looked good that day), and templates for the most common collateral. The Amazon listing images alone had been produced by at least three different designers with three different aesthetic vocabularies. A customer who saw the product on Amazon, then visited the website, then opened the packaging was encountering what felt like three different companies. Each transition was a fluency violation. Each violation was an erosion of trust. It wasn’t exciting work. It was the typography equivalent of pouring a foundation (and about as glamorous). Nobody photographs foundations. But nothing stands without one. Year two, I started building the seasonal creative system. SSH sold gift-oriented products, which meant their visual identity had to flex across Valentine’s Day, Mother’s Day, Father’s Day, Christmas, and everything in between. Without a brand system, each season’s creative was a reset. New designer, new look, new everything. The Valentine’s Day campaign looked like it came from a different company than the Christmas campaign. With a system in place, each season became a variation on a theme. Same typography. Same layout grid. Different color palette, different photography mood, but always recognizably SSH. By year three, the brand system was mature enough to support sub-brands. ClicSmart, the smart home security line. myLuma, the digital frame line. Disney-licensed products that had to live inside Disney’s brand guidelines and SSH’s simultaneously. Brookstone partnership products with their own visual constraints. Each sub-brand had its own personality but shared the family resemblance. You could tell they were siblings. That family resemblance is processing fluency at the brand architecture level: when a customer encounters a new SSH product line, the visual system says “you already know us, this is just a new room in the same house.” Year four, the system was infrastructure. Seasonal palettes generated from the master system. Marketing teams could produce collateral without me touching every piece because the rules were documented, the templates existed, and the visual language was stable enough to be extended without breaking. Revenue tripled. Same products. Same factories. Same price points. The perception changed. The thing most people miss about processing fluency: the violations that hurt the most are the ones that are almost right. There’s a paper from Bujack and colleagues in 2022 that changed how I think about color in design. They demonstrated that perceptual color space is non-Riemannian, which is a technical way of saying something designers need to understand: color perception doesn’t work the way our digital tools suggest it does. In CSS, if you pick a blue and then shift it 3% in one direction, you’d expect the perceived difference to be proportional to the shift. It’s not. Color perception is non-linear and asymmetric. Some small shifts are imperceptible. Others, equally small in the math, are jarring. And the worst part (the part that bit me multiple times before I understood why): near-miss color deviations are disproportionately more disruptive than far-miss deviations. Think about it this way. If your brand blue is #2563EB and someone uses #2866F0 on a banner, that’s a near miss. The brain catches it. Not consciously, not in a way the user can articulate, but the prediction error fires. “This is almost the right blue but something is off.” The brain can’t resolve whether it’s the same or different, so it keeps processing. Fluency drops. Trust drops. The feeling of “this is right” gets replaced by a subtle, unnameable sense that something is wrong. Now imagine using a completely different color. Red instead of blue. The brain doesn’t try to reconcile that. It processes it as a different thing entirely. Different category, different prediction, no conflict. A far miss is less disruptive than a near miss because the brain doesn’t waste energy trying to match it to the existing prediction. This is why brand police who obsess over exact hex codes aren’t being pedantic. They’re protecting processing fluency at the level where violations are most expensive. And it’s why the move toward perceptually uniform color spaces like OKLCH matters. The tools need to model how the eye actually works, not how the math says it should. I keep coming back to the SSH tagline as a processing fluency case because it demonstrates something that’s hard to explain in the abstract. I created “smart home decor” as a product category. It wasn’t a category that existed. Nobody was using that phrase. But it felt right, and that’s the point. The products were digital photo frames and smart displays. “Consumer electronics” was the accurate category. But consumer electronics processing triggers a whole set of predictions: comparison shopping, spec sheets, cheaper alternatives on Amazon, disposable tech. None of that matched the product’s actual value proposition, which was emotional connection between family members. “Smart home decor” did something different. It triggered decor predictions: aesthetic value, fits your living space, worth displaying, permanent rather than disposable. The processing was fluent because the category name matched the experience of owning the product, even though it didn’t match the product’s technical classification. I wasn’t inventing a false category. I was finding the one that processed more fluently against the truth of what the product actually did for people. A digital frame sits on your shelf next to other decor. It displays photos of your family. You pick the frame style that matches your room. That’s decor behavior. Calling it “consumer electronics” was the processing fluency violation, not the other way around. Two years after I coined the phrase, Aura Frames started using nearly identical positioning. When your competitor steals your language, you know you found the category that the brain was already looking for. This is processing fluency applied to naming, which most people don’t think of as a design decision but absolutely is. Every name, every category label, every tagline is a processing event. The brain encounters the words and either processes them fluently against existing mental models or stumbles. “Consumer electronics” processed fluently against the wrong mental model for SSH’s products. “Smart home decor” processed fluently against the right one. Same product. Different cognitive path. The path that matches reality converts better. Typography is the most underrated processing fluency tool in design. George Lakoff and Mark Johnson published Metaphors We Live By in 1980, and it’s one of those books that, once you read it, changes how you see language forever. Their core finding: abstract thought isn’t abstract. It’s built on bodily experience. We don’t just use physical metaphors for convenience. We think through them. “Happy is up, sad is down” isn’t poetic decoration. It maps to actual physical posture in those emotional states. “Argument is war” isn’t a creative choice. It structures how we actually experience disagreement (attacking positions, defending claims, winning or losing). This matters for design because the way we describe visual and auditory experience reveals how the brain processes it. We use sensory words, the cross-modal kind, for everything. A color is “warm” or “cool.” A font is “heavy” or “light.” A layout is “clean” or “cluttered.” A brand voice is “rough” or “smooth.” None of these are literal. Colors don’t have temperature. Fonts don’t have mass. But the brain processes them through the same sensory channels it uses for physical experience. Charles Spence demonstrated this empirically in 2011: people consistently match high-pitched sounds with small, bright objects located high in visual space. Low-pitched sounds match large, dark objects located low. These aren’t learned associations. They’re cross-modal correspondences baked into how the nervous system integrates information from different senses. When the correspondences align, processing is fluent. When they conflict, processing stutters. This is why typography isn’t decoration. It’s the music layer of design. The spacing cadence, the weight distribution, the rhythm of headings to body text to whitespace. These create a temporal pattern that the brain processes the same way it processes a melody. Consistent rhythm equals fluent processing equals trust. Inconsistent rhythm equals stuttering equals doubt. Two fonts maximum. Not because it’s a rule someone invented. Because each additional typeface is a new voice in the conversation, and the brain has to work to integrate voices. Two voices can have a dialogue. Three becomes a panel. Four becomes noise. The processing load compounds with each addition, and the fluency cost is steeper than most designers realize because they’ve already adapted to their own designs. The fresh eye hasn’t. I think about this every time I see a site that uses a different font for the logo, the navigation, the headings, the body, and the call-to-action buttons. Five typefaces means five visual voices. The designer might have a rationale for each one. But the user’s brain doesn’t care about the rationale. It’s hearing five people talking at once and trying to figure out who to listen to. That cognitive work is invisible to the designer who chose the fonts deliberately, but it’s real for every first-time visitor. The McGurk effect, discovered by McGurk and MacDonald in 1976, is maybe the most visceral demonstration of how cross-modal processing works. They showed people a video of someone mouthing one syllable while the audio played a different syllable. The result: people heard a third syllable, neither the visual one nor the auditory one, but a fusion of both that didn’t exist in either input. The brain didn’t choose between eyes and ears. It merged them. And the merge produced something neither channel sent. This matters for design because it proves that users don’t process your visual design and your copy and your interaction patterns separately and then combine them. The brain integrates across channels continuously, producing a fused perception that’s different from any individual element. Your typography style, your color palette, your image quality, your copy voice, your interaction timing: these aren’t separate design decisions. They’re inputs into a single perceptual fusion. When those inputs are congruent, when the typography feels the same weight as the copy voice, when the color temperature matches the brand personality, when the image quality matches the price point, the fusion is clean. Processing is fluent. The thing feels true. When they’re incongruent, the fusion breaks. And the result isn’t “the user notices the typography is wrong.” The result is a vague, unnameable sense that something is off. The user can’t point to it. They’ll tell you the site seems fine if you ask them. But they won’t convert. The prediction error is happening below the threshold of conscious access, and it’s driving behavior that neither the user nor the designer can explain through surface-level analysis. This is why “fix the copy” or “fix the colors” is almost always the wrong diagnosis when conversion is underperforming. The problem isn’t in any single channel. It’s in the fusion. The channels are sending conflicting signals, and the brain is producing a fused perception that feels wrong without any individual element being obviously broken. You have to audit the congruence across channels, not within them. A $500 product on a $50 website. That’s the most common processing fluency violation I see, and it’s the one that costs the most revenue. Think about what happens cognitively. The user arrives with a price expectation (or discovers the price early in the visit). The brain immediately generates a prediction about what a product at that price point should look like, feel like, be presented like. This prediction is built from every other $500 purchase they’ve made. The Apple Store. A nice restaurant’s website. A boutique hotel’s booking page. That’s the reference class. Now they’re looking at a template site with stock photography and three different font sizes that don’t relate to each other. The prediction error fires hard. Not “this site is ugly.” The evaluation is faster and more damaging than that: “something is wrong here.” The price says premium. The presentation says cheap. The brain can’t reconcile those signals, so it defaults to the safer interpretation. The product probably isn’t worth $500. The reverse happens too. A $15 product on a site that looks like it costs $500 creates a different kind of dissonance: “this is too good to be true.” The brain predicts a catch, a scam, a bait-and-switch. Processing fluency demands alignment. The visual presentation must match the price point, which must match the product quality, which must match the brand voice, which must match the customer service experience. Every mismatch is a prediction error. Every prediction error costs fluency. Every fluency loss costs trust. I see this constantly in my consulting work. A SaaS company charges $200 per month but their marketing site uses free templates and system fonts. A luxury goods retailer has beautiful packaging and a website that looks like it was built in 2014. A professional services firm charges $300 an hour and sends proposals in plain-text email. In each case, the people running the business don’t see the mismatch because they know the value of what they’re selling. But the customer doesn’t know that yet. The customer is processing the presentation, and the presentation is sending a signal that conflicts with the price. This is what I was fixing at SSH for four years. The products were $100 to $180. Good products, thoughtful engineering, real value. But the visual presentation was saying $30 clearance bin. Every time a potential customer landed on the site, their brain was running the price-to-presentation comparison, and the site was failing. The brand system wasn’t cosmetic improvement. It was the systematic elimination of price-perception mismatches across every touchpoint. The practical takeaway sounds boring, but it compounds faster than anything else in design. Boring consistency beats exciting inconsistency. Every time. I know that’s not what designers want to hear. The exciting inconsistency is fun to design. The unexpected color. The surprising typeface. The layout that breaks the grid for emphasis. And there are times when controlled prediction errors create activation points (that’s Chapter 2 territory). But the ratio matters enormously. For every one place where you strategically break a pattern, there should be fifty places where the pattern holds. Processing fluency is not a one-time fix. You don’t redesign a brand and walk away. It’s infrastructure that requires maintenance, enforcement, and gradual refinement. Each year at SSH, the system got tighter. The seasonal palettes got more systematic. The sub-brand guidelines got more specific. The templates got more detailed. And each year, the perceived quality of the brand went up, even when the products themselves didn’t change much. That’s what compounding processing fluency looks like. Not a dramatic before-and-after. A slow, steady increase in how trustworthy, how professional, how right everything feels. The individual improvements are often invisible. A slightly better letter-spacing ratio. A more consistent padding rhythm across pages. The photography style guide getting one more rule that prevents a common mistake. None of these are portfolio pieces. All of them are moving the needle. Tripling revenue over roughly two and a half years wasn’t one design decision. It was the compound effect of making everything feel like it belonged together, so the brain could stop working to reconcile conflicting signals and start evaluating the product on its merits. Processing fluency isn’t glamorous. It’s the plumbing of perception. But when the plumbing works, everything above it works better. And when it doesn’t, nothing you build on top of it will fully land. If you want to test this on your own brand tomorrow, here’s where to start. Pick one touchpoint, any touchpoint, and compare it to another. Your website and your business card. Your homepage and your about page. Your product packaging and your Amazon listing. Do they feel like they came from the same place? Not “do they use the same logo.” Do they feel the same? Same visual weight. Same typographic voice. Same level of quality. If they don’t, that’s your first processing fluency violation, and it’s costing you more than you think. Next: The Gap, on what happens when your marketing team describes what the product does instead of what having it feels like, and how a single tagline shift tripled revenue. Key Terms Processing fluency Reber & Schwarz (1999). Easy-to-process information is judged as more true, more likable, more trustworthy. The brain uses processing ease as a heuristic for credibility. Illusory truth effect Dechêne et al. (2010). Repeated exposure increases perceived truth via processing fluency. Why brand consistency compounds trust over time. Subjective felt fluency Forster, Leder & Ansorge (2013). What drives L2 outcomes is the felt experience of fluency, not the objective measurable fluency. Lighthouse scores and lived fluency are correlated but not identical; L2 lives in the gap. Near-miss color deviation Bujack et al. (2022). Their finding that perceptual color space is non-Riemannian implies that near-miss brand color deviations are disproportionately more disruptive than far-miss deviations. The brain can’t resolve whether it’s the same or different. Cross-modal correspondence Spence (2011). The brain maps sensory channels onto each other (warm colors, heavy fonts). When correspondences align, fluency increases. When they conflict, processing stutters. McGurk effect McGurk & MacDonald (1976). Cross-modal fusion: the brain merges visual and auditory input into a third percept that exists in neither channel. Design channels fuse the same way. Visual-price coherence The perceived quality of your design must match the perceived value of what you’re selling. A $500 product on a $50 website creates dissonance. References Reber & Schwarz (1999) Effects of perceptual fluency on judgments of truth. Consciousness and Cognition, 8(2), 338–342. Alter & Oppenheimer (2009) Uniting the tribes of fluency to form a metacognitive nation. Personality and Social Psychology Review, 13(3), 219–235. Dechêne et al. (2010) The truth about the truth: A meta-analytic review of the truth effect. Personality and Social Psychology Review, 14(2), 238–257. Forster, Leder & Ansorge (2013) It felt fluent, and I liked it: Subjective feeling of fluency rather than objective fluency determines liking. Emotion, 13(2), 280–289. Bujack et al. (2022) The non-Riemannian nature of perceptual color space. PNAS, 119(18). Lakoff & Johnson (1980) Metaphors We Live By. University of Chicago Press. Spence (2011) Crossmodal correspondences: A tutorial review. Attention, Perception, & Psychophysics, 73(4), 971–995. McGurk & MacDonald (1976) Hearing lips and seeing voices. Nature, 264(5588), 746–748. A note on how this was written: This series is AI-assisted. I provide the stories, the methodology, the case studies, and the editorial direction. AI helps me structure and draft. This is consistent with Perception-First Design’s own transparency principle: if I’m writing about perception, I should be honest about how the writing itself is produced. --- ## Chapter 7: The Gap URL: https://aurochs.agency/writing/make-me-think/the-gap/ I had a meeting where someone on the team told me my tagline was wrong. Not wrong as in grammatically wrong or factually wrong. Wrong as in they didn’t like it. The product was a digital photo frame, a smart frame you could send pictures to from your phone so your parents or grandparents could see new photos of the kids without having to figure out how to use an iPad. And my tagline was: “Stay connected, even when you’re apart.” The pushback wasn’t articulate. It wasn’t “this doesn’t match our brand positioning” or “our user research says otherwise.” It was more like a feeling. A shrug. The team couldn’t put their finger on what was wrong with it, but they knew they didn’t love it. Partly because they hadn’t thought of it. Partly because nobody on the team could come up with anything in the same league, which made the discomfort worse. Nobody likes being outwritten on their own product. I recognized the dynamic from improv. One of the first things my teacher taught us: if you don’t come up with an idea you like and someone else does, you get a negative pull toward shooting it down. Not because of anything wrong with the idea. Because you didn’t think of it first, and that makes you feel something about yourself. The fix is the same as every other social instinct we were learning to override in that class. Notice it, name it, choose yes-and instead of no-but. I kept it anyway. Campaigns with that tagline outperformed everything else the company ran. Not by a little. By enough that the conversation about whether it was the right tagline just stopped. The numbers ended the argument. Two years later, Aura Frames, one of the biggest competitors in the space, started using nearly identical messaging. “Stay connected” language. The same emotional architecture. They didn’t arrive at it by coincidence. They arrived at it because it works, and when something works that well, competitors find it. My rule of thumb: when your competitor steals your tagline, you found the emotional core. That company, Simply Smart Home, tripled its annual revenue. Disney licensing. Costco pallet placement. Walmart shelf space. The product didn’t change. It was the same hardware the whole time. Same screens, same app, same wood casing. What changed was the perception. And the perception changed because I stopped telling people what the product did and started telling them what having the product would feel like. This is Layer 3: Perception Bias Optimization, and it’s the most politically difficult principle in the entire framework. Because it means telling stakeholders that their instincts about their own customers may be wrong. Not their instincts about the product, not their instincts about the market. Their instincts about what their customers actually care about and why they buy. A hard conversation to have. People identify with their products. They built the thing. They know the specs, the engineering trade-offs, the supply chain nightmares. And when someone walks in and says “none of that is what your customer cares about,” it feels dismissive. Even when it’s true. The core insight comes from a 1977 study by Richard Nisbett and Timothy Wilson that should be required reading for anyone who designs anything. They demonstrated, across multiple experiments, that people are often completely unaware of what actually influenced their decisions. When asked to explain their choices, subjects generated plausible post-hoc rationalizations, stories that sounded reasonable but had nothing to do with the actual cause. The subjects weren’t lying. They genuinely believed their explanations. They just didn’t have access to the real process. Kahneman formalized this in Thinking, Fast and Slow (2011) with the System 1 / System 2 framework I introduced in Chapter 2. System 1 makes the decision. Fast, automatic, emotional, below the surface. System 2 constructs the explanation. Slow, deliberate, rational, satisfying to articulate. This matters for design: survey data captures the System 2 rationalization. Analytics captures the System 1 decision. The gap between them is where perception problems hide. Ask a customer why they bought a digital photo frame and they’ll tell you about the screen resolution, the app reviews, the price comparison they did. That’s the story. System 2 constructed it after the fact because humans need reasons for their behavior. But the actual purchase decision, the moment they went from “maybe” to “yes,” happened in System 1. It happened because the brand made them feel something. Connected to their family. Less guilty about distance. Like a thoughtful gift-giver. Ask a customer why they didn’t buy, and the gap gets even wider. “The price was too high.” “I wasn’t sure about the reviews.” “I’ll think about it.” All perfectly rational explanations. All System 2 reconstructions of a System 1 verdict that already fired: “This doesn’t feel right.” Maybe the site looked cheap. Maybe the messaging was all specs and no soul. Maybe the product photos showed hardware instead of humans. The customer felt a perception barrier, bounced, and then constructed a reason that had nothing to do with what actually happened. I’ve seen this on every project I’ve worked on. The stated reason and the actual reason almost never match. Ran Hassin extended this in a 2013 review I introduced in Chapter 2 as the pre-verbal arousal backbone. It should unsettle anyone who thinks the unconscious is limited. The old frame: the unconscious handles fast, automatic processing. Pattern recognition, emotional response, snap judgments. The important stuff, the deliberation, the reasoning, the decision-making, all happens in System 2 where we can watch it. That frame is wrong. Hassin reviewed the evidence and concluded that the unconscious performs every high-level cognitive function previously assumed to require awareness. Decision-making. Goal pursuit. Cognitive control. Self-regulation. All of it. The conscious, deliberate mind is not where these things happen. It’s where we learn they happened. This reframes what “the gap” actually is. The gap between survey data and analytics isn’t the gap between what the user knows and what they’re hiding. It’s the gap between what the user’s conscious mind can report and what the rest of their cognition is doing without it. Pre-verbal arousal isn’t an emotional residue on top of conscious thought. It’s the primary processing mode. Conscious thought arrives late, and often just to write the story. The Simply Smart Home story is the clearest example I have, so let me tell it fully. When I came on, the company was doing modest revenue. The product was functional, the price was competitive (for the category), and the team was working hard. But the presentation told a completely different story. The site was a template. The branding was scattered. The marketing was entirely feature-focused: screen resolution, WiFi connectivity, photo capacity, app compatibility. Every layer of perception was failing. The site didn’t answer a visitor’s basic questions (cognitive load problem). No faces, no warmth, no emotional context in the imagery (first impression problem). A template site at a premium price point created a quality mismatch that screamed “knockoff” (processing fluency problem). Feature specs instead of emotional connection (perception bias problem). And the homepage was structured around the company’s org chart, not the customer’s goals (decision architecture problem). But the perception bias problem, Layer 3, was doing the most damage. Because it infected everything upstream. The feature-focused messaging shaped the photography choices (product shots instead of lifestyle shots). The photography shaped the visual identity (cold, technical, sterile). The visual identity shaped the first impression (this feels like a $15 tablet, not a $150 smart home product). Each layer was failing partly because the perception bias at the messaging level was wrong. The product was actually $20 to $35 worth of hardware in a wood casing, sold for $100 to $180. That’s not a criticism (I’m not dunking on the product here, I worked with it for four years because I believed in what it did). That’s every consumer electronics product. The gap between hardware cost and retail price is called brand. It’s called perceived value. And perceived value isn’t a lie. It’s the emotional architecture that makes the price feel fair. So I built one. I created “smart home decor” as a category. Not “digital photo frame.” Not “WiFi-enabled picture display.” Smart home decor. A thing that belongs in your living room because it makes the room feel warmer, more connected, more like a home where people who love each other actually live. That phrase repositioned the entire product. It wasn’t a gadget anymore. It was decor. It wasn’t competing with tablets. It was competing with the framed photos on your mantle, except these ones updated automatically when your daughter posted new pictures of the grandkids. And the tagline, “Stay connected, even when you’re apart,” told you why you wanted it. Not what it did. Why you wanted it. Because you miss people. Because distance is hard. Because seeing a new photo of someone you love, appearing on a frame in your kitchen without you having to do anything, makes you feel closer to them. Perception bias optimization. Finding the emotional core that drives the actual purchase decision, the System 1 decision, and making it the centerpiece of the brand. Instead of burying it under specs. The team didn’t like it at first. That’s the politically difficult part. The people closest to the product are the least equipped to see it from the customer’s perspective, because they know too much. They know about the WiFi chip and the display resolution and the firmware update process. They think those things matter to the buyer because those things matter to them. This is one of the hardest things I have to communicate to stakeholders: it’s not for you. It’s for them. I’ve said those exact words. Looked someone in the eye and told them: this product isn’t for you. You’re not the customer. Your feelings about the messaging are real, but they’re not data about what your audience needs to hear. There’s a famous comparison that illustrates this perfectly, and I use it in almost every client conversation. The iPod and the Zune were the same product category. Both were portable music players. Both played MP3s. The Zune actually had some features the iPod didn’t, like wireless sharing between devices and a larger screen on certain models. From a pure feature comparison, the Zune was competitive. In some ways it was better. The iPod won. Completely. Overwhelmingly. Not because of the hardware. Because of the perception architecture. Apple’s marketing never told you what the iPod did. It told you what having an iPod made you feel. Freedom. Self-expression. Your entire music library in your pocket, available whenever you wanted it, wherever you were. The silhouette ads, the white earbuds as a cultural signifier, the integration with iTunes as a music discovery experience. Every touchpoint reinforced the same emotional message: this product is about who you are, not what it does. Microsoft’s marketing for the Zune told you what it did. Screen size. Storage capacity. Sharing features. FM radio. All real features. All things that System 2 can evaluate and compare. And all completely irrelevant to the System 1 decision that was already made the moment someone saw a white earbud cord on a stranger in a coffee shop and felt something. Same hardware category. Radically different perception architecture. One became a cultural icon. The other became a case study in how to lose. The gap between what users say they want (features, specs, rational comparison) and what actually drives their behavior (feeling, identity, emotional resonance) is the gap. It’s the space where perception bias operates, invisible to everyone except the people paying attention. Nisbett and Wilson’s 1977 finding plays out in a design context like this. You run a user survey. You ask people what matters to them when choosing a product in your category. They tell you: price, features, reviews, reliability. All rational. All System 2. All true, in the sense that people genuinely believe these are the factors driving their decisions. Then you look at your analytics. The highest-converting landing page has the fewest feature mentions and the strongest emotional hook. The product photo that drives the most clicks shows a person using the product, not the product alone. The email subject line with the highest open rate doesn’t mention a single spec. The customers who pay full price without waiting for a sale came in through channels that emphasized lifestyle, not value. The survey says one thing. The data says another. The gap between them is where your actual design problem lives. Most teams resolve this by trusting the survey. It’s explicit. It’s quotable. You can put it in a slide deck and say “customers told us they want X.” It feels like evidence because it comes from the customer’s own mouth. But it’s not evidence of what they want. It’s evidence of what they think they want. And those are different things. I’m not saying surveys are useless. They’re useful for what they actually measure: conscious rationalizations, stated preferences, articulated pain points. That data matters. But treating it as ground truth about purchase behavior is like asking someone why they married their spouse and taking the answer at face value. The real reasons are deeper, messier, and largely inaccessible to the person giving the explanation. There’s a specific mechanism behind the gap that most people miss. In 2015, Seckler and colleagues at the University of Basel published a paper that measured how trust actually forms on websites. Their finding reframes what L3 intervention looks like. Trust isn’t a rational output. It’s a perceptual one. Color coherence, typographic quality, layout consistency: these drive trust judgments before any content is read. The visual system flags the presentation as high-quality or low-quality, and everything the user reads afterward is filtered through that prior. Which means: users don’t distrust your site because of a weak argument. They distrust it because the visual system already flagged the presentation as cheap, and your carefully written argument is now showing up inside a frame that says “this is probably a scam.” This is why branding investment compounds. Each visit that passes the visual trust gate lowers the cognitive cost of the next visit. Each visit that fails raises the bar for the argument that follows. The practical implication: when trust data is poor, fix visual coherence before rewriting copy. The reader’s nervous system is telling them “something is off” long before they’ve gotten to your words. No argument survives that verdict. There’s a specific pattern I watch for that tells me a company has a Layer 3 problem. It shows up the same way every time. Marketing focuses on features. The homepage leads with specs, capabilities, technical details. The assumption is that if people understand what the product does, they’ll buy it. But understanding and wanting are different processes. I can understand exactly what a product does and feel nothing. Stakeholders love the design, but customers don’t buy. The team looks at the site and says “this is great, this represents us well.” And they’re right, it does represent them. It represents how the company sees itself. But the customer doesn’t care how the company sees itself. The customer cares how the product makes them feel. The design passes the internal test and fails the external one because the internal test is checking the wrong thing. Copy describes what the product does, not why you’d want it. “WiFi-enabled digital photo frame with 10.1-inch display, 16GB storage, and free cloud service.” That’s a product description. It’s accurate. It’s also dead on arrival, because it activates System 2 (evaluation, comparison, deliberation) when it should be activating System 1 (feeling, desire, identity). Nobody lies awake at night wishing they had a WiFi-enabled digital photo frame. People lie awake wishing they felt closer to the people they love. When all three of these patterns show up together, I know the problem isn’t the product, the price, or the funnel. The problem is perception bias. The company is designing for what they know about the product instead of what the customer feels about the product. And the fix isn’t better features or lower prices. It’s better perception architecture. The SSH story has a payoff that I think about a lot. After the rebrand, after “stay connected, even when you’re apart,” after “smart home decor” as a category, after the emotional photography and the lifestyle-first marketing, the company tripled revenue. Disney wanted licensing deals. Costco gave them pallet placement. Walmart put them on shelves. None of those retailers cared about the WiFi chip. They cared about perceived value. They cared that the brand told a story customers would respond to. They cared that the packaging looked like it belonged next to $200 products instead of next to $15 knockoffs. The product was the same product. The engineering team didn’t redesign the hardware. The supply chain didn’t change. The price didn’t drop. What changed was the story, and the story changed what people were willing to pay, where they were willing to buy it, and which companies were willing to put it on their shelves. I put enough lipstick on the pig that was Simply Smart Home to persuade enough people on the fence that it was worth the buy. I’m honest about that framing. The product was fine. Not remarkable. Fine. The perception architecture made it remarkable. And the people who were almost going to buy, the fence-sitters who needed one more reason to say yes, those people started buying. That’s the conversion impact of perception bias optimization. You’re not creating demand out of nothing. You’re removing the perception barriers that were blocking demand that already existed. Perceived value meant more than anything else sales or marketing tried to do to increase sales. That’s a direct quote from my own notes. I don’t say things like that unless I mean them. There’s a subtlety here that I want to name because it matters for how you apply this. Perception bias optimization is not lying. It’s not inventing qualities the product doesn’t have. It’s not manipulating people into buying something they don’t want. It’s finding the true emotional value of the product and making it visible. “Stay connected, even when you’re apart” is true. The digital frame does connect people across distance. That’s what it actually does, at the emotional level, for the person who uses it. I didn’t invent that value. I identified it. The product team was too close to the hardware to see it. They saw WiFi chips and screen resolutions. I saw a grandmother in Ohio looking at photos of her grandkids in California and feeling less alone. The ethical line is clear to me: perception bias optimization means aligning the perception with the best true version of the product. Making the emotional reality visible. Not fabricating an emotional reality that doesn’t exist. When the perception you build doesn’t match the experience, people find out. They always find out. And the trust damage is worse than if you’d never built the perception in the first place. Layer 3 only works when it’s grounded in something real. I want to come back to the team resistance for a moment, because if you’re going to practice this, you need to be ready for it. The people closest to the product will almost always push back on perception-first messaging. Not because they’re wrong about the product, but because they’re evaluating the messaging from the wrong perspective. They’re asking “does this accurately describe what we built?” when the right question is “does this make someone want what we built?” Those are fundamentally different questions. The first is an accuracy check. The second is a desire check. And the answers are often in tension, because the things that make someone want a product (emotional connection, identity alignment, aspiration) are not the things that accurately describe it (specs, capabilities, technical details). I’ve had this conversation dozens of times. The stakeholder says “but we need to mention the 16GB storage.” And I say “does your customer know what 16GB means in the context of photo storage?” They don’t. “Does your customer care?” They don’t. “What does your customer care about?” That their mom can see pictures of the grandkids. The stakeholder knows this. At some level, they’ve always known it. But saying it out loud feels like a betrayal of the engineering effort. Like all those months of firmware development and hardware sourcing don’t matter. They do matter. They matter enormously, because without the engineering the product doesn’t work. But the engineering is the table stakes. It’s the foundation. The customer assumes it works. What the customer needs help with is wanting it. The gap. What the team values and what the customer values are different things, and the team’s instinct is to communicate what they value. Perception bias optimization means having the discipline to communicate what the customer values instead. One more thing about the SSH story that I think is underappreciated. The revenue growth wasn’t a one-time spike from a campaign launch. It built over roughly two and a half years. Every year, the brand system got more refined. The photography got more emotionally resonant. The messaging got more precise. The seasonal palettes, the sub-brand mood boards for Disney and Brookstone and Costco co-branding, the collateral for sales teams, the packaging updates. All of it compounding. Processing fluency isn’t a one-time fix. It’s brand infrastructure that compounds. And perception bias optimization sits right on top of it, benefiting from every improvement in visual consistency and brand coherence. Each year, the perceived quality matched the price point more precisely. Price objections decreased. Vendors got more interested. The flywheel spun faster. That growth wasn’t a redesign payoff. It was a perception compound interest payoff. The redesign set the foundation, and the compounding did the rest. So take this from the chapter. Your customers are making decisions you can’t see, for reasons they can’t explain. The gap between what they tell you and what they actually do is not noise. It’s signal. It’s the most important signal in your data, and most teams ignore it because the survey results are easier to act on than the behavioral patterns. Find the emotional core. Not the features, not the specs, not the technical differentiator. The feeling. What does someone feel after they use your product? What problem does it solve that they can’t articulate but absolutely feel? That’s your message. That’s your brand. That’s your entire perception architecture. And be ready for the room to push back. Because the emotional core almost never sounds like what the team expects to hear. It sounds too simple. Too soft. Too far from the engineering. “Stay connected, even when you’re apart” doesn’t mention screens or apps or cloud storage. It mentions a human feeling. And human feelings are what drive human decisions. It’s not for you. It’s for them. Next: You Wouldn’t Vibe Code a Car, on why memes work, why video doesn’t, and what the credibility gap in AI-assisted development actually looks like. Key Terms Perception Bias Optimization (Layer 3) Finding the emotional core that drives the actual purchase decision (System 1) and making it the centerpiece of the brand, instead of burying it under specs. System 1 / System 2 gap Kahneman (2011). System 1 makes the decision fast and emotionally. System 2 constructs the rational explanation after the fact. Survey data captures System 2. Analytics captures System 1. Post-hoc rationalization Nisbett & Wilson (1977). People generate plausible explanations for their decisions that have nothing to do with the actual cause. The stated reason and the actual reason almost never match. Visual trust Seckler et al. (2015). Color coherence, typographic quality, and layout consistency drive trust judgments before any content is read. Trust is a perceptual output, not a rational one. When trust data is poor, fix visual coherence before rewriting copy. Emotional core The feeling that drives the actual purchase decision. Not the features, not the specs. What someone feels after they use your product. Smart home decor A category Stefan invented for Simply Smart Home that repositioned digital photo frames from consumer electronics to home decor, triggering different (and more accurate) mental models. References Nisbett & Wilson (1977) Telling more than we can know: Verbal reports on mental processes. Psychological Review, 84(3), 231–259. Kahneman (2011) Thinking, Fast and Slow. Farrar, Straus and Giroux. Hassin (2013) Yes it can: On the functional abilities of the human unconscious. Perspectives on Psychological Science, 8(2), 195–207. Seckler et al. (2015) Trust and distrust on the web: User experiences and website characteristics. Computers in Human Behavior, 45, 39–50. A note on how this was written: This series is AI-assisted. I provide the stories, the methodology, the case studies, and the editorial direction. AI helps me structure and draft. This is consistent with Perception-First Design’s own transparency principle: if I’m writing about perception, I should be honest about how the writing itself is produced. --- ## Chapter 8: The Trail URL: https://aurochs.agency/writing/make-me-think/the-trail/ Sing down a scale in front of a crowd and stop one note before the end. Do, re, mi, fa, sol, la, ti... Everyone in the room knows what comes next. They can feel it. The resolution is inevitable, and the pause before it creates a physical tension in the body. Your brain has predicted the final note so completely that withholding it generates a kind of pleasurable discomfort. And when it lands, when the last note arrives exactly where you knew it would, the satisfaction isn’t intellectual. It’s visceral. Pattern completed. Tension resolved. A great call to action feels exactly like that. Not a button that screams BUY NOW, not a popup that guilt-trips you into subscribing, not a countdown timer manufactured to simulate urgency. A resolution. The only possible next step after everything that came before it. The note your brain was already singing. I don’t build funnels. I conduct. The word matters. A funnel is a mechanical thing. You pour people in the top, gravity does the work, and whatever doesn’t leak out the sides arrives at the bottom. The metaphor tells you everything about how most marketers think about their users: as liquid to be channeled. Passive. Directionless without the container. Conducting is something else entirely. Think about a musician who’s reached unconscious mastery, the point where they stop thinking about technique and focus entirely on how the music makes you feel. They’re not forcing notes. They’re shaping an experience in real time, reading the room, adjusting tempo, building toward resolution. The audience doesn’t feel controlled. They feel moved. I’m after the opposite of a funnel. Not a container that forces a path, but an experience that makes the path feel inevitable. The bouncer story from Chapter 1 is the conductor model in action, and I didn’t realize it until years later. At 1:10, twenty minutes before close, I’d walk table by table. “Hey, last call’s coming up. Get any drinks you want now. Thanks for coming out tonight.” I primed each table with a personal visit. Set the tempo by starting early. Gave them a clear next action: get your drinks. Resolved the experience with genuine thanks. By the time the lights came up at 1:30am, half the venue was already heading for the door. Nobody was forced. Nobody was funneled. They were conducted. Decision architecture is the formal name for what I was doing at that nightclub, and it sits at the top of the Perception-First Design stack as Layer 4. Richard Thaler and Cass Sunstein laid the groundwork in Nudge (2008), establishing that there is no neutral way to present choices. Every design is a choice architecture. Defaults, framing, order, visibility, the number of options, the way options are labeled. All of it shapes behavior whether you intend it to or not. The honest caveat: a Bayesian re-analysis by Maier and colleagues (2022) suggests that nudge effects may be near-zero after correcting for publication bias. The sexy headline results, the ones where a small tweak in framing doubles compliance, don’t hold up as robustly as the original literature promised. Worth naming. But the structural principles survive the correction. Defaults matter. Framing matters. How many options you present at a decision point matters. The effect may be smaller than Thaler and Sunstein suggested, but the mechanism is real. And in design, where you’re making hundreds of micro-decisions about how to present information, small effects compound. Similarly, the choice overload effect turns out to be more complicated than the famous jam study implied. Scheibehenne, Greifeneder, and Todd ran a meta-analysis in 2010 and found the effect was near-zero across many conditions. Sometimes more options help. Sometimes they paralyze. It depends on context: the person’s expertise, the stakes, the complexity of the choice, whether the options are meaningfully different. I want to sit with that for a moment, because the nuance matters for practice. “Fewer choices is always better” is bad advice. What’s true: at high-stakes single decisions, at unfamiliar purchase moments, at points where a visitor is deciding whether to commit, reducing options reduces paralysis. What’s not true: that you should strip options from expert tool palettes, browse interfaces, or workspace contexts where option density is a feature. A surgeon doesn’t want a simplified toolbar. A developer doesn’t want three menu items (I’m building a product right now where the whole point is information density). The principle is about decision points, not interfaces in general. This is the final layer for a reason. It requires everything below it to be functional. A perfectly designed trail doesn’t help if the visitor is cognitively overloaded (L0). Without trust on first impression (Layer 1), they’ll never follow the trail. Inconsistent or hard-to-process presentation (Layer 2) turns the trail into friction instead of flow. And if the messaging doesn’t connect emotionally (Layer 3), the trail leads to a destination nobody wants to reach. Decision architecture is the capstone. It assumes you’ve already cleared the bandwidth, earned the first impression, established fluency, and connected through perception. Now you conduct. Now you build the trail. They’re hunters looking for prey, and it’s my job to make a trail. That line has become my shorthand for the whole layer. Your visitors aren’t wandering. They arrived with intent. They’re looking for something specific, even if they can’t articulate exactly what. My job isn’t to decide what they want. It’s to make the path between their intent and their goal feel so natural that following it feels like their idea. I think about this through what I call the Mental Waterfall, the unconscious sequence of questions every visitor processes when they land on a site: The hero section answers questions one through five. If it does that in the first second or two, the visitor is primed. Then the rest of the page conducts them through six through nine as they scroll. Each section resolves one question and creates the momentum for the next. Like notes in a scale. Each one implies what follows. This is the same prediction mechanism from the last chapter—the brain generating expectations, matching each section against them. Match means flow. Violation means friction. The waterfall isn’t a checklist. It’s a description of what’s already happening in your visitor’s head whether you designed for it or not. If your page doesn’t answer these questions in roughly this order, the visitor doesn’t stop and wait for answers. They leave. Autopilot carries them right past you, the same way it carried people past the nightclub entrance in Chapter 2. If the hero fails, it’s the lights slamming on at 1:30am. They’re gone before you’ve started conducting. The practical rules are simpler than the theory. Navigation reflects user goals, not your org chart. This is the most common L4 failure I see, and it’s almost universal. Companies organize their navigation by how the company thinks about itself: “Products,” “Solutions,” “Resources,” “Company.” Users think in problems: “How does this work?” “What does it cost?” “Is it right for me?” When your navigation matches the company’s internal structure instead of the user’s mental model, every click requires translation. The visitor has to guess which department has the answer to their question. That’s a prediction error in the wrong direction: confusion instead of curiosity. I rebuilt an e-commerce site where the old version was a homepage and a product list. That was the entire trail: land, browse, hope you find something. The new site organized products by how customers actually search—by use case, not by SKU. Someone looking for a commercial vacuum sealer and someone looking for a home food saver have completely different intents, different budgets, different urgency. The navigation has to respect that from the first click, or the trail is broken before it starts. One primary CTA per viewport section. Not one CTA per page. One primary action per visible area. Too many buttons creates choice paralysis at the exact moment you need decisiveness. Zero buttons creates a dead end. One button, the right one, at the right moment, creates resolution. CTAs at natural decision points. After value is established, not before. After social proof, not before. After the price, not before. A “Buy Now” button above the fold, before you’ve explained what the product is or why anyone should care, is like asking someone to marry you at the door. The tempo is wrong. Show the premium option first. Anchoring is one of the most robust findings in behavioral economics. When the first number someone sees is high, everything after it feels like a deal. This isn’t manipulation. It’s honest framing. If your best offering genuinely is the best fit for most people, showing it first serves both of you. Two refinements to the practical rules that matter more than they sound. First: the description-experience gap. Ralph Hertwig and Ido Erev published a paper in 2009 that identified a systematic asymmetry in how people weigh risk. Decisions from description underweight rare events. Decisions from experience weight them accurately. Told the pill has a 1% side-effect rate, the user ignores it. After the pill causes nausea once, the user quits. Same probability. Different decision. The implication for L4: for anything whose value lives in its edge cases (risk, pricing, uncertainty, future commitment), demonstrate via experience rather than describing the feature. Interactive previews. Free trials. Samples. Calculators. Anything that moves the decision from description to experience. Description-only copy loses the trust argument for any product whose value lives in what might go wrong or what might go right. Second: construal-level theory. Yaacov Trope and Nira Liberman synthesized a decade of research in 2010 around a single finding. The brain processes near and far commitments differently. Near-term, concrete actions (“buy now,” “sign up,” “start the free trial”) trigger concrete-construal processing: immediate, specific, action-oriented. Abstract, future commitments (“strategic partnership,” “long-term growth,” “your company’s future”) trigger high-construal processing: categorical, broad, value-oriented. L4 copy that mismatches the commitment distance loses conversion. A partnership pitch with “sign up” language feels transactional. A free-trial CTA with “build your company’s future” language feels bloated. Match the linguistic register to the psychological distance of the ask. Near and concrete. Far and abstract. Get the pairing wrong and the reader feels it even when they can’t name it. I know this layer is failing when users reach for the search bar instead of navigation. Clearest signal there is. The search bar is a bail-out. It means the trail broke and the visitor has resorted to hunting on their own. If your search bar gets heavy use, your information architecture isn’t doing its job. Other signals: the main action is below the fold or lost in a forest of competing buttons. There’s no clear default option. Users say “I didn’t know what to click.” High conversion but low retention, which smells like dark patterns rather than genuine trail design. And the inverse mistake: experts frustrated by oversimplified interfaces. Misapplied decision architecture to a context where option density is the feature. A Photoshop user doesn’t want three tools. A stock trader doesn’t want one chart. Know your audience’s expertise level and design the trail density to match. Think about how this plays out on an actual page. A visitor lands on your homepage. The hero answers what this is, who it’s for, and why it matters. One CTA: “See how it works” or “Start free” or whatever your primary action is. They click, or they scroll. Either way, the next section deepens the value proposition, maybe with a feature breakdown or a visual explanation. One CTA: same action, different framing, now with more context. Then social proof: testimonials, logos, numbers. One CTA: same action again, now with credibility behind it. Then pricing or a comparison. One CTA: the commitment point. Each section is a note in the scale. Each one resolves the question the visitor was carrying into it and creates momentum toward the next. The CTA doesn’t change. The context around it does. By the time they reach the final section, clicking that button isn’t a leap of faith. It’s the resolution they were already moving toward. This is what separates trail design from the common mistake of scattering CTAs everywhere and hoping one sticks. When you put different actions in every section (“Watch the video!” “Download the whitepaper!” “Schedule a demo!” “Follow us on LinkedIn!”), you’re not giving people options. You’re giving them noise. Each competing action forces a conscious decision: “Is this the right thing to click?” That’s a prediction error at a decision point. Exactly the wrong place for friction. One trail. One destination. Multiple on-ramps at natural decision points. The visitor chooses when to commit, but the commitment is always the same action. I want to be careful about something here, because this layer is where the ethical questions get sharpest. The line between conducting and manipulating is real, and it’s thinner than most designers acknowledge. Dark patterns are decision architecture too. The pre-checked box that subscribes you to a newsletter. The “Are you sure? You’ll miss out on savings!” modal when you try to close a popup. The pricing page where the “free” option is deliberately crippled to push you toward the paid tier. All of these are trails. They’re just trails designed to serve the business at the user’s expense. Thaler and Sunstein actually addressed this distinction. They called their approach libertarian paternalism: structuring choices to guide people toward outcomes that benefit them, while preserving their freedom to choose otherwise. You can argue about whether that framing is honest (and plenty of people have), but the core principle is sound. The question isn’t whether you’re influencing behavior. You are. The question is whether the outcome serves the person being influenced. My test is simple. If the customer understood exactly what I was doing and why, would they still feel good about the experience? If the answer is yes, I’m conducting. If the answer is no, I’m manipulating. Your customers don’t know you restructured the navigation around their mental model instead of your org chart. But if you told them, they’d say “good, that’s how it should work.” The trail serves them. When the trail only serves the business, you’re not conducting. You’re funneling. And we’re back to the mechanical metaphor, pouring people through a container and hoping they don’t leak out. I’ll go deeper on ethics in Chapter 11. But the seed belongs here, because decision architecture is where the temptation is strongest. Every other layer in the stack is about perception: how things look, how they feel, what they signal. This layer is about action. It’s the point where design becomes persuasion. And persuasion without ethics is just manipulation with better typography. There’s a concept from improv that captures the entire philosophy of this layer. One of my improv teachers gave me a formula: (Listen) x (Act + React). The formula for being in the moment. If Listen equals zero, if the user isn’t attending, everything else multiplies by zero. The first impression IS the moment that activates listening. Everything depends on it landing. But once they’re listening, once the hero has done its job and the visitor is engaged, the rest of the experience is Act plus React. You present information (act). The visitor processes it and responds (react). You present the next piece (act). They respond again (react). Each exchange builds on the one before it, and if you’ve been listening, if you’ve actually understood what they came here for, the tempo feels natural. This is why I said at the beginning of this chapter that a great CTA feels like the resolution of a scale. It’s the final Act in a sequence of exchanges where each step was calibrated to what the visitor needed at that moment. Not what you wanted to sell them. What they were ready to hear. The bouncer story is the same pattern. I listened (understood that people wanted to enjoy their last twenty minutes). I acted (walked table by table with useful information). They reacted (ordered drinks, started wrapping up). I acted again (genuine thanks). The resolution (leaving at close) felt natural because every step respected their autonomy and served their interests alongside mine. The trail is not the funnel. A funnel assumes passivity. It assumes users will flow where you channel them, and your job is to minimize leakage. That metaphor produces designs that feel coercive, because they are. Every “nudge” is a wall. Every “optimization” is a way to prevent the user from exercising judgment. A trail assumes agency. It assumes users are hunters with intent, and your job is to make the path between their intent and their goal so clear that following it feels like their own decision. That metaphor produces designs that feel natural, because they are. Every section resolves one question and creates momentum for the next. Every CTA arrives at the moment the visitor is ready for it. Every navigation label speaks the user’s language, not yours. The difference isn’t in the outcome. Both approaches can achieve high conversion. The difference is in what happens after. Funnel-optimized experiences generate buyer’s remorse, support tickets, refunds, and one-time customers. Trail-designed experiences generate repeat purchases, word of mouth, and decade-long partnerships. This is the final layer because it’s the one that turns everything else into results. L0 clears the bandwidth. First impressions earn the right to be heard. Fluency makes the message feel true. Perception bias connects the message to what people actually care about. And decision architecture builds the trail from caring to acting. Do, re, mi, fa, sol, la, ti, do. Resolution. Next: Feel, Unpack, Diagnose, Prescribe, on the four-step process I actually use when I sit down with a client’s site for the first time. Key Terms Decision architecture (Layer 4) Thaler & Sunstein (2008). There is no neutral way to present choices. Every design is a choice architecture. The capstone layer that turns everything below it into results. Unconscious mastery The point where technique becomes invisible and the practitioner focuses entirely on how the experience lands. The goal for any PFD practitioner. Mental Waterfall The unconscious sequence of 9 questions every visitor processes when they land: from “What am I trying to accomplish?” to “Does this give me what I want?” Trail design Building a path between user intent and action that feels so natural, following it feels like the user’s own decision. Contrasted with funnel optimization. Choice overload Scheibehenne et al. (2010). The effect is context-dependent: at high-stakes single decisions, reducing options reduces paralysis. Expert tool contexts benefit from option density. Description-experience gap Hertwig & Erev (2009). Decisions from description underweight rare events; decisions from experience weight them accurately. For products whose value lives in edge cases (risk, pricing, uncertainty), demonstrate via interactive preview or trial rather than describing the feature. Construal-level Trope & Liberman (2010). The brain processes near and far commitments differently. Near/concrete actions (“buy now”) need concrete language. Abstract/future commitments (“partnership”) need high-construal language. Mismatches feel transactional or bloated. Libertarian paternalism Thaler & Sunstein (2008). Structuring choices to guide people toward beneficial outcomes while preserving freedom to choose otherwise. References Thaler & Sunstein (2008) Nudge: Improving Decisions About Health, Wealth, and Happiness. Yale University Press. Maier et al. (2022) No evidence for nudging after adjusting for publication bias. PNAS, 119(31). Scheibehenne, Greifeneder & Todd (2010) Can there ever be too many options? A meta-analytic review of choice overload. Journal of Consumer Research, 37(3), 409–425. Hertwig & Erev (2009) The description-experience gap in risky choice. Trends in Cognitive Sciences, 13(12), 517–523. Trope & Liberman (2010) Construal-level theory of psychological distance. Psychological Review, 117(2), 440–463. A note on how this was written: This series is AI-assisted. I provide the stories, the methodology, the case studies, and the editorial direction. AI helps me structure and draft. This is consistent with Perception-First Design’s own transparency principle: if I’m writing about perception, I should be honest about how the writing itself is produced. --- ## Chapter 9: Feel, Unpack, Diagnose, Prescribe URL: https://aurochs.agency/writing/make-me-think/feel-unpack-diagnose-prescribe/ A client sent me a link to their homepage last year with the subject line “something’s off.” No other context. No analytics. No brief. Just a URL and three words. I opened the page. My stomach tightened. My eyes couldn’t land anywhere. Dense text above the fold, a rotating carousel of stock photos, eleven navigation items, and a chatbot popup that appeared before the page finished loading. The overall feeling: I’m being talked at by someone who doesn’t realize they’re shouting. That feeling happened in about two seconds. I hadn’t read a word of copy. I hadn’t checked the analytics. I hadn’t looked at the mobile experience. But I already knew the shape of the problem. Here’s the thing: that two-second gut response wasn’t a hunch. It was a diagnosis. I just didn’t have the vocabulary for it yet. The Doctor’s Glance In medical education, there’s a concept called the doorway diagnosis. An experienced physician walks into an exam room, and before they ask a single question, before they touch a stethoscope, they’ve already formed a working hypothesis. Skin color, posture, breathing pattern, facial expression. The patient hasn’t spoken yet, but the doctor has already started narrowing the possibilities. This isn’t magic. It’s pattern recognition running on a database built from thousands of prior patients. Ericsson and colleagues spent decades studying expert performance and found that what separates experts from novices isn’t speed of thought. It’s the size and organization of their pattern library (Ericsson, Krampe & Tesch-Römer, 1993). A cardiologist who has listened to ten thousand heart sounds doesn’t consciously compare each murmur to a textbook diagram. The recognition fires automatically. The pattern matches before the conscious mind catches up. Gary Klein studied this across firefighters, military commanders, and intensive care nurses. His Recognition-Primed Decision model (Klein, 1998) showed that experts in high-stakes environments almost never compare options the way decision theory says they should. They don’t line up alternatives and weigh pros and cons. Instead, they recognize the situation as a variant of something they’ve seen before, mentally simulate the first plausible response, and act. The recognition happens fast. The deliberation happens after, as a check, not as the primary engine. Design diagnosis works the same way. When I open that client’s homepage and my stomach tightens, that’s my pattern library firing. Fifteen years of looking at websites, thousands of pages across hundreds of clients, compressed into a gut signal that says: this page is fighting itself. The feeling isn’t soft. It’s the fastest diagnostic instrument I have. Why the Sequence Matters Over those fifteen years, I’ve distilled the diagnostic process into four steps. They sound obvious when I list them. The sequence seems almost too simple to call a methodology. But the order is load-bearing. Skip a step, reverse two, or collapse them into one, and the diagnosis breaks. The four steps: Feel. Unpack. Diagnose. Prescribe. Each one operates in a different cognitive mode. Feel is pre-verbal. Unpack is analytical. Diagnose is forensic. Prescribe is architectural. They use different parts of the brain, and they need to happen in sequence because each step generates the input for the next one. This isn’t just my personal preference for how to work. The sequence mirrors how expert reasoning actually operates. Croskerry (2009) mapped dual-process theory onto clinical diagnosis and found that effective diagnosticians oscillate between intuitive recognition (System 1) and analytical verification (System 2), but they start with intuition. The gut response narrows the search space. The analysis confirms or corrects it. Starting with analysis, without an intuitive anchor, produces what Croskerry calls “premature closure,” locking onto the first measurable thing you find instead of the thing that actually matters. In design, premature closure looks like this: someone opens the analytics, sees a high bounce rate, and immediately starts redesigning the hero section. They skipped feeling the page. They skipped unpacking which principle is violated. They jumped from symptom to treatment and missed the disease. Step 1: Feel I arrive at the page and I let the emotional response fire before language kicks in. Before I start naming things, before I look for specific problems, before I open DevTools or check the analytics dashboard. I just feel it. That “something’s off” homepage? My Step 1 took two seconds. Stomach tightened. Eyes couldn’t land anywhere. The compressed signal my System 1 returned: “being shouted at.” I wrote that down. I didn’t name the eleven nav items or the chatbot popup yet. Not yet. The vocabulary of Feel is simple and physical. “This makes me feel rushed.” “This makes me feel like I can’t find the thing I need.” “This feels cheap.” “This feels cold.” “This feels like nobody cared.” Notice the language. “Rough.” “Heavy.” “Sharp.” “Warm.” “Sour.” When we describe first impressions, we instinctively reach for sensory words. This is not a quirk of casual speech. It is grounded in how cognition actually works. Lakoff and Johnson (1980) laid the foundation: abstract concepts are structured through physical metaphors. We understand “heavy” topics, “warm” people, “rough” experiences because cognition is embodied. It runs on a body that has weight, temperature, texture. The metaphors are not decorative. They are the architecture. The experimental evidence is striking. Williams and Bargh (2008) had participants hold either a warm or cold cup of coffee, then rate a stranger’s personality. Warm cup holders rated the stranger as having a “warmer” personality. Physical sensation directly influenced social judgment. Ackerman, Nocera, and Bargh (2010) extended this across multiple senses: the weight of a clipboard made résumés feel more “important,” rough textures made social interactions feel more “difficult,” and hard objects made negotiators more “rigid.” Haptic sensation shaped abstract evaluation, reliably, across studies. This is why the Feel step works. When a designer says “this site feels heavy” or “that layout feels rough,” they are not being vague. They are accessing a perceptual system that maps physical sensation to abstract evaluation. Sensory language IS diagnostic language. It is the fastest path to what the unconscious evaluation actually detected. The feeling itself is the diagnostic data. Not an interpretation. Not a theory about what’s causing it. The raw emotional response. In an industry obsessed with data and measurable outcomes, “I felt a thing” sounds like it belongs in a therapy session, not a design review. But that feeling IS your System 1 processing the entire page simultaneously. It’s the predictive model I described in Chapter 2, running a full-stack pattern match against thousands of stored templates and returning a single compressed signal. “Off.” “Good.” “Wrong.” “Trustworthy.” “Sketchy.” Damasio’s somatic marker hypothesis (1994) explains why this works. Emotional responses aren’t separate from rational decision-making. They’re integral to it. The body tags situations with feelings based on prior experience, and those tags function as rapid-fire heuristics that narrow the decision space before conscious analysis even begins. Patients with damage to the ventromedial prefrontal cortex, the region that generates these somatic markers, can reason perfectly well in abstract terms but make catastrophic real-world decisions. They lost the feeling, and without it, they lost the ability to prioritize. The feeling is the fastest diagnostic instrument you have. It’s also the most honest one, because it fires before your conscious brain starts rationalizing, making excuses, or looking for things to praise. Emotional Primacy: Why the Feeling Arrives First This isn’t a metaphor. Feelings are the fastest response system the central nervous system produces. They arrive before language can catch up, before your brain can put words to what it detected. That’s not a design choice I made for this methodology. It’s how human neurology actually works. LeDoux (1996) mapped the pathway. When sensory input hits the thalamus, it splits. One route goes up through the cortex for detailed, conscious analysis. The other route, what LeDoux called the “low road,” shoots straight to the amygdala, bypassing the cortex entirely. That low road delivers an emotional response in as little as 12 milliseconds. The cortical route, the one that gives you conscious awareness and the words to describe what you’re seeing, takes 200 to 300 milliseconds. By the time your conscious mind knows what it’s looking at, your emotional system has already responded. Zajonc (1980) argued this from the behavioral side before the neuroscience confirmed it. His paper “Feeling and thinking: Preferences need no inferences” presented evidence that affective reactions can occur independently of, and prior to, cognitive processing. Emotions are not post-cognitive. They are pre-cognitive. You don’t evaluate something and then feel about it. You feel about it and then evaluate it. The preference forms before the reasoning can justify or explain it. Now compare this to language. Wernicke’s area, the region of the brain responsible for language comprehension, operates on a completely different timescale. Kutas and Hillyard (1980) discovered the N400 component, an electrical signature the brain produces when it encounters a word and tries to make meaning of it. That component peaks around 400 milliseconds after stimulus onset, with activity spanning roughly 200 to 500 milliseconds. That is the speed of language comprehension: a quarter-second to half-second for your brain to process a single word in context. The gap between feeling and naming is neurologically real and measurable. Your amygdala responds at 12 milliseconds. Your language system responds at 200 to 500 milliseconds. That is a 10x to 40x speed difference. This is why you laugh before you can explain why something is funny. Why you recoil before you can name what’s wrong. Why you feel a certain way about a webpage, a person, a room, before you can articulate it. The 50-millisecond verdict I keep returning to in this book isn’t just fast. It’s pre-linguistic. The feeling arrives first. Language follows. Damasio’s somatic marker work (1994), already referenced above, converges on the same point. The body tags situations with feelings before the conscious mind reasons about them. Those somatic markers are the CNS’s shortcut, and they operate on a timescale that language cannot match. This is what I mean by “50 milliseconds.” The Feel step is not about being imprecise or touchy-feely. It is about accessing the fastest, most honest evaluation system the brain has. By the time you can say what’s wrong, you’ve already felt it. The diagnostic sequence starts with Feel because that is where the data arrives first. My ADHD is actually an advantage here. I feel cognitive load more acutely than most users. Friction that a neurotypical designer might walk past without noticing, a cluttered sidebar, an ambiguous label, registers for me as genuine discomfort. The same sensitivity that makes filling out government forms painful makes me a better diagnostic instrument. My perceptual system is tuned to a lower threshold. That’s a liability in daily life and an asset in design work. The critical discipline in Step 1 is not acting on the feeling. Not yet. The feeling is data. It tells me something is wrong and points in a direction. But if I jump from “this feels cheap” straight to “let’s fix the typography,” I’ve collapsed four steps into one. I’ve skipped the unpacking, the diagnosis, and the requirements gathering. I’ll probably land on a reasonable fix. But I’ll miss the systemic issue underneath it. Sit with the feeling. Name it. Write it down. Then move to Step 2. Step 2: Unpack “Why does this feel X?” Now I’m looking for the broken rule. Not the specific element (though I’ll get there), but the underlying principle the page violates. Something made me feel rushed. Why? Something made me feel like the site was untrustworthy. What pattern is generating that signal? This is the transition from System 1 to System 2. The gut response has narrowed the search space. Now the analytical mind works within that narrowed field instead of trying to evaluate everything at once. Klein (1998) found that experts who skipped this step, who acted purely on recognition without unpacking, made faster decisions but more errors. The experts who paused to mentally simulate their intuitive read before acting caught problems that pure recognition missed. Back to that homepage. Why “being shouted at”? I started cataloging which principles were broken. Don’t overload. Eleven navigation items, three competing calls to action, a rotating carousel, and a chatbot popup. The brain’s working memory holds roughly four chunks at a time (Cowan, 2001). This page was demanding twelve. Every element beyond the limit becomes noise that the brain has to actively filter, spending processing bandwidth on things that don’t carry meaning. Hick’s Law (Hick, 1952) predicted this decades ago: decision time increases logarithmically with the number of choices. More options doesn’t mean more engagement. It means more paralysis. A caveat: less doesn’t always mean more. In professional tooling, code editors, design software, analytics dashboards, information density IS the product. Kalyuga and colleagues (2003) documented the expertise reversal effect: instructional designs that help novices can actively hurt experts. Simplified interfaces strip away the context that experienced users rely on to work efficiently. The overload principle applies to consumer-facing design where the user is deciding whether to engage. It does not apply to tools where the user has already committed and needs context to perform. Know the difference. Be specific. The copy said “innovative solutions” and “best-in-class service” and “driven by passion.” None of it told me what the company actually does or why I should care. Vague language triggers the same suspicion as vague eye contact. If you can’t be specific about what you do, I start wondering why. Pennebaker’s text analysis research (2011) shows that concrete, specific language correlates with perceived honesty, while abstract language correlates with perceived deception. Your visitors’ brains are running this evaluation whether you want them to or not. Get to your point. Three scrolls deep and I still didn’t know what this business sells. The hero section is the table visit at 1:10 in a restaurant. You have seconds, not scrolls. Lindgaard’s research (2006) on rapid aesthetic judgments found that users form reliable impressions of websites within 50 milliseconds. If the first viewport doesn’t communicate what you are and who you’re for, everything below it is a conversation with someone who already left. Signal-to-noise. This is essentially contrast between elements. Half the elements on this page were decorative filler. Stock photography that could be on any site in any industry. Animated flourishes that moved without purpose. When everything is visually loud, nothing stands out. When nothing stands out, the brain has no hierarchy to follow and falls back on scanning, which is slow and effortful. Every element that doesn’t carry meaning is a processing fluency tax on every visitor. Reber, Schwarz, and Winkielman (2004) demonstrated that processing fluency, the ease with which information is processed, directly affects preference judgments. Harder to process means less trustworthy, less likeable, less credible. Visual noise isn’t neutral. It actively degrades the experience. Expectation mismatch. The trickiest violation. Sometimes the design isn’t breaking any obvious rule. The cognitive load is reasonable, the copy is specific enough, the structure is clear. But the feeling doesn’t match the brand’s positioning. A luxury brand that feels generic. A startup that feels corporate. A personal service that feels automated. That gap between what the brand says and what the design signals is a processing fluency violation at the identity level. The verbal system promises one thing and the perceptual system delivers another. The visitor can’t name the conflict, but they feel it. Four of those five violations were present in a single viewport. The feeling was right. It just needed names. Over fifteen years, I’ve noticed the same violations show up again and again. Different industries, different budgets, different teams. The surface symptoms change. The broken rules don’t. That’s what makes unpacking a skill rather than a checklist: the ability to trace a feeling back to the principle it’s violating, even when the surface looks nothing like the last time you saw that principle break. Step 3: Diagnose Unpacking tells me which rules are broken. Diagnosis tells me which layer is failing. This is where data enters the picture. Steps 1 and 2 are perceptual and analytical. Step 3 is forensic. I’m mapping the broken rules I identified in Step 2 to specific layers of the framework and cross-referencing against behavioral evidence. The distinction matters because different layers produce similar symptoms. Low conversion could be a Layer 1 failure (people bounce before they see anything) or a Layer 4 failure (people engage but the path to action is broken). The treatment is completely different. Prescribing without diagnosing the layer is like treating a headache without checking whether it’s dehydration or a brain tumor. The symptom is the same. The cause determines the intervention. I pulled up that client’s analytics. High bounce rate, low time on site. Layer 1: the first impression was doing the damage before anything else got a chance to work. But visitors who made it past the homepage showed decent engagement with low conversion. Layer 4: the trail was broken too. Two layers failing, and the dependency stack told me which one to address first. Over years of doing this, I’ve built a pattern library that maps data signatures to layers. High bounce rate, low time on site: the first impression is failing. Visitors are landing, getting a gut response (Step 1 at user scale), and leaving before they process anything else. That’s Layer 1, the 50-millisecond verdict. The page looks wrong, feels wrong, or fails to communicate what it is before the visitor’s autopilot carries them away. High time on site, low conversion: the visitor is engaged but not acting. They’re reading, scrolling, maybe exploring. But something isn’t compelling them to take the next step. That’s usually Layer 4, decision architecture. The trail is broken. There’s no clear path from interest to action, or the path exists but doesn’t feel natural, so the visitor stalls. High add-to-cart, low checkout completion: might be L0 (cognitive load in the checkout flow itself) or Layer 3 (price perception, where the perceived value doesn’t match the number at the payment screen). Strong desktop, poor mobile: Layer 2, processing fluency. The visual system works at one scale but breaks at another. Spacing that breathes on desktop collapses into a wall on mobile. Typography that anchors hierarchy on a wide screen loses its rhythm on a narrow one. What users say they want doesn’t match what they do. The classic Layer 3 signal, perception bias. Users tell you in surveys they want more features, more information, more options. Their behavior shows them converting on the simplest, most focused version of the page. Nisbett and Wilson (1977) demonstrated this gap decades ago: people regularly misidentify the causes of their own behavior. The gap between stated preference and revealed preference is where perception bias lives, and recognizing it is one of the most important skills a designer can develop. This diagnostic step is where the physician analogy becomes most precise. A good diagnostician doesn’t just recognize symptoms. They map symptoms to systems. They understand which organs are upstream and downstream of each other. They know that treating a downstream symptom while the upstream cause persists is futile. The PFD layer stack functions the same way: each layer depends on the ones below it. Diagnosis tells me where in the stack the failure lives, so the prescription targets the right level. Step 4: Prescribe One rule governs every prescription: fix the lowest failing layer first. This is the dependency stack from the framework, applied as a treatment protocol. The five layers aren’t independent. They build on each other. Layer 1 depends on L0. Layer 2 depends on Layer 1. And so on. If L0 is broken, if the user can’t figure out how to navigate, then fixing Layer 4 (the decision trail) is pointless. The user never gets there. Fixing decision architecture when the first impression is broken is like optimizing a trail that nobody enters. You can build the most elegant conversion path in the world, but if the visitor bounced in 50 milliseconds because the site looked like a scam, the path is invisible. That homepage had Layer 1 and Layer 4 both failing. Layer 1 is lower, so that’s where I started. The stock carousel became a single hero image with one clear sentence: what the company does and who it’s for. Eleven nav items became five. The chatbot popup got killed entirely. Those changes alone moved the bounce rate, because the first impression stopped triggering “being shouted at” and started triggering “I can see what this is.” Then I moved up to Layer 4 and restructured the page so the path from “what is this” to “I want this” felt like walking, not searching. This sounds obvious. It’s also where most design processes go wrong. Teams want to fix the exciting problem: the conversion funnel, the checkout flow, the pricing page. Those are Layer 4 problems, and they’re genuinely important. But if the data shows the real drop-off happening at the homepage, at the hero, at the first impression, then the exciting problem isn’t the actual problem. The actual problem is less glamorous: typography that doesn’t signal professionalism, stock photography that feels generic, a value proposition buried below three screens of scroll. Tversky and Kahneman (1974) identified this as the availability heuristic: people overweight problems that are mentally accessible (the pricing page everyone’s been arguing about) and underweight problems that are harder to articulate (the vague feeling that the homepage “doesn’t feel right”). The dependency stack is the antidote. It replaces subjective priority with structural priority. Which layer is it? Fix the lower one first. Not because it’s more important in some abstract sense, but because downstream layers literally cannot function until the upstream ones work. The Last 10% One pattern I see constantly, across industries and budgets. The difference between a C-minus design and an A-plus design is usually typography, spacing, and styling. Small systematic tweaks that take something from 90% done to fully resolved. Clients often think they need a redesign when they actually need refinement. Those last 5 to 10 percent of changes are disproportionately impactful because processing fluency compounds. One slightly wrong font weight is barely noticeable. Twenty slightly wrong details create a cumulative feeling of “something’s off.” Fix the twenty small things and the whole page snaps into focus. Alter and Oppenheimer (2009) demonstrated this compounding effect experimentally. They found that even minor fluency manipulations, changing a font from slightly hard to read to slightly easy to read, shifted judgments of truth, confidence, and trust. The effect isn’t about any single element. It’s about the accumulated ease or difficulty of processing the whole. Twenty small friction points don’t add up linearly. They multiply, because each one makes the next one feel slightly worse. This is why the diagnostic sequence matters even for “small” fixes. A designer who skips the feeling step and jumps straight to “the font weight is wrong” will fix the font weight. But they’ll miss the nineteen other small things that are compounding with it, because they never felt the cumulative weight. The feeling told me “something’s off.” The unpacking told me it’s twenty small things, not one big thing. The diagnosis told me they’re all Layer 2. The prescription: systematic refinement, not redesign. Where I Learned This I first noticed this sequence in action at the door of a nightclub in Santa Barbara, years before I had a name for it. When someone walked up, I felt something before I could articulate it. Something about their energy, their posture, the way they approached the line. I unpacked why: what specifically was triggering that read? Eyes glassy, speech slurred, posture unstable. I diagnosed: too intoxicated, not “bad person.” Then I prescribed: slow them down with conversation, redirect to water, reassess in ten minutes. Or turn them away cleanly, before the situation escalated. Feel, Unpack, Diagnose, Prescribe. That order. Every time. When I skipped steps, things went wrong. If I jumped from feeling to prescribing (gut says trouble, turn them away), I’d sometimes refuse someone who was just nervous. If I skipped feeling entirely and went straight to diagnosis (checking ID mechanically without reading the person), I’d miss the guy with a valid ID who was clearly about to start a fight. The steps existed because each one caught things the others missed. Bouncing taught me something that took years to fully understand. The best diagnostic decisions aren’t purely intuitive and they aren’t purely analytical. They oscillate between both. Kahneman (2011) described this as the interplay between System 1 and System 2, where intuition generates candidates and analysis evaluates them. Klein and Kahneman eventually co-authored a paper (2009) reconciling their frameworks, agreeing that expert intuition is reliable when two conditions are met: a regular environment with valid cues, and sufficient practice to learn those cues. A nightclub door at 1 AM on a Saturday is a regular environment with valid cues. So is a website. The patterns repeat. The cues are learnable. The intuition, once trained, is fast and accurate. But it needs the analytical check. It needs the unpacking and the diagnosis. Without them, intuition is just guessing with confidence. The Whole Sequence Feel, Unpack, Diagnose, Prescribe. Four steps, always in order. That client’s “something’s off” became a page that converts. Not because I had a magic fix. Because I felt it first, unpacked why, diagnosed which layers were breaking, and prescribed in dependency order. The feeling led to the rules. The rules led to the layers. The layers led to the fix. The alternative to this sequence is every shortcut I’ve watched fail in fifteen years. Jumping to solutions before feeling the problem. Copying competitors instead of unpacking the real violation. Treating the exciting symptom instead of the structural cause. The steps exist because every way of skipping them has a name, and I’ve seen each one cost someone real money. The sequence is the methodology. The layers give you the map. And the discipline of not proposing solutions until you understand the problem is the hardest skill in design, because every instinct tells you to jump to the fix. Fifteen years in, I still have to force myself to slow down. To feel first. To let the feeling teach me something before I decide what it means. Next: Music, Humor, Stories, on the three tools that change how people feel about anything, ranked by power. Key Terms Feel, Unpack, Diagnose, Prescribe The four-step diagnostic sequence. Feel is pre-verbal (System 1). Unpack is analytical (System 2). Diagnose is forensic (mapping to layers). Prescribe is architectural (fixing in dependency order). Sequence is load-bearing. Doorway diagnosis Medical concept: an experienced physician forms a working hypothesis before asking a single question. Pattern recognition running on thousands of prior cases. Design diagnosis works identically. Recognition-Primed Decision Klein (1998). How experts actually decide in high-stakes environments: recognize the situation type, mentally simulate the first plausible response, act. Recognition first, deliberation as a check. Emotional primacy Zajonc (1980), LeDoux (1996). Affective responses arrive before cognitive processing. The amygdala’s low road responds in ~12ms; language comprehension operates at 200–500ms. The feeling is pre-linguistic. It arrives before you can name it. Somatic markers Damasio (1994). The body tags situations with feelings based on prior experience. These markers function as rapid heuristics that narrow the decision space before conscious analysis. Feeling is not the opposite of reasoning. It is reasoning’s fast lane. Premature closure Croskerry (2009). Locking onto the first measurable finding instead of the actual root cause. The diagnostic error that occurs when you skip from symptom to treatment without the intermediate steps. Viewport count Layer 0 test: count everything competing for attention in a single viewport. Working memory holds roughly four chunks (Cowan, 2001). If a screen demands twelve, L0 is failing. Dependency stack The treatment protocol: fix the lowest failing layer first. Downstream layers cannot function until upstream ones work. Structural priority replaces subjective priority. References Zajonc (1980) Feeling and thinking: Preferences need no inferences. American Psychologist, 35(2), 151–175. Kutas & Hillyard (1980) Reading senseless sentences: Brain potentials reflect semantic incongruity. Science, 207(4427), 203–205. Lakoff & Johnson (1980) Metaphors We Live By. University of Chicago Press. LeDoux (1996) The Emotional Brain: The Mysterious Underpinnings of Emotional Life. Simon & Schuster. Ericsson, Krampe & Tesch-Römer (1993) The role of deliberate practice in the acquisition of expert performance. Psychological Review, 100(3), 363–406. Klein (1998) Sources of Power: How People Make Decisions. MIT Press. Damasio (1994) Descartes’ Error: Emotion, Reason, and the Human Brain. Putnam. Croskerry (2009) A universal model of diagnostic reasoning. Academic Medicine, 84(8), 1022–1028. Cowan (2001) The magical number 4 in short-term memory. Behavioral and Brain Sciences, 24(1), 87–114. Hick (1952) On the rate of gain of information. Quarterly Journal of Experimental Psychology, 4(1), 11–26. Pennebaker (2011) The Secret Life of Pronouns. Bloomsbury Press. Lindgaard et al. (2006) Attention web designers: you have 50 milliseconds to make a good first impression! Behaviour & Information Technology, 25(2), 115–126. Williams & Bargh (2008) Experiencing physical warmth promotes interpersonal warmth. Science, 322(5901), 606–608. Reber, Schwarz & Winkielman (2004) Processing fluency and aesthetic pleasure. Personality and Social Psychology Review, 8(4), 364–382. Nisbett & Wilson (1977) Telling more than we can know: verbal reports on mental processes. Psychological Review, 84(3), 231–259. Tversky & Kahneman (1974) Judgment under uncertainty: heuristics and biases. Science, 185(4157), 1124–1131. Alter & Oppenheimer (2009) Uniting the tribes of fluency to form a metacognitive nation. Personality and Social Psychology Review, 13(3), 219–235. Ackerman, Nocera & Bargh (2010) Incidental haptic sensations influence social judgments and decisions. Science, 328(5986), 1712–1714. Kahneman (2011) Thinking, Fast and Slow. Farrar, Straus and Giroux. Klein & Kahneman (2009) Conditions for intuitive expertise: a failure to disagree. American Psychologist, 64(6), 515–526. Kalyuga et al. (2003) The expertise reversal effect. Educational Psychologist, 38(1), 23–31. A note on how this was written: This series is AI-assisted. I provide the stories, the methodology, the case studies, and the editorial direction. AI helps me structure and draft. This is consistent with Perception-First Design’s own transparency principle: if I’m writing about perception, I should be honest about how the writing itself is produced. --- ## Chapter 10: Music, Humor, Stories URL: https://aurochs.agency/writing/make-me-think/music-humor-stories/ I took my first improv class in Chicago because a guy I met at a meetup about emotions in games became a friend and told me to go see a show at iO Theater. I was captivated by what I saw, and how improvisation led to so much laughter and people paying close attention to whatever that bouncing ball of attention the actors held compelled them to. I’d gotten my autism diagnosis a few months earlier. I was just starting to realize that my systems for socializing were miscalibrated. I wasn’t picking up on when people were disinterested, or reading other cues. Tone deaf in the social sense. When I saw improv, I saw all those social instincts being called on in patterns that other improvisers picked up on, took in at full blast, and then acted and reacted on. To the audience it looked like magic, people making things up out of nothing. But improv is less about making things up and more about using your experience to act in the moment. And these performers were using it to tell stories and jokes and evoke the full spectrum of emotions. I was captivated, and I needed to know what was actually happening. Taking classes literally spelled it out. The rules, the patterns, the dynamics, and the kind of group mind that teams develop after working with each other long enough. Everything the audience experienced as spontaneous magic had structure underneath it. Not metaphors. Tools. There are three ways to change how someone perceives something. I rank them by power: music, humor, and stories. This ranking comes from 15 years of applied work and it’s backed by converging research across music cognition, humor psychology, and narrative persuasion. But I want to be clear: the unified hierarchy is my practitioner synthesis. Each level is grounded independently in the literature. The combined ranking is how I use them, not a published model. Music first. Music: Bypassing Rational Processing When I say “music” I don’t mean put a soundtrack on your landing page. I mean the thing that music does to the brain, and how that same mechanism operates in visual design. Music communicates emotion, momentum, and how to think and feel. And it does all of this by bypassing rational processing entirely. Koelsch showed in 2014 that music modulates subcortical emotion structures (amygdala, nucleus accumbens, hippocampus) rapidly and potentially before conscious evaluation even begins. Your brain responds to musical structure before you’ve decided to respond. The emotional hit lands before the analysis has time to form. Juslin and Västfjäll (2008) identified six mechanisms for music-evoked emotion, and the majority of them, brainstem reflexes, evaluative conditioning, emotional contagion, bypass conscious processing entirely. Huron mapped this in 2006 as the ITPRA theory: Imagination, Tension, Prediction, Reaction, Appraisal. Of those five stages, only appraisal is fully conscious. Four-fifths of the response happens before you know you’re responding. It’s not a metaphor for design, it IS design. There’s a video I send people when I’m trying to explain what I mean by the gap between technique and feeling. It’s Benjamin Zander, conductor of the Boston Philharmonic, giving a talk about classical music and why most people think they don’t like it. He starts by mocking a kid stuck in piano lessons, grinding through a Mozart sonata with no progress because the lessons are boring. Then he grabs someone from the audience, pretends they’re that kid, and tells them: “You’re a two-buttock player. You need to be a one-buttock player.” He sits down at the piano and shows the difference. A two-buttock player is planted, rigid, technically correct and emotionally dead. A one-buttock player leans into the music. The technique is invisible because it’s been internalized so deeply that conscious attention is free to focus entirely on how the music makes you feel. He plays Chopin’s Prelude in E minor, Op. 28, No. 4, the one sometimes called the “Suffocation” prelude. He walks through what Chopin is doing: delaying resolution, using deceptive cadences, sustaining tension longer than your ear expects. At one point a man in the audience sighs audibly when the resolution finally lands, and Zander picks him out. That sigh is the point. The body responded before the mind caught up. He does this thing where he sings down a scale and stops before the last note. The whole audience fills in the tonic without being asked. Nobody told them to. The pattern created an expectation so strong that completing it felt involuntary. That’s not education. That’s the nervous system responding to musical structure the same way it responds to gravity. Then he plays the whole prelude and asks the audience to think of someone they’ve lost. Follow the delayed resolution, he says. Feel where Chopin holds the tension and where he finally lets it go. The room goes still. “Classical music is for everyone,” he says. “They say 3% of the population likes classical music. I say everyone loves classical music. They just haven’t found out about it yet.” And then the line that stuck with me: “As a performer, ask yourself: who am I being, that my listeners’ eyes are not shining? It’s not about wealth and power. It’s about how many shining eyes are around me.” Unconscious mastery and flow state. That’s what I’m watching in that video. Someone who has internalized decades of technique so completely that none of it is visible in the performance. All you see is the effect. When I sit down and focus these days, that’s what happens with this methodology. The layers, the diagnostics, the pattern library, the perception analysis. It runs automatically. I don’t think about how things are done. I think about how they land. I’m not bragging. The point is that the goal for any practitioner learning this framework is to internalize it deeply enough that it becomes unconscious. The conscious version is slow and effortful. The unconscious version is responsive, adaptive, and present. When the technique becomes invisible, the outcome becomes everything. Think about what typography rhythm does to a page. The spacing between headlines and body copy. The cadence of short paragraphs followed by a longer one. The visual flow that pulls your eye from hero to proof to CTA without you consciously tracking the movement. The timing of a scroll-triggered animation, whether it feels abrupt or inevitable. The music of a page. It controls pacing and emotional state the same way a tempo change controls the energy in a room. Fast sections create urgency. Slow sections create gravity. Pauses create anticipation. And the transition between them, the moment where a dense technical section gives way to a single short sentence standing alone, that’s a key change. The reader feels it in their body before they process it consciously. That scale Zander sang, stopping before the last note. The pattern demands completion. You don’t have to explain it. You don’t have to label it. The brain fills in the resolution because the expectation is so strong that completing it feels inevitable. A well-designed page works the same way. Each section creates the expectation that naturally resolves in the next. The hero sets up the problem. The proof section confirms the authority. The testimonials provide social validation. And the CTA is the last note of the scale. By the time someone reaches it, clicking feels like completing a thought they were already having. You don’t convince anyone to click a CTA. You design a sequence where clicking is the resolution of accumulated momentum. The button isn’t persuasion. It’s the final note that was always going to come. Humor: Controlling What People Notice Music controls how people feel. Humor controls what they pay attention to. Jokes change how people feel about something by pointing out absurdity and giving permission to think without rigidity. The structure of a joke is deceptively simple. The setup creates an expectation. The punch reveals a pattern. And what happens between setup and punch, the tiny gap where the brain is anticipating the resolution, that’s anxiety. Laughter is the anxiety being released. It works the same way a prediction error works in the broader framework. The brain predicts what’s coming. The punchline either confirms the prediction in a surprising way (“I’d do that!”) or violates it in a way that reframes the setup (“I’d never do that!”). Either way, the surprise generates engagement. The brain leans in because something unexpected happened in a safe context. In design, this shows up as pattern interrupts, unexpected delight, personality in copy, and moments where the experience subverts expectation in a satisfying way. Not literal jokes. Don’t put puns on your pricing page. Don’t get cute in your terms of service. Anywhere the user needs clarity, give them clarity. But the mechanism of humor, expectation, subversion, release, that’s a design tool. A loading animation that does something unexpected. A 404 page that makes you smile instead of sigh. A piece of copy that zigs when every competitor zags. These aren’t humor in the traditional sense, but they use the same cognitive machinery. Anticipation, surprise, delight. In improv, there’s something called the rule of eventually. It’s the technique of starting a word that sounds like it’s going one direction and landing somewhere else. “That’s horri-bly... good!” The audience’s brain hears “horri-” and braces for “horrible.” The anxiety of an insult builds. Then “good” lands and the relief creates a burst of energy. The joke isn’t the content. It’s the trajectory. This is a micro-version of what great design does. Create an expectation, sustain it long enough for the brain to commit, then deliver something that resolves the tension in a way that’s better than what was predicted. Not confusing. Not random. Better. Three comedy escalation patterns map directly to UX: Linear repetition. Same intensity, building familiarity. This is consistent brand messaging. Every page, every touchpoint, same tone, same quality, same feel. The repetition doesn’t escalate but it accumulates. Trust builds through predictability. After the fifth time a brand delivers on its implicit promise, the sixth time feels like a relationship. Sequential escalation. Building momentum toward a peak. This is the marketing funnel narrative. Each step in the journey is slightly more compelling than the last. The blog post is interesting. The email series goes deeper. The case study is specific and concrete. The discovery call feels inevitable. Each beat builds on the one before, creating a sense of forward motion. Exponential escalation. Each beat dramatically amplifies the last. This is the hero section into social proof into testimonial into CTA pattern. The hero makes a bold claim. The logos beneath it silently say “these companies trusted that claim.” The testimonial says “and here’s a real person confirming it.” By the time the CTA appears, the accumulated proof has gone from assertion to evidence to human validation, and each layer multiplied the credibility of the one before. Stories: Shaping What People Believe Music shapes how people feel. Humor shapes what they notice. Stories shape what they believe. Stories are how we simulate experiences without having to live them. Green and Brock demonstrated this in 2000 with their research on narrative transportation, the phenomenon where absorption into a story produces belief changes consistent with the narrative. You read about someone’s experience, you get pulled into it, and when you emerge, your beliefs have shifted toward the story’s perspective. Through simulation, not argument. Van Laer and colleagues confirmed this across 132 effect sizes in a 2014 meta-analysis. Braddock and Dillard found the same pattern in narrative persuasion studies. The mechanism holds up: stories change beliefs and attitudes. They do it reliably. They’ve been doing it for as long as humans have been communicating. But most people miss the important part about stories. A story isn’t the events. It’s how someone changes from A to B. Lisa Cron nails this in Wired for Story: the brain is wired to track transformation, not sequence. “I went to the store and bought milk” is a sequence of events. “I was terrified of grocery stores until I realized the produce section was designed to make me feel calm” is a story. The difference is transformation. This applies directly to how a website should work. The user is the protagonist. They arrive with a problem (point A) and the experience should transform their understanding (point B). The hero section is the inciting incident, the moment something disrupts their autopilot and activates engagement. The content is the rising action, building understanding and trust. The CTA is the climax, the moment of decision. Testimonials are supporting characters who’ve already completed the journey and came back to tell you it was worth it. That makes the page a transformation arc. The user is the lead. The brand is supporting cast. Most landing pages get this backwards. They tell the company’s story. Founded in 2003. Trusted by millions. Our mission is to blah blah blah. The company is the protagonist and the user is the audience. It takes a bit of perspicacity to flip that. To put yourself in your audience’s shoes and figure out what their actual mental flow is when they land on your page. What their goals are. What emotions they’re carrying. Most people skip this step because it means thinking about someone who isn’t you. Cast the user as the lead. Everything else is supporting cast. And here’s where this gets practical. The typical homepage is a hero section with a value proposition and a call to action. “Buy now.” “Learn more.” “Schedule a demo.” All of those are asks. They take from the visitor. What if you gave something instead? What if the first thing someone encountered wasn’t a demand but an acknowledgment of why they’re here? Same intent, shaped for how it’s perceived. The Improv Rules That Became Design Tools All of this came from improv class. Not all at once. The notebook I kept from 2012 and 2013, a bound hardcover I carried to every class and show, was mostly notes about scene work and character exercises. But the principles I was writing down kept showing up in my design work, and eventually I stopped treating them as separate domains. The improv rules became operational design tools. These are the ones I use constantly. “Yes, and.” The foundational rule of improv, and the foundational rule of this methodology. You agree with what exists and add to it. In improv, this means taking whatever your scene partner gives you and building on it. In design, this means accepting the constraints (the client’s brand, the existing platform, the budget, the timeline) and building within them. Don’t throw everything out when something feels wrong. Brick by brick, not castle at a time. In stakeholder conversations, yes-and is how I tell clients they’re correct while disagreeing with them. I accept the validity of their concern (yes, that IS a reasonable worry), then redirect toward a better solution (and here’s how we address it without compromising the design). Nobody likes being told they’re wrong. Everyone likes being told their concern is valid and here’s what we do about it. The fun thing about this rule, and all these rules, is that knowing them lets you bend them, break them, and get around them. You can agree to disagree. You can yes-and someone right into seeing why their own position doesn’t serve them. The rules aren’t constraints. They’re instruments. Once you know how they work, you can play them however the situation demands. “Chicken or egg.” Playing with what comes before and what comes after in a sequence. Things need to make temporal and pattern sense. If your hero section talks about results but your proof section hasn’t established credibility yet, the sequence is wrong. The claim lands before the evidence, and the brain rejects it. Reorder the same content and suddenly it flows. The information didn’t change. The sequence did. “Hold topics and return.” In improv, you introduce something early in a scene, let it breathe, develop other threads, then call back to it when the moment is right. The callback always gets a reaction because the audience has been holding the thread unconsciously. But sometimes you hear something worth returning to and you need to let it go in the moment. Not because it’s unimportant, but because if you’re focused on what you’re going to say next, you’re not listening. Hold it. Let it breathe. Return when it’s right. If the moment doesn’t come, let it go entirely. In design, this is progressive disclosure. Introduce a concept in the hero. Develop it in the body. Call it back at the CTA. The user’s brain has been holding the thread, and the return creates a sense of completion. “Callbacks.” Some improv forms are built entirely around sourcing from an opening scene. The Harold, which came out of iO Theater, works like this: the ensemble does an opening, then every following scene is inspired by something from that introduction. Three rounds of scenes, each separated by a group interlude, and the whole thing coheres because every beat traces back to the same source material. It sounds formulaic on paper but it feels spontaneous, because the audience recognizes the pattern without being able to name it. This works within a single scene or interaction too. Someone drops something interesting, you let it pass, and when the opportunity arises you bring it back. The chord resolves. It feels satisfying because the pattern was set up and then completed. That’s exactly how a well-designed site should work. The hero introduces the core idea. Every section that follows calls back to it, develops it, reframes it. By the end, the whole experience feels like one coherent thought, not a collection of separate sections. “(Listen) x (Act + React).” One of my improv teachers gave me this formula for being in the moment. Listen is the prerequisite. Act is what you do. React is the emotional response that drives the action. It makes total sense to me with most people being on autopilot and not being able to lead or control conversations, let alone their thoughts. In design terms: Listen is what the user perceives in the first moment. Act is what they do in response (click, scroll, leave). React is the unconscious emotional response that determines the action. The formula is a multiplication, not an addition. And that matters enormously, because of what it means when one of the terms is zero. “If this is true, what else is true?” One insight cascades. If this demographic responds to warmth and faces, what else follows? They probably respond to testimonials with photos. They probably trust video over text. The brand voice should probably be conversational, not clinical. In improv, this is how you build a world from a single detail. In design, it’s how a single user research finding generates an entire design direction. “It’s not for you, it’s for them.” This is the one I repeat most often. To myself. To clients. To anyone who says “I don’t like that blue.” The design isn’t for you. It’s for the person you’re trying to reach. Yes-and them into it: if we go with the color you prefer, here’s what happens with your target audience. If we go with what the research says, here’s the likely result. Which outcome do you want to choose? Design for the target’s autopilot, not the stakeholder’s comfort. “What’s familiar to you is novel to others.” This cuts both ways. You’ve looked at your product page 400 times. Your users haven’t. You know every feature by heart. They’re seeing it for the first time. But you can also be nose-deep in something and miss what’s obvious to a fresh pair of eyes. And you can explain things to someone that go completely over their head because you’ve forgotten what it’s like to not have your context. Design for their first encounter, not your hundredth. The curse of knowledge is real and persistent. I’ve watched teams spend weeks debating copy that a user will glance at for half a second. Designing for your own context is the most natural mistake in the world, and it takes active discipline to resist it. All of those improv rules, the one-buttock playing, and every theme running through this book operate under the same mechanism: how something is perceived matters more than what it contains. And that perception starts before anyone consciously evaluates anything. Walk into a restaurant. Before you read the menu, before you talk to anyone, you’ve already decided something. How maintained and clean does it feel? How do the menus and the tables feel when you touch them, sticky, wobbly, or unnoticeable because everything’s fine? The lighting temperature. Whether there are plants. What the soundscape is doing. All of it feeds a single unconscious question: is this experience going to be worth my time and money? Prove me wrong or prove me right. That evaluation happens within a second of arriving and sitting down. Websites work the same way. So do product packages, movies, books, art. Everything you encounter uses this first impression mechanism, and you feel something because of it before you’ve consciously decided to. Gray, Gray, and Wegner’s research (2007) identified two dimensions of mind perception: Agency (competence, the capacity to act) and Experience (empathy, the capacity to care). We judge entities, whether human or digital, on whether they seem capable and whether they seem like they care. Think about the last time you met someone new and felt immediate interest. Why them? What was different? Now think about everyone else in that room you didn’t notice at all. Same space, same lighting, same context. Your brain was running the same evaluation on everyone and only flagging the ones that passed. The red flags work the same way. A person who doesn’t take their appearance seriously signals low agency. A website with broken layouts and pixelated images signals the same thing. Someone who hedges every statement reads the same way a website with vague value propositions reads. A person who buries the point under walls of small talk triggers the same impatience as dense paragraphs and buried CTAs. A person who centers themselves with no empathy for what you need triggers the same rejection as a website that talks endlessly about its own history without addressing the user’s problem. But perception goes both ways. Simons and Chabris demonstrated this in 1999: participants watching a video of basketball passes completely missed a person in a gorilla suit walking through the middle of the frame. Their attention was so directed at counting passes that the most obvious thing in the room was invisible to them. Inattentional blindness. You can design for attention, but attention directed at one thing means blindness to another. That’s another example of knowing the rules well enough to bend them. Once you understand how perception works, you can use it to surface what matters, and you can use it to keep what doesn’t serve the experience out of the way. All of these evaluations operate at the same speed: unconsciously, in the first seconds, before deliberate analysis begins. How seriously a business takes its presentation is a direct indicator of how seriously they’ll take interacting with you. Everything Multiplies by Zero This is where the improv formula becomes the most important insight in the entire methodology. (Listen) x (Act + React). It’s a multiplication. If any term is zero, the product is zero. If Listen equals zero, if the user isn’t attending, if the first impression failed, if the activation point didn’t fire, then it doesn’t matter how good your content is, how compelling your offer is, or how beautiful your design is. Zero times anything is zero. The user is on autopilot. They’re gliding past you. Nothing lands. This is the listen multiplier. It’s why first impressions aren’t just important. They’re multiplicative. Everything downstream, every piece of copy, every testimonial, every carefully designed CTA, all of it depends on the user being in a state of active attention. And active attention depends on the first impression successfully breaking autopilot. When I was at the nightclub, I figured this out physically. If my greeting didn’t land, the rest of the interaction was dead. I could say the same words, give the same instructions, apply the same enforcement, but if the person wasn’t listening, none of it registered. They’d argue. They’d push back. They’d make the interaction ten times harder than it needed to be. Not because they were difficult people, but because I’d never activated their attention in the first place. And what seemed to work at the time was saying “I hope you’re having a good night” instead of asking “how are you doing tonight?” Same thing communicated. One gives something. The other asks for something. The difference in how people responded was immediate and consistent. The greeting that gave something activated listening. The greeting that asked for something activated defense. Try it. Next conversation, next interaction. Give something in your opening instead of asking for something. Watch how the response changes. In design, the hero section is the greeting. The rest of the page is the conversation that follows. If the greeting fails, the conversation never starts. And if the greeting asks (“Buy now,” “Sign up,” “Learn more”) instead of gives (“Here’s what I think is going on with your problem”), the visitor’s defense activates before attention does. The greeting activated everything. Without it, everything multiplies by zero. I learned all of this by accident. A kid with autism who couldn’t make eye contact, taking improv classes in Chicago because a friend from a meetup about emotions in games told me to go see a show at iO. I was captivated. The improv didn’t change my analytical nature. It gave the analysis something to operate on. Writing notes in a bound notebook about scene work principles that turned out to be design principles. Breaking down comedy sets with friends at Second City on Saturday nights, trying to figure out why some things land and others don’t. Dissecting the frog, my best friends called it. The retro after every class, every show, every performance. What worked. Why it worked. Who it worked on. What the audience was actually responding to versus what the performer thought they were doing. That gap, between intent and reception, between what the performer was trying to do and what the audience actually experienced, that’s the same gap I’d been working with since the nightclub. The perception gap. Turns out that’s how pages work. How brands work. How products, software, anything interactive or spectation-based works. The audience is always right. Not about what they say they want, but about what they actually respond to. The gap between those two things is where all of this lives. So figure out when you have that bouncing ball of attention in your hands while you’re on stage, or while a viewer is on your site, or while someone is using your product. That’s when you’ve got eyes glued to you. Let’s actually design for that. Music sets the emotional state. Humor creates the engagement. Stories provide the transformation. And the whole thing depends on listening, on the first moment of genuine attention, because without it, everything multiplies by zero. It’s simpler than it sounds. Before you worry about what your page says, figure out what it feels like. Before you optimize the copy, check the rhythm. Before you A/B test the CTA, make sure the user is actually listening when they get there. And before any of that, ask the question that makes everything else possible: for whom? The improv rule underneath all the other improv rules. Every principle in this chapter, the music, the humor, the stories, the relationship model, the listen multiplier, all of it comes back to the same question. Who is this for, and what are they actually experiencing? Not what you think they should experience. Not what the stakeholder wants them to experience. What they’re actually, physically, unconsciously experiencing in the first second they encounter your design. Design the music first. Then write the lyrics. Next: The Oath, on why engineers of thoughts and emotions need a Hippocratic oath, and the time I turned down Fortune 500 money. Key Terms Communication hierarchy Music, humor, stories. Three tools ranked by power for changing how people perceive something. Music (rhythm/pacing) is most powerful, humor second, stories third. ITPRA theory Huron (2006). Five stages of musical response: Imagination, Tension, Prediction, Reaction, Appraisal. Only appraisal is fully conscious. Four-fifths of the response happens before you know you’re responding. Rule of eventually Improv technique of starting a word in one direction and landing somewhere else. Creates anticipation, surprise, delight. Maps to design pattern interrupts. The Harold Long-form improv structure from iO Theater. Three rounds of scenes sourced from a single opening, creating cohesion through callbacks. Maps to site architecture where every section traces back to the hero’s core idea. Narrative transportation Green & Brock (2000). Absorption into a story produces belief changes consistent with the narrative. Stories change beliefs through simulation, not argument. (Listen) x (Act + React) The improv formula for being in the moment. If Listen equals zero, everything multiplies by zero. The listen multiplier is why first impressions are multiplicative, not additive. Mind perception Gray et al. (2007). Two dimensions: Agency (competence) and Experience (empathy). Users evaluate websites the same way they evaluate people. Inattentional blindness Simons & Chabris (1999). Attention directed at one thing creates blindness to another. Perception goes both ways: you can design for what people see and what they miss. References Koelsch (2014) Brain correlates of music-evoked emotions. Nature Reviews Neuroscience, 15(3), 170–180. Juslin & Västfjäll (2008) Emotional responses to music. Behavioral and Brain Sciences, 31(5), 559–575. Huron (2006) Sweet Anticipation: Music and the Psychology of Expectation. MIT Press. Green & Brock (2000) The role of transportation in the persuasiveness of public narratives. Journal of Personality and Social Psychology, 79(5), 701–721. Van Laer et al. (2014) The extended transportation-imagery model. Journal of Consumer Research, 40(5), 797–817. Gray, Gray & Wegner (2007) Dimensions of mind perception. Science, 315(5812), 619. Cron (2012) Wired for Story. Ten Speed Press. Zander (2008) The transformative power of classical music. TED Talk. Braddock & Dillard (2016) Meta-analytic evidence for the persuasive effect of narratives on beliefs, attitudes, intentions, and behaviors. Communication Monographs, 83(4), 446–467. Simons & Chabris (1999) Gorillas in our midst: sustained inattentional blindness for dynamic events. Perception, 28(9), 1059–1074. A note on how this was written: This series is AI-assisted. I provide the stories, the methodology, the case studies, and the editorial direction. AI helps me structure and draft. This is consistent with Perception-First Design’s own transparency principle: if I’m writing about perception, I should be honest about how the writing itself is produced. --- ## Chapter 11: The Oath URL: https://aurochs.agency/writing/make-me-think/the-oath/ A comedian I used to watch in Chicago had a bit about confidence men. The original con, the one the term comes from, was a guy in 1840s New York who’d walk up to strangers, strike up a conversation, and then ask: “Would you have the confidence in me to trust me with your watch until tomorrow?” And people would hand it over. No trick. No sleight of hand. Just a clean, warm, fluent social interaction that felt so natural, so trustworthy, that the mark’s critical evaluation never fired. That story has bothered me for years. Because the mechanism the confidence man used is the same mechanism I design with. Processing fluency. Social warmth cues. Pattern-matching that says “this feels right.” The brain’s prediction engine processing smoothly, no errors, no red flags, System 1 gliding right through. Everything I’ve spent ten chapters teaching you to deploy, the confidence man deployed first. He just pointed it at the wrong target. Here’s the thing about designing for perception: it works. The 50-millisecond verdict, the prediction engine, the fluency effect, the activation points. These mechanisms are real, they’re powerful, and they’re value-neutral. A fluent lie is more persuasive than a disfluent truth. A warm, trustworthy-looking interface can sell a product that doesn’t deserve trust. An activation point can be calibrated to trigger impulse purchases that users regret the next morning. If you’re a designer working in the perception layer, you are, whether you like the framing or not, an engineer of thoughts and emotions. You decide where attention goes. You decide what feels trustworthy. You decide which predictions to match and which to break. You decide what the user thinks about, and when, and how it makes them feel. That’s power. And the design industry has essentially no shared ethical framework for wielding it. The ACM Code of Ethics exists. It’s thoughtful, comprehensive, and almost nobody outside of academia has read it. I’ve asked designers at conferences. I’ve asked them in interviews. I’ve asked them in workshops. Blank stares. The closest thing most practicing designers have to an ethical framework is “I wouldn’t do anything I felt gross about,” which is not a framework. It’s a vibes check. Vibes checks fail when the incentives push hard enough. Three Tests in Sixty Seconds So I built one. Not a manifesto. Not a poster for the office wall. Three tests you can run on any design decision in sixty seconds, right there in the review session, before anything ships. The Alignment Test: Does this design bring perception closer to reality, or further from it? Berdichevsky and Neuenschwander proposed this in 1999, in one of the earliest papers on the ethics of persuasive technology. They argued that any technology designed to change attitudes or behaviors has an obligation to change them in the direction of truth. Not the designer’s truth. Not the client’s truth. Actual reality. When Simply Smart Home’s site looked like a $25 overseas knockoff despite selling a $150 product that actually worked, perception was below reality. Closing that gap upward is correction, not manipulation. The product was good. The perception was bad. I brought them into alignment. The violation goes in the other direction. If the product is mediocre and the site makes it look premium, if the service is slow and the messaging promises speed, if the company culture is toxic and the careers page radiates warmth, that’s inflating perception above reality. That’s the confidence man’s watch trick with better typography. The same mechanism shows up in physical products. A cheap gadget with unnecessary weight added to the housing so it “feels” substantial. A mid-tier brand that plasters a celebrity endorsement across the packaging so the product inherits status it didn’t earn. These are perception inflation through physical and social signals: manufacturing the feeling of quality without the substance of it. The channel changes. The trick doesn’t. The Sincerity Test: If the user fully understood what this design choice does, would they feel served or exploited? This one comes from Friestad and Wright’s Persuasion Knowledge Model (1994). Their research showed something that most designers don’t realize: people don’t object to being persuaded. We know ads try to sell us things. We know stores are designed to move us through a path. We know salespeople are working an angle. We’re fine with it, mostly, as long as we believe the persuasion is sincere. What triggers resistance, what Friestad and Wright call the “change of meaning” moment, is when the user realizes the persuasion was designed to benefit someone other than them. Campbell and Kirmani expanded this in 2000, showing that users who detect an ulterior motive don’t just resist the specific tactic. They downgrade their entire evaluation of the source. Trust doesn’t bend. It breaks. Perceived intent changes everything. Langer, Blank, and Chanowitz demonstrated this in 1978 with a study so clean it still holds up nearly fifty years later. A researcher approached people waiting at a copy machine and asked to cut in line. Three conditions. “Excuse me, may I use the machine?” got about 60% compliance. “Excuse me, may I use the machine, because I’m in a rush?” got 94%. And “Excuse me, may I use the machine, because I need to make copies?” got 93%. That last reason is meaningless. Everyone at a copy machine needs to make copies. But the word “because” signaled intentionality, signaled a reason existed, and that was enough. People processed the structure of the request and waved it through without evaluating the content. But only for small asks. When the request got costly (20 or more pages), the placebic reason stopped working. People started evaluating what came after the “because.” This is the dynamic that matters for design ethics. A tiny dark pattern, a pre-checked newsletter box, a slightly confusing unsubscribe flow, gets waved through the same way. The user processes the structure (“this looks like a standard form”) and moves on. But when the cost escalates, when the stakes get personal, users start reading the actual intent behind the design. And when they realize the “because” was empty, when they see that the unsubscribe button was tiny on purpose, that the cancellation flow was deliberately obstructed, trust doesn’t erode gradually. It collapses. The same action perceived as accidental gets patience. Perceived as deliberate, it gets fury. This response is not cultural. It is developmental. Fehr, Bernhard, and Rockenbach (2008) demonstrated egalitarianism in children as young as three, published in Nature. McAuliffe and colleagues (2017) reviewed the developmental evidence and found that children across cultures engage in costly punishment of unfairness, rejecting unequal offers even when rejection costs them personally. Fairness is not a learned preference. It is a developmental constant that appears before conscious reasoning is fully online. When a design promises one thing and delivers another, the response is visceral and immediate because it triggers mechanisms that predate the user’s ability to articulate why they’re angry. So the sincerity test isn’t “is this persuasive?” Everything I design is persuasive. The test is: if I pulled back the curtain, if I showed the user exactly how this layout directs their attention, exactly how this color palette builds trust, exactly how this copy sequence moves them toward the CTA, would they feel like I was working for them? Or against them? The Golden Rule: Would I consent to being influenced by this technique if I were the user? This one doesn’t need a citation. It needs honesty. I browse the web. I buy products. I fill out forms and sign up for trials and click “accept” on terms of service. When I encounter a design that respects my time and intelligence, that makes a genuinely good thing easy to find and evaluate and buy, I appreciate it. When I encounter a design that hides the unsubscribe button, pre-checks the newsletter box, makes the “decline” option look like a guilt trip (“No thanks, I don’t want to save money”), I remember. And I don’t come back. Every time Microsoft uses an update cycle to ask whether I want OneDrive as my default storage, I feel the same violation. Every time Adobe pushes a new feature tooltip on app launch that I have to close before I can work, same thing. Software that uses updates to reset user preferences or push unwanted services is the digital equivalent of a restaurant that brings you a more expensive dish than you ordered and hopes you won’t send it back. Microsoft and Adobe survive it because of monopoly lock-in, not because users forgive them. The golden rule test fails every time. No one at Microsoft wants their OS settings overwritten by a vendor during a security update. The golden rule is the final filter because the other two tests can be gamed. You can argue that a misleading progress bar “brings perception closer to the reality of completion.” You can argue that an aggressive upsell “serves the user’s need for the premium tier.” But you can’t honestly say you’d want to be on the receiving end of a confirm-shaming modal. Not if you’re being straight with yourself. Dark Patterns Are Cheap Laughs Dark patterns are the cheap laughs of design. I mean this literally, not metaphorically. In comedy, a cheap laugh is a joke that gets a reaction through shock, crudeness, or punching down. It works once. The audience laughs. But they don’t respect you for it, and they don’t come back because of it. A cheap laugh trades long-term audience trust for a short-term metric (laughter in the room right now). Harry Brignull coined the term “dark patterns” in 2010, formalizing the concept in A List Apart the following year. Gray, Kou, Battles, Hoggatt, and Toombs published the definitive taxonomy in 2018: nagging, obstruction, sneaking, interface interference, forced action. Every one of these patterns converts short-term. The hidden subscription charges. The roach motel you can check into but can’t check out of. The misdirection that moves the “cancel” button between screens. They all produce the number someone asked for on the dashboard this quarter. And they all destroy trust. Cheap laughs get old fast. I get it, I get it, yeah yeah, another one, wow. The audience checks out because the pattern is transparent and the comedian isn’t risking anything. Del Close and Charna Halpern wrote the book on this, literally. Truth in Comedy (1994), the manual that came out of iO Theater. Their core principle: “The truth is funny. Honest discovery, observation, and reaction is better than contrived invention.” One of the biggest mistakes a performer can make is trying to be funny. If the audience senses you’re reaching for the laugh, you’ve made the job harder. Real humor comes from finding the joke in the reality of the moment, not from sacrificing reality to crack a cheap one. Close called the cheap version “going for the joke.” Sacrificing the truth of a scene to get a reaction. It works once. But generosity, making your scene partner and their ideas look as good as possible, that’s what builds the group mind. That’s what makes an ensemble feel like they’re reading each other’s thoughts. The audience can’t fake-detect generosity because it’s not a technique. It’s an orientation. Tina Fey is the clearest example. She came up through iO and Second City, trained directly in the Del Close lineage, and her whole approach is this principle in action. Her run on SNL, first as head writer then Weekend Update with Amy Poehler, was built on elevating the people around her rather than centering herself. The 30 Rock writers’ room ran on “yes, and” as a production methodology, not just a comedy rule. She designed the systems that made other people funnier. That’s generosity as architecture, not generosity as performance. Expensive laughs are sustainable and scalable because they’re rooted in truth. Comedians build entire careers on them. It takes longer to set up. It requires understanding the room. It doesn’t always land on the first try. But when it works, it builds a relationship between the comedian and the audience that compounds over time. They come back. They bring friends. They buy the special. PFD is the expensive laugh. Genuine emotional resonance built on real value, designed to compound. It takes more skill. It requires the product to actually be good. It doesn’t produce a spike on the dashboard this quarter. But the clients I’ve worked with for five, seven, nine years didn’t stay because I tricked their users into converting. They stayed because their users kept coming back. But expensive laughs depend on the comedian’s judgment in the moment. What happens when the comedian gets tired, or the club owner is pressuring them to do the crowd-pleasing bit they know is cheap? That’s the limit of individual ethics. It took me years to articulate this as a distinction: operational ethics versus structural ethics. Operational vs. Structural Ethics The three tests are operational ethics. One designer, one review session, one decision. Does this button placement pass the alignment test? Does this copy pass the sincerity test? Would I accept this modal as a user? One practitioner, one moment, one call. Operational ethics are necessary. They’re also insufficient. They assume the designer has good values, clear judgment, and the organizational power to act on both. In practice, designers get tired. They get pressured. They get incentivized. A PM says “we need this metric up by Thursday.” A client says “make the cancel flow harder.” A stakeholder says “can we just default them into the annual plan?” And the designer, who passed the ethics test on Monday, cuts a corner on Friday because the sprint needs to close. Structural ethics work differently. They build the ethical constraints into the system itself, so they hold regardless of any individual practitioner’s judgment on any given Tuesday. My business model is structural ethics. Build, Host, Retain. I don’t do one-off projects without ongoing hosting. I don’t build a site and disappear. If I inflate perception beyond reality, I’m the one who has to maintain the gap. I’m the one fielding support tickets from users who feel misled. I’m the one watching the analytics when the return rate climbs and the reviews turn negative. That structure filters out a specific type of client before the engagement even starts: the ones who want perception inflated beyond reality. They don’t want a long-term relationship. They want a quick flip. They want someone to make their mediocre product look premium, take the money, and move on. My model doesn’t serve that. It’s not designed to. It’s not that I’m more ethical than other designers. It’s that my business model makes it structurally expensive to be unethical. The incentives are aligned with the tests. That’s the point. Operational ethics assumes good values. Structural ethics enforces them. A framework that relies only on operational ethics fails when the practitioner is rushed or incentivized to cut corners. A framework that relies only on structural ethics fails when a novel situation falls outside the rules. PFD uses both. What happens when structural ethics are absent Cory Doctorow named it in 2023: enshittification. Platforms follow a three-stage decay cycle. First, be good to users to attract them. Then, abuse users to benefit business customers (advertisers, agencies, merchants). Finally, abuse business customers to extract maximum value for shareholders. The business model incentivizes progressive exploitation. Each stage feels rational to the people running it. The structure makes the unethical outcome the default outcome. Facebook is the textbook case. Between 2015 and 2018, they inflated video viewing metrics by 150 to 900 percent. Not a methodology dispute. Not a rounding error. The amended class-action complaint revised the initial estimates upward; Facebook settled for $40 million. The inflated numbers told media companies that video was the future. CollegeHumor gutted its editorial staff. Funny or Die laid off most of its writers. Mic, a news outlet with tens of millions of monthly readers, fired its entire editorial team and pivoted to video production. Within two years, Mic sold for a reported $5 million after being valued at $100 million. That’s the confidence man’s watch trick at planetary scale. Facebook fabricated the perception of video value, and an entire industry restructured around the lie. These companies had profitable text and image operations. They destroyed them to chase metrics that were fake. Build-Host-Retain is the inverse structure. My model makes it expensive to inflate perception because I’m the one who lives with the consequences. Enshittification makes it expensive not to inflate perception because every stage of the decay cycle rewards the company that extracts more from its users. The incentives point in opposite directions. That’s not an accident. It’s the whole point of structural ethics. And the harm compounds. Ardoline and Lenzo (2025) published the first peer-reviewed paper connecting enshittification to cognitive harm. They introduced the concept of cognitive deskilling: when platforms degrade, users who have offloaded cognitive tasks to those platforms lose the capacity to perform those tasks independently. The damage is not transactional. Users don’t just lose a service. They lose the ability to do what the service used to do for them. Doctorow endorsed the paper. The American Dialect Society had already named “enshittification” its 2023 Word of the Year. The Fluency Trap There’s a thing I need to name honestly, because if I don’t, someone else will, and they’ll be right to. I call it the fluency trap. Processing fluency, the mechanism at the heart of Layer 2, is value-neutral. “If it’s easy to process, it feels true.” That’s Reber and Schwarz, 1999. The effect doesn’t care whether the thing being processed is actually true. A fluent lie feels truer than a disfluent truth. A clean, well-designed scam site feels more trustworthy than a legitimate but poorly designed one. The mechanism doesn’t evaluate content. It evaluates processing ease. And it assigns truth-value based on that ease. This means that by optimizing for fluency, by making my designs smooth, consistent, easy to process, I’m making it harder for users to engage the critical evaluation that would catch deception. The very thing that makes PFD effective (reducing prediction error, keeping the user’s processing smooth through the desired path) is the same thing that can prevent them from stopping to question whether they should. This is ontic occlusion baked into the mechanism. I borrowed that term from a colleague and friend, Cory Knobel, who uses it to describe how one representation of reality blocks another from being seen. My framework makes certain things visible (perception gaps, trust signals, fluency) and in doing so makes other things harder to see (the question of whether fluency itself should be interrupted). The harm compounds over time. Ardoline and Lenzo’s cognitive deskilling research (2025) applies directly here. When a platform optimizes for fluency unethically, users who offload evaluation to that platform’s cues (the smooth interface, the trustworthy layout, the frictionless flow) gradually lose their capacity to evaluate independently. Unethical fluency optimization does not just suppress critical evaluation in the moment. It degrades the user’s capacity for critical evaluation over time. The dark pattern makes the user worse at recognizing dark patterns. That is not a side effect. It is the mechanism working as designed. Pennycook and Rand (2019) found that susceptibility to fake news was driven by insufficient analytical thinking, not partisan bias. People who scored higher on analytical reasoning tests were better at distinguishing fake from real news regardless of political alignment. The problem was not ideology. The problem was insufficient cognitive engagement. Designs that reduce cognitive engagement, that make everything smooth and easy and frictionless, make users more vulnerable to deception. That is not a misuse of fluency. It is fluency doing exactly what it does, applied without the ethical constraints that determine whether it serves the user or exploits them. I don’t have a clean solution for this. The three ethical tests help. If the alignment test is honest, then fluency is being deployed on behalf of truth, not against it. But the tests depend on the designer’s judgment about what “truth” and “reality” are, and those judgments are themselves subject to the fluency trap. I know my own biases better than most people know theirs (the autism helps with that, actually; I’m relentlessly self-monitoring). But knowing your biases doesn’t eliminate them. It just means you can name what you’re probably missing. The honest statement is this: PFD is a loaded weapon. The safety is the designer. If the designer’s judgment is compromised, by incentives, by self-deception, by honest ignorance, the weapon still fires. Steve Krug didn’t address ethics. He didn’t need to. His book is about usability, about reducing cognitive friction so users can accomplish their goals. There’s not much ethical ambiguity in making a navigation menu clearer or reducing the number of form fields. Usability optimization is, for the most part, unambiguously good for the user. PFD doesn’t operate in the usability layer. It operates in the persuasion layer. The perception layer. The layer where you’re not just removing barriers to what the user already wants to do, but actively shaping what they notice, what they trust, what they feel, and what they do next. That layer demands ethical guardrails in a way that usability never did. That omission isn’t a criticism. It’s a scope observation. His book covers its territory completely and well. But if you take the techniques in this book and apply them without the ethical framework in this chapter, you are building confidence tricks. Fluent ones. Effective ones. The kind that don’t trigger the persuasion knowledge response because you’ve designed them not to. That’s not what this framework is for. The mandate, stated plainly: Perception-First Design removes perception barriers between users and genuine value. It does not create the perception of value where none exists. The designer is responsible for the perception layer. The organization is responsible for the value layer. When these diverge, the designer’s obligation is to the user. Simply Smart Home is the clean case. The product was good. Families genuinely stayed more connected using those digital picture frames. The tablets worked. The price was fair for the functionality. But the website looked like an overseas knockoff. The marketing led with feature specs instead of emotional connection. The brand system was a template that communicated “$25” when the product justified “$150.” Perception was below reality. I closed the gap upward. Revenue tripled. Not because I manufactured desire. Because I removed the perception barriers that were blocking people from seeing value that actually existed. The violation occurs in the opposite direction. If the product was mediocre and I made the site look premium, if the service was unreliable and I designed the experience to feel seamless, if the value wasn’t there and I manufactured the perception of it, that’s the confidence man with the watch. That’s the cheap laugh. That converts this quarter and craters next year. I think about this more than most designers do. Partly because of the framework. When you name the mechanisms explicitly, when you write down “processing fluency makes things feel true whether they are or not,” you can’t pretend you don’t know what you’re holding. The knowledge creates the obligation. Partly because of the nightclub. At the door, I had power over people’s nights. I decided who got in and who didn’t. I could have abused that power, and I saw other bouncers who did. The ego trip. The petty gatekeeping. The “that violates our dress code” or turning people away for some made-up reason as a proxy for “I don’t like you.” That was perception manipulation too, just crude. I chose to use the position differently. To make the experience better for everyone who showed up, regardless of who they were. To conduct instead of command. And partly because I’ve spent years building for communities that have been on the wrong end of perception manipulation for generations. Communities that know what it feels like when someone else controls how they’re seen. You develop a different relationship with the tools when the people you serve have been shaped by other people’s perceptions their entire lives. You don’t get to be cavalier about the power. The oath isn’t complicated. Three tests. Sixty seconds. Before anything ships. Does this bring perception closer to reality? If the user knew what this does, would they feel served? Would I accept this as the user? If all three pass, ship it. If any one fails, redesign it. That’s it. No manifesto. No certification. No twelve-step program. Just three questions and the honesty to answer them. The hard part was never the test. The hard part is the honesty. Next: What I Don’t Know Yet, on the questions this framework can’t answer and the blind spots I haven’t figured out how to see past. Key Terms The Alignment Test Does this design bring perception closer to reality, or further from it? Based on Berdichevsky & Neuenschwander (1999). The Sincerity Test If the user fully understood what this design choice does, would they feel served or exploited? Based on Friestad & Wright’s Persuasion Knowledge Model (1994). The Golden Rule Would I consent to being influenced by this technique if I were the user? Dark patterns Brignull (2010/2011), Gray et al. (2018). Design patterns that convert short-term by exploiting users: nagging, obstruction, sneaking, interface interference, forced action. The cheap laughs of design. Operational vs. structural ethics Operational: three tests run by one designer on one decision. Structural: ethical constraints built into the business model itself, holding regardless of individual judgment. The fluency trap Processing fluency is value-neutral. A fluent lie feels truer than a disfluent truth. By optimizing for fluency, PFD can make it harder for users to engage critical evaluation. Ontic occlusion Knobel. Any representation of reality blocks other representations from being seen. PFD makes perception visible and in doing so occludes other concerns. Enshittification Doctorow (2023). Three-stage platform decay: good to users, then abuse users for business customers, then abuse business customers for shareholders. The structural ethics failure case. Cognitive deskilling Ardoline & Lenzo (2025). When platforms degrade, users who offloaded cognitive tasks lose the capacity to perform those tasks independently. Harm compounds over time. References Berdichevsky & Neuenschwander (1999) Toward an ethics of persuasive technology. Communications of the ACM, 42(5), 51–58. Friestad & Wright (1994) The Persuasion Knowledge Model. Journal of Consumer Research, 21(1), 1–31. Campbell & Kirmani (2000) Consumers’ use of persuasion knowledge. Journal of Consumer Research, 27(1), 69–83. Langer, Blank & Chanowitz (1978) The mindlessness of ostensibly thoughtful action: The role of “placebic” information in interpersonal interaction. Journal of Personality and Social Psychology, 36(6), 635–642. Brignull (2011) Dark Patterns: Deception vs. Honesty in UI Design. A List Apart, November 1, 2011. Gray, Kou, Battles, Hoggatt & Toombs (2018) The dark (patterns) side of UX design. CHI ’18 Proceedings. Doctorow (2023) TikTok’s Enshittification. Wired / Pluralistic. Ardoline & Lenzo (2025) The Cognitive and Moral Harms of Platform Decay. Ethics and Information Technology, 27, 37. Fehr, Bernhard & Rockenbach (2008) Egalitarianism in young children. Nature, 454(7208), 1079–1083. McAuliffe et al. (2017) The developmental foundations of human fairness. Nature Human Behaviour, 1, 0042. Pennycook & Rand (2019) Lazy, not biased: Susceptibility to partisan fake news is better explained by lack of reasoning. Cognition, 188, 39–50. Reber & Schwarz (1999) Effects of perceptual fluency on judgments of truth. Consciousness and Cognition, 8(3), 338–342. Knobel (2010) Ontic Occlusion and Exposure in Sociotechnical Systems. Doctoral dissertation, University of Michigan. A note on how this was written: This series is AI-assisted. I provide the stories, the methodology, the case studies, and the editorial direction. AI helps me structure and draft. This is consistent with Perception-First Design’s own transparency principle: if I’m writing about perception, I should be honest about how the writing itself is produced. --- ## Chapter 12: What I Don't Know Yet URL: https://aurochs.agency/writing/make-me-think/what-i-dont-know-yet/ Every framework is a flashlight. It illuminates what it points at and leaves the rest in the dark. I’ve spent eleven chapters telling you what Perception-First Design sees. The five layers, the diagnostic, the ethics, the cases, the science. And all of it is real. It works. I’ve watched it work for fifteen years across dozens of projects, from a nightclub in Santa Barbara to Costco pallets to improv theaters to the site you’re reading this on. The pattern holds. But a framework that doesn’t name its own blind spots isn’t confident. It’s careless. So this final chapter is about what I don’t know, what PFD doesn’t see, and what I haven’t figured out yet. Not as a performative gesture of humility. As an honest accounting from someone who takes this seriously enough to tell you where the edges are. The WEIRD Problem The evidence base for PFD is predominantly WEIRD: Western, Educated, Industrialized, Rich, Democratic. That’s the term Henrich, Heine, and Norenzayan coined in 2010 to describe the narrow sample that most psychology research draws from. And it applies directly to everything I’ve built. The underlying cognitive architecture, predictive processing, working memory limits, first-impression formation, that stuff is expected to generalize across humans. Brains are brains. Clark’s prediction machine model doesn’t stop working when you cross an ocean. But the parameters are culturally shaped. Which visual patterns signal trust. What counts as “too much” cognitive load. How processing fluency maps to credibility. Whether a face in the hero section reads as warmth or as surveillance. These are not universal constants. They’re learned associations, and they’re learned differently in Osaka than in Ohio. I’ve applied PFD in Western e-commerce and entertainment. That’s my dataset. Whether the five-layer dependency order holds in collectivist markets, in cultures with non-Latin typographic traditions, in markets where trust signals are fundamentally different from what I’ve calibrated against, that was an open question when I first wrote this book. Recent cross-cultural research has resolved part of it. Takahiko Masuda and Richard Nisbett showed in 2001 that East Asian observers attend to background and relational context while Western observers focus on foreground and focal objects. Four years later, Hannah Chua, Julie Boland, and Nisbett replicated this with eye-tracking: the same scenes produce different scanning patterns depending on the viewer’s cultural background. Joshua Goh and his colleagues found cultural differences in face-processing cortex activation in 2010. And Li-Jun Ji, Nisbett, and Yanjie Su had shown earlier, in 2001, that prediction direction itself varies cross-culturally. Westerners tend to assume linear continuation. East Asians more often anticipate non-linear reversals. What this evidence resolves: the cognitive architecture is universal, the calibration is cultural. The five-layer stack holds across populations. What differs is the parameters. L1 calibration (what triggers trust) and L3 calibration (what predictive priors load automatically) have to be validated against target-market users before any design rolls across regions. Meta-rule: layer requirements are universal; layer calibration is cultural. PFD cannot claim cross-cultural universality without this step. What the evidence doesn’t resolve: L2 processing-fluency effects across non-Latin typographic traditions, L4 decision-architecture priors across markets with different default authority structures, and whether the dependency-stack ordering itself holds invariant across high-context vs low-context cultures. Those stay open. Partially resolved is not the same as resolved. But it’s also not the same as unknown. The architecture-vs-calibration distinction is now a working premise, not a guess. The Measurement Gap The 5-Minute Perception Audit is the most accessible thing in PFD. It’s the part I can teach a business owner in an hour. It’s directionally correct, fast, and useful. It’s also not validated. I mean that in the technical sense. There’s no inter-rater reliability data. I haven’t run a study where two PFD practitioners independently audit the same site and I measure whether they converge on the same layer diagnosis. I haven’t compared audit outcomes against controlled user research to see if the heuristic catches what the empirical methods catch. The audit is based on my pattern library, my sensitivity, my years of doing this. It works for me. I believe it works for others who learn it. But I don’t have the numbers to prove it. The gap between “expert intuition with a framework” and “teachable, repeatable diagnostic” is real. Right now PFD sits closer to the first one. Getting it to the second would mean running the kind of validation studies that turn a practitioner framework into something publishable in a peer-reviewed journal. That’s work I want to do. I haven’t done it yet. I’m naming this because I think the honesty is more useful than the spin. A lot of design methodologies claim rigor they don’t have. The “proprietary process” that’s really just “the person who built it is good at design.” I don’t want to be that. If PFD’s diagnostic is going to claim it’s teachable and repeatable, I need the evidence. I’m working on it. First Encounters vs. Visit 100 PFD describes first encounters well. That’s what the five-layer stack was built for. Activation points, first impressions, the mental waterfall from cognitive load to decision architecture. All of it is optimized for what happens when someone arrives at your site for the first time. But user relationships evolve. And I’m not sure how the framework applies to visit 100. At visit 100, the prediction errors that activated on visit 1 are now expected. The user’s brain has updated its model. Your site is no longer novel. The fluency that felt trustworthy on first encounter has become invisible, the way you stop noticing the hum of your refrigerator. The activation points that broke autopilot on the first visit are now part of the autopilot. Does the five-layer stack apply the same way to retention? To long-term engagement? To the moment when a loyal user starts to drift? I think it does, but in a different mode. The dependency order probably holds. But the interventions at each layer would need to be different. You’re not designing for first-impression trust anymore. You’re designing against habituation. Against the slow fade of attention that comes from familiarity. There’s research on this. Berlyne’s work on novelty and arousal from the 1970s. More recent work on hedonic adaptation. But I haven’t integrated it into PFD in any systematic way. The framework is weighted toward first contact, and I haven’t done the work to extend it to the full lifecycle. That’s a gap, and it matters for anyone using PFD on a product with a long user relationship. The Calibration Problem Here’s the one that keeps me up at night. The diagnostic starts with Feel. Arrive at the page. Let the emotional response fire before your conscious brain translates it. “This makes me feel X.” That pre-verbal read is the primary instrument of the entire methodology. But that instrument is calibrated by my own culture, class, neurotype, and aesthetic history. I named this in Chapter 3: the autism gives me analytical social cognition, the ADHD gives me friction sensitivity. I called it a tuning fork. And it is. But every tuning fork has a frequency, and mine is not universal. I’m a half-Mexican, half-Ukrainian, college-educated, neurodivergent man who grew up in Southern California and lives in San Francisco. My “feel” is calibrated by that. The things that trigger my friction response, the things that register as warm or cold, the things that read as trustworthy or sketchy, those responses are shaped by my life, not by some objective standard. I named my neurodivergence as an advantage. Heightened sensitivity to friction. And it is. But heightened sensitivity to friction could also mean under-sensitivity to other things. Social warmth cues that are obvious to neurotypical designers. Cultural signifiers outside my experience. Accessibility needs I don’t personally feel. Emotional tones that register differently for people whose nervous systems are wired differently from mine. The diagnostic instrument has its own biases. And PFD doesn’t yet have a protocol for calibrating them. I’ve told you to “know your instrument” and “compensate for its tuning.” That’s honest advice. But it’s also incomplete. Knowing your own biases is necessary. It’s not sufficient. You need external input, diverse perspectives, real user data, to catch what your instrument misses. And right now, PFD doesn’t formalize that. The Feel step is personal. The correction for the Feel step’s blind spots is... informal. That needs to get better. Accessibility: Optimization vs. Rights Accessibility sits in an uncomfortable place in PFD. The framework handles it through Layer 0 (cognitive load reduction) and Layer 2 (processing fluency). This produces good accessible design in practice. Sites built with PFD tend to be more accessible than average because the framework naturally prioritizes clear hierarchy, readable text, manageable information density, and consistent interaction patterns. The curb cut effect I described in Chapter 3 is real, and it runs through the whole methodology. But disability justice scholars make a different argument, and I think they have a point. They say accessibility is a rights issue, not an optimization target. And PFD’s lens is fundamentally utilitarian. It frames accessibility through “does this reduce cognitive load” and “does this improve processing fluency.” Those are optimization questions. They produce good outcomes, but the reasoning is instrumental. You optimize because it converts better, because it reaches more users, because it reduces friction across the board. The disability justice framing says: you make things accessible because people deserve access. Full stop. Not because it improves your conversion rate. Not because it happens to benefit other users too. Because excluding people from designed experiences is a harm, and the obligation to not cause that harm doesn’t depend on whether it helps your bottom line. I think both frames are true. I think PFD’s utilitarian lens gets you 80% of the way to good accessible design, and the rights-based frame covers territory the optimization lens never reaches. But I also think there are cases where they conflict. Where the utilitarian calculation says “this accessibility investment doesn’t improve outcomes for enough users to justify the cost” and the rights-based calculation says “doesn’t matter, do it anyway.” PFD currently lives in the first calculation. The framework gives you tools to build accessible design. It doesn’t give you a reason to build accessible design when the tools say it’s not worth it. That’s a gap. And I’m not sure a practitioner framework can close it. It might require something more like a value commitment that sits outside the framework entirely. The AI Question And then there’s the question I get asked most often by other designers: what happens when AI generates the design? PFD’s diagnostic starts with a human body. The Feel step, the pre-verbal emotional read, the “neurodivergent tuning fork” that resonates at the slightest vibration. The methodology is built around the idea that the designer’s own perceptual system is the primary instrument. You feel the design before you analyze it. You feel the friction before you name the layer. Can an AI do that? I’m genuinely uncertain. Not as a rhetorical gesture. I don’t know. Current AI systems can analyze visual hierarchy, check contrast ratios, evaluate reading level, flag inconsistent spacing. They can run the mechanics of the Perception Audit faster than I can. But the Feel step isn’t mechanics. It’s the pre-verbal emotional response of a nervous system that has spent decades being shaped by culture, aesthetics, social interaction, and the particular way my brain processes sensory input. If the diagnostic can be automated, PFD scales beyond the individual practitioner. That’s exciting. But what’s lost when the instrument is no longer a human body? The whole methodology grew from a specific kind of embodied sensitivity. The friction I feel in a cluttered interface. The warmth I notice when a hero section gets the first impression right. The unease I can’t name until I’ve sat with it for ten seconds and let the feeling resolve into language. An AI running the diagnostic would be running a different diagnostic. Maybe a better one in some dimensions. Faster, more consistent, less biased by the designer’s own calibration. But the pre-verbal feel, the part that catches things before they have names, I don’t know if that transfers. And if it doesn’t, then PFD has a ceiling on how far it scales without a practitioner in the loop. It’s a methodology and tool for practitioners to automate work they can vet, not a system that replaces the practitioner entirely. I think the future is both. AI handles the analysis, the layer-by-layer audit, the pattern matching. The human handles the feel. But I haven’t figured out where the handoff goes, and I’m not pretending I have. Ontic Occlusion Cory Knobel introduced me to a concept that applies directly to this chapter: ontic occlusion. The idea that any representation of reality blocks other representations from being seen. Every lens that makes something visible makes something else invisible. PFD makes perception visible. It makes the gap between how things are and how things are experienced into something you can diagnose and design for. But in doing that, it occludes things. Structural power dynamics. PFD’s ethical test says “align perception with reality.” But it doesn’t interrogate who decides what “reality” is. When Simply Smart Home’s product page undersold a good product, the reality was clear. But what about a company whose product is mediocre and whose “reality” is ambiguous? The designer choosing which reality to align with is itself a power move. PFD doesn’t surface that. Individual vs. collective effects. PFD models one user’s cognitive journey. It doesn’t model what happens when thousands of users are conducted through the same waterfall simultaneously. Individual perception optimization can produce collective harms: engagement loops, attention extraction at scale, homogenization of choice. The methodology’s unit of analysis is the user. The effects that emerge at population scale are outside its frame. The fluency trap. Processing fluency is Layer 2’s core mechanism: if it’s easy to process, it feels true. That mechanism is value-neutral. It works for honest brands and dishonest ones. A fluent lie is more persuasive than a disfluent truth. PFD’s ethical tests catch deliberate manipulation. But they don’t address the structural problem: by optimizing for fluency, the framework makes it harder for users to engage the critical evaluation that would catch deception. The very thing that makes PFD effective (reducing prediction error, keeping users on the path) is the same thing that can prevent them from stopping to question. That’s ontic occlusion baked into the mechanism. And I don’t have a clean answer for it. I started this book with a story about a barback getting chewed out by his boss on a Tuesday night. The boss saw one snapshot, formed an impression, acted on it like it was the full picture. His perception didn’t match reality. Twelve chapters later, I’ve given you a framework for seeing those gaps and designing for them. Five layers. A diagnostic. An ethical test. Specific cases where it worked. Specific science that explains why. And now I’ve given you the list of things I haven’t figured out. The cultural limits of my evidence base. The measurement gap in my diagnostic. The temporal dynamics I haven’t modeled. The biases in my own instrument. The accessibility frame that my utilitarian lens can’t fully hold. The AI question that could either extend the methodology or break its foundational assumption. The structural blind spots that the framework occludes by the very act of making perception visible. A living methodology names its edges. That’s what this chapter is for. Not to undermine what came before, but to mark the places where the next work needs to happen. Mine, or someone else’s. I believe this framework is true. I believe the layers are real, the dependency order is real, the diagnostic works. I believe the science supports it and the results demonstrate it. I also believe that the version of PFD I’ve described in these twelve chapters is an early version of something that could be much more complete. The bouncer who got promoted because he noticed a perception gap, the designer who turned that instinct into a methodology, the person writing this sentence right now, we’re all the same person at different points in the same long process of figuring out how humans actually perceive, and what to do about it. Design is intentionality from our minds into reality. That’s stupidly profound when you sit with it. Design is a lens of empathy to solve with. For whom. Always for whom. Bill Watterson, Calvin and Hobbes Afterword Thank you for reading this. The whole thing. That means more than I can put into a sentence. If something landed, or if something didn’t, send me an email and tell me your thoughts. I’d love to hear from you. This book started as notes in a bound notebook and became a methodology and then became whatever this is. It wouldn’t exist without the people who believed in me while I was figuring it out. My loved ones, my chosen family, the friends who listened to me talk about processing fluency at dinner and didn’t change the subject. You know who you are. Thank you for being there along the journey and for believing in me when the evidence was mostly just enthusiasm and a spiral of half-formed ideas. And to you, the reader: I hope something in these twelve chapters changed how you see. That’s all I was trying to do. Stefan Kovalik San Francisco, 2026 Key Terms WEIRD Henrich, Heine & Norenzayan (2010). Western, Educated, Industrialized, Rich, Democratic. The narrow cultural sample most psychology research draws from. PFD’s evidence base is predominantly WEIRD. Calibration vs. architecture The cognitive architecture (predictive processing, working memory limits, face processing) generalizes across humans, confirmed by cross-cultural research (Masuda & Nisbett, 2001; Chua et al., 2005; Goh et al., 2010; Ji et al., 2001). The calibration (what signals trust, what counts as too much load) is culturally shaped and must be validated against target-market users before cross-regional rollout. Hedonic adaptation The tendency for the impact of repeated stimuli to diminish over time. PFD is optimized for first encounters but hasn’t been systematically extended to visit 100. Ontic occlusion Knobel. Any representation of reality blocks other representations from being seen. Every lens that makes something visible makes something else invisible. The fluency trap Processing fluency is value-neutral. Optimizing for fluency can prevent users from engaging critical evaluation. The safety is the designer. References Henrich, Heine & Norenzayan (2010) The weirdest people in the world? Behavioral and Brain Sciences, 33(2–3), 61–83. Masuda & Nisbett (2001) Attending holistically versus analytically: Comparing the context sensitivity of Japanese and Americans. Journal of Personality and Social Psychology, 81(5), 922–934. Chua, Boland & Nisbett (2005) Cultural variation in eye movements during scene perception. Proceedings of the National Academy of Sciences, 102(35), 12629–12633. Goh et al. (2010) Culture differences in neural processing of faces and houses in the ventral visual cortex. Social Cognitive and Affective Neuroscience, 5(2–3), 227–235. Ji, Nisbett & Su (2001) Culture, change, and prediction. Psychological Science, 12(6), 450–456. Clark (2013) Whatever next? Predictive brains, situated agents, and the future of cognitive science. Behavioral and Brain Sciences, 36(3), 181–204. Berlyne (1971) Aesthetics and Psychobiology. Appleton-Century-Crofts. A note on how this was written: This series is AI-assisted. I provide the stories, the methodology, the case studies, and the editorial direction. AI helps me structure and draft. This is consistent with Perception-First Design’s own transparency principle: if I’m writing about perception, I should be honest about how the writing itself is produced. --- ## The 5-Minute Perception Audit URL: https://aurochs.agency/writing/methods/the-5-minute-perception-audit/ Perception-First Design has five layers. This is one test per layer. The full diagnostic is in Make Me Think. This is the version you can run in five minutes. These tests work because perception is faster than language. LeDoux (1996) mapped the pathway: sensory input hits the thalamus and splits. The “low road” shoots straight to the amygdala in roughly 12 to 20 milliseconds. The cortical route, the one that gives you conscious awareness and words, takes 200 to 500 milliseconds (Kutas & Hillyard, 1980). The feeling arrives first. Zajonc (1980) argued from the behavioral side that affective reactions occur independently of, and prior to, cognitive processing. These five tests are designed to catch what that first feeling detected. For the full neuroscience, see Chapter 9: Feel, Unpack, Diagnose, Prescribe. The Five Tests Viewport Count (Layer 0: Cognitive Load) What to do: Pick any screen on your site. Count everything competing for attention at once. Nav items, CTAs, images, text blocks, form fields, popups. Cowan (2001) established that working memory holds roughly three to five chunks, not the commonly cited seven from Miller. That’s the budget. What it tells you: If a single viewport demands more than five focal points, L0 is failing and everything downstream is compromised. Every element beyond the limit becomes noise the brain has to actively filter, spending processing bandwidth on things that don’t carry meaning. Time: ~30 seconds 5-Second Test (First Impression, L1) What to do: Show your homepage to someone who has never seen it. Give them five seconds. Take it away. Ask three questions: What does this company do? How did it make you feel? Would you trust it with your credit card? What it tells you: Lindgaard et al. (2006) found that users form reliable aesthetic judgments of websites within 50 milliseconds. Five seconds is generous. If they still can’t answer the first question, your value proposition is invisible. If they wouldn’t trust it, your first impression is failing before anything else gets a chance to work. Time: ~60 seconds Squint Test (Processing Fluency, L2) What to do: Blur your eyes while looking at the page. What stands out? If everything looks equally important, nothing is. The squint test strips away details and shows you the raw visual hierarchy. What it tells you: If the hierarchy is clear when blurred, processing fluency is working. If it’s a uniform field, the brain has to sort it manually. Reber, Schwarz, and Winkielman (2004) demonstrated that processing fluency directly affects preference judgments: harder to process means less trustworthy, less likeable, less credible. Visual noise is not neutral. It actively degrades the experience. Time: ~30 seconds Analytics Gap (Perception Bias, L3) What to do: Compare what your users say versus what they do. Surveys say “love it” but conversion is flat? Users want “more features” but the simplified version converts better? What it tells you: Nisbett and Wilson (1977) demonstrated that people regularly misidentify the causes of their own behavior. The gap between stated preference and revealed preference is where perception bias lives, and recognizing it is one of the most important skills a designer can develop. Time: ~90 seconds Trail Test (Decision Architecture, L4) What to do: Open your site in incognito mode. Pretend you have never been there. Hunt for the core offer. Count clicks. Did you need the search bar? Were there dead ends? What it tells you: If the trail takes more than three clicks, or you needed the search bar, decision architecture needs work. Time: ~90 seconds Five tests, five layers, five minutes. This catches the biggest failures fast. For the full diagnostic: Feel, Unpack, Diagnose, Prescribe. For how requirements turn into solutions without contamination: The Generative Protocol. For the complete series: Make Me Think. Or run your site through the automated version: Forge. Key Terms Viewport Count Layer 0 test for simultaneous cognitive load. Count everything competing for attention in a single screen. 5-Second Test First impression trust evaluation. Five seconds of exposure, three questions, immediate recall. Squint Test Visual hierarchy and processing fluency check. Blur removes detail and exposes structure. Analytics Gap Stated versus revealed preference divergence. What users say they want versus what they actually do. Trail Test Navigation clarity and decision path assessment. Incognito, fresh eyes, count the clicks. References Zajonc (1980) Feeling and thinking: Preferences need no inferences. American Psychologist, 35(2), 151–175. Kutas & Hillyard (1980) Reading senseless sentences: Brain potentials reflect semantic incongruity. Science, 207(4427), 203–205. Damasio (1994) Descartes’ Error: Emotion, Reason, and the Human Brain. Putnam. LeDoux (1996) The Emotional Brain: The Mysterious Underpinnings of Emotional Life. Simon & Schuster. Nisbett & Wilson (1977) Telling more than we can know: verbal reports on mental processes. Psychological Review, 84(3), 231–259. Cowan (2001) The magical number 4 in short-term memory: a reconsideration of mental storage capacity. Behavioral and Brain Sciences, 24(1), 87–114. Reber, Schwarz & Winkielman (2004) Processing fluency and aesthetic pleasure: is beauty in the perceiver’s processing experience? Personality and Social Psychology Review, 8(4), 364–382. Lindgaard et al. (2006) Attention web designers: you have 50 milliseconds to make a good first impression! Behaviour & Information Technology, 25(2), 115–126. A note on how this was written: This post is AI-assisted. I provide the methodology, the diagnostic tools, and the editorial direction. AI helps me structure and draft. This is consistent with Perception-First Design’s own transparency principle: if I’m writing about perception, I should be honest about how the writing itself is produced. --- ## The Generative Protocol URL: https://aurochs.agency/writing/methods/the-generative-protocol/ In Make Me Think, I walk through five perception layers and a four-step diagnostic for reading a website the way a physician reads symptoms. That diagnostic, Feel, Unpack, Diagnose, Prescribe, tells you what’s broken and which layer is failing. This is the other half: how to build what’s right, not just diagnose what’s broken. The diagnostic and the generative protocol are different cognitive operations. Evaluating a design is pattern recognition: you see what’s there, you feel the response, you trace the failure to its source. Generating a design is constraint satisfaction: you gather requirements from each layer and find the solution that satisfies all of them simultaneously. One is forensic. The other is architectural. They need different rules. The Mistake Early in my career, I kept making the same error. I’d see a problem and immediately propose a solution. Client’s homepage feels cluttered? “Let’s simplify the navigation and add a stronger hero section.” Checkout conversion is low? “Let’s reduce the form fields and add trust badges.” Those aren’t bad recommendations. They’re reasonable, experienced, and often correct. They’re also contaminated. In clinical diagnosis, Pat Croskerry calls this premature closure: locking onto the first measurable finding instead of the actual root cause (Croskerry, 2009). The design version is identical. The moment you propose a solution, your brain stops looking for requirements. You’ve collapsed the problem space into a single point. Instead of asking “what does this design need to do?” you’re now asking “does my proposed solution work?” Those are different questions, and the second one is dangerously easier to answer. Confirmation bias takes over. You start seeing evidence that supports your proposal and filtering out evidence that contradicts it. Requirements that your solution doesn’t address become invisible, not because they don’t exist, but because you’re no longer looking for them. The solution has become the lens, and the lens only shows you what it can focus on. Tversky and Kahneman (1974) identified the mechanism: the availability heuristic. People overweight problems that are mentally accessible (the pricing page everyone’s been arguing about) and underweight problems that are harder to articulate (the vague feeling that the homepage “doesn’t feel right”). Once you’ve proposed a solution, that solution becomes the most available mental object. Everything else recedes. I watched this happen to myself dozens of times before I named it. I’d propose a solution in the first meeting, spend the next three weeks refining it, and then discover at launch that the real problem was two layers below where I’d been working. The fix was fine. It just fixed the wrong thing. Rule Zero Do not propose any solution until all five layers have been analyzed and requirements accumulated. That’s it. One rule. Everything else in the generative protocol follows from this. Solution-first thinking contaminates the requirement space. You stop seeing what the layers demand and start seeing what the solution provides. You pattern-match the solution against your expectations rather than deriving the solution from the constraints. And you miss requirements that no existing proposal addresses, because you never looked for them. Rule Zero is hard to follow because experience works against you. The more projects you’ve done, the faster your brain pattern-matches to a familiar solution. Gary Klein’s research on Recognition-Primed Decision making (Klein, 1998) showed that experts in high-stakes environments almost never compare options the way decision theory says they should. They recognize the situation as a variant of something they’ve seen before and mentally simulate the first plausible response. That’s powerful for firefighters. It’s dangerous for designers, because the “first plausible response” skips the requirement-gathering that makes the solution correct rather than just familiar. Senior designers are more susceptible to this than juniors, not less. Their pattern library is deeper, so the match happens faster, and the confidence that comes with experience makes it harder to hold the solution at arm’s length while you finish gathering requirements. The discipline is not “don’t have ideas.” Ideas will show up. The discipline is “don’t commit to any of them until the layers have spoken.” The Derivation In practice, I state the design problem in terms of user experience, not features. “Users can’t find the primary action” is a design problem. “We need a bigger button” is a solution masquerading as a problem. This distinction matters because the solution might not be a bigger button. It might be fewer competing elements (L0), a stronger visual hierarchy (Layer 2), or a better trail (Layer 4). Then I work each layer bottom-up, from cognitive load (L0) through Layer 4. For each layer, I ask three questions. What is the constraint (the biological or psychological hard limit this layer represents)? How does the current design violate it? What requirement does this generate? The requirement is a “must,” not a “should.” The design MUST do this. “Should” is negotiable. “Must” is not. If working memory holds three to five chunks (Cowan, 2001), then a viewport that demands twelve is not a suggestion to simplify. It is an L0 violation with a measurable cognitive cost. L0 (Cognitive Load): Can a first-time user complete the primary task without getting stuck? This is Sweller’s territory: Cognitive Load Theory (1988) distinguishes between intrinsic load (the complexity of the task itself) and extraneous load (the complexity your design adds on top of it). If navigation is ambiguous, if the information architecture reflects the org chart instead of user goals, if the user has to think about how to use the thing before they can think about whether to use the thing, L0 is failing. Every unit of extraneous load your interface demands is a unit of working memory stolen from the actual task. Requirement: the structural layer must be invisible. Layer 1 (First Impression): Does the page activate the right emotion in the first 50 milliseconds? The visitor’s System 1 fires a verdict before conscious processing begins. If that verdict is “cheap,” “cluttered,” or “I don’t trust this,” no amount of good copy downstream will reverse it. Requirement: the first impression must match the brand’s intended positioning. Layer 2 (Processing Fluency): Is every visual and verbal element consistent, hierarchical, and easy to process? Reber and Schwarz (1999) demonstrated that processing ease directly affects judgments of truth: if it’s easy to process, it feels true. Typography, spacing, color, alignment. The things that make a page feel “professional” without anyone being able to name why. Alter and Oppenheimer (2009) showed that even minor fluency manipulations shift judgments of truth, confidence, and trust. The effect compounds: twenty slightly wrong details create a cumulative feeling of “something’s off” that no single element accounts for. Requirement: processing effort must be minimized so cognitive bandwidth is spent on the message, not the medium. Layer 3 (Perception Bias): Does the design speak to what users respond to, not what they say they want? Nisbett and Wilson (1977) demonstrated that people regularly misidentify the causes of their own behavior. The gap between stated preference and revealed preference is where perception bias lives. Users say they want more options. Their behavior shows them converting on fewer. Requirement: design for observed behavior, not reported preference. Layer 4 (Decision Architecture): Does the trail lead to conversion without manipulation? The path from “I see what this is” to “I want this” must feel like walking, not searching. Requirement: the decision trail must be clear, natural, and honest. When all five layers have been analyzed, I have five requirements. All non-negotiable. The solution has to satisfy all of them simultaneously. When requirements conflict (and they sometimes do), lower-layer requirements win. The dependency stack is the tiebreaker. Only then do I derive the solution. What satisfies requirement 1 AND requirement 2 AND requirement 3 AND requirement 4 AND requirement 5? If no single intervention covers all five, what minimal set does? The solution emerges from the constraint intersection, not from brainstorming. Not from competitive analysis. Not from “what would Apple do?” This is slower than proposing a solution and iterating. Significantly slower. But it surfaces things that solution-first thinking systematically misses. The requirement that nobody had on their list because they were already thinking in terms of sidebars and carousels and hero images. The constraint that only becomes visible when you ask the layers in order. The Gate The Gate is the last piece. It runs before shipping. Each layer is pass/fail. If any layer below Layer 4 fails, do not ship. Downstream layers cannot compensate for upstream failures. This is the dependency stack applied as a quality gate, and it’s the hardest one to enforce because of schedule pressure, because of stakeholder expectations, because shipping something imperfect feels better than shipping nothing. Each layer has a specific gate question. Can a first-time user complete the primary task without getting stuck? That’s cognitive load (L0). Does the first impression activate the right emotion? That’s Layer 1. Is every visual and verbal element consistent? Layer 2. Does the design speak to what users respond to, not what they say they want? Layer 3. Does the trail lead to conversion without manipulation? Layer 4. I’ve shipped work that failed the gate. I’m not going to pretend otherwise. Sometimes the client’s timeline wins. Sometimes the budget doesn’t allow for the iteration needed. But I name the failure. I tell the client which layer isn’t passing and what the likely cost will be. That honesty is part of the practice. Berdichevsky and Neuenschwander (1999) argued that any technology designed to change attitudes or behaviors has an obligation to change them in the direction of truth. Not the designer’s truth. Not the client’s truth. Actual reality. The gate doesn’t exist to make shipping impossible. It exists to make failure visible and intentional rather than accidental and invisible. If I’m inflating perception above reality, I should know it and say it. That’s what The Oath demands. Anti-Patterns There are patterns of failure I see repeated across projects, across industries, across experience levels. Each one sounds reasonable until you see the downstream damage. Solution-first. “We need a sidebar.” “We need a chatbot.” “We need to redesign the homepage.” These are all solutions. None of them are problems. When the team starts with a solution, every subsequent conversation evaluates that solution rather than exploring the problem space. Requirements get shaped to justify the solution instead of the solution being derived from requirements. This is premature closure (Croskerry, 2009) applied to design: locking onto the first measurable intervention and never looking past it. Competitive copying. “Miro does X, so we should too.” “Our competitor has a three-column pricing page.” Competitors are solving their own problems in their own contexts with their own constraints. Copying their solutions imports their assumptions, and their assumptions might not match your users, your brand, or your layer stack. This is domain-driven design, not perception-driven design. Velocity-first. “We need results this quarter.” Brand systems take months. Campaigns take weeks. When the pressure hits, the campaign always wins, and the product work gets pushed to next quarter. Then the quarter after that. Then never. The campaign won the quarter. The product would have won the year. Checklist mode. “Does it have a hero section? Check. Does it have social proof? Check. Does it have a CTA above the fold? Check.” Evaluation by checklist catches obvious omissions but generates no insight. You can pass every item on a checklist and still have a page that feels wrong, because checklists evaluate components and perception operates on the whole. The Feel step exists for exactly this reason: the compressed System 1 signal catches what no checklist can. Skipping layers. Jumping from “the bounce rate is high” to “let’s add more CTAs” without diagnosing which layer is actually failing. The bounce rate might be a Layer 1 problem (the page doesn’t look trustworthy), but the intervention is Layer 4 (decision architecture). The fix doesn’t address the failure. The layer stack is a dependency chain. You can’t skip links. The generative protocol and the diagnostic are two halves of the same methodology. The diagnostic tells you what’s broken. The generative protocol tells you what to build. Both work the layers in order. Both refuse to skip ahead. Both treat the dependency stack as non-negotiable. The full diagnostic is in Chapter 9 of the Make Me Think series. The chapter on cognitive load (L0) that everything else depends on is Chapter 4. The ethical framework that governs both halves is in Chapter 11. If you want to run a quick version right now, the 5-Minute Perception Audit gives you one test per layer. And if you want to see the protocol applied to a live site, Forge runs it automatically. Key Terms Generative Protocol The rule that no solution should be proposed until all five layers have been analyzed and requirements accumulated. The generative counterpart to the Feel, Unpack, Diagnose, Prescribe diagnostic. Rule Zero Do not propose any solution until all five layers have been analyzed and requirements accumulated. Solution-first thinking contaminates the requirement space. Premature closure Croskerry (2009). Locking onto the first measurable finding instead of the actual root cause. The diagnostic error that occurs when you skip from symptom to treatment without the intermediate steps. Intrinsic vs. extraneous load Sweller (1988). Intrinsic load is the complexity of the task itself. Extraneous load is the complexity your design adds on top. Your job is to eliminate every unit of extraneous load so the visitor’s working memory is spent on the task, not the interface. The Gate Pre-ship quality check. Each layer is pass/fail. If any layer below Layer 4 fails, do not ship. Downstream layers cannot compensate for upstream failures. Availability heuristic Tversky & Kahneman (1974). People overweight problems that are mentally accessible and underweight problems that are harder to articulate. Once you propose a solution, it becomes the most available mental object. Solution-first Anti-pattern. Starting with a solution instead of a problem. Every subsequent conversation evaluates the solution rather than exploring the problem space. The design version of premature closure. Competitive copying Anti-pattern. Importing another company’s design decisions without importing their context, constraints, or users. Velocity-first Anti-pattern. Choosing the campaign that wins this quarter over the product work that would win the year. Checklist mode Anti-pattern. Evaluating components instead of perception. You can pass every item and still have a page that feels wrong. Skipping layers Anti-pattern. Applying a Layer 4 fix to a Layer 1 problem. The dependency stack is a chain. You can’t skip links. References Tversky & Kahneman (1974) Judgment under uncertainty: heuristics and biases. Science, 185(4157), 1124–1131. Nisbett & Wilson (1977) Telling more than we can know: verbal reports on mental processes. Psychological Review, 84(3), 231–259. Sweller (1988) Cognitive load during problem solving: effects on learning. Cognitive Science, 12(2), 257–285. Klein (1998) Sources of Power: How People Make Decisions. MIT Press. Berdichevsky & Neuenschwander (1999) Toward an ethics of persuasive technology. Communications of the ACM, 42(5), 51–58. Reber & Schwarz (1999) Effects of perceptual fluency on judgments of truth. Consciousness and Cognition, 8(3), 338–342. Cowan (2001) The magical number 4 in short-term memory: a reconsideration of mental storage capacity. Behavioral and Brain Sciences, 24(1), 87–114. Croskerry (2009) A universal model of diagnostic reasoning. Academic Medicine, 84(8), 1022–1028. Alter & Oppenheimer (2009) Uniting the tribes of fluency to form a metacognitive nation. Personality and Social Psychology Review, 13(3), 219–235. Cowan (2010) The magical mystery four: how is working memory capacity limited, and why? Current Directions in Psychological Science, 19(1), 51–57. A note on how this was written: This post is AI-assisted. I provide the methodology, the case studies, and the editorial direction. AI helps me structure and draft. This is consistent with Perception-First Design’s own transparency principle: if I’m writing about perception, I should be honest about how the writing itself is produced. --- ## Forge: Design Intelligence, Automated URL: https://aurochs.agency/writing/methods/forge-design-intelligence/ Forge didn’t start as a product. It started as a command I typed when I wasn’t sure something was going to work. /analyze That’s it. A skill file in my CLI workflow. I’d be building a landing page, or reviewing a client’s dashboard, or staring at a mockup that felt off but I couldn’t articulate why. So I’d type /analyze and let it run. A few seconds later: a scored report. What’s working, what isn’t, and why, grounded in the five perception layers I’d been using for 15 years of design audits. I never planned to extract it. I just kept using it. Hundreds of times. How I Actually Use It The formal name is Perception-First Design analysis. In practice, I say “PFD this.” It runs on anything. A URL. A screenshot. A mockup file. A directory of HTML and CSS. Raw copy. A plan. An impasse. It doesn’t care about the input format. It cares about the question: will this make sense to a real person, and will it convert? The output is a scored report across five cognitive layers: L0 Layer 0: Cognitive Load Cognitive load. Can the brain even process this without overload? L1 First Impression The 50-millisecond judgment. Pre-attentive pattern recognition and initial trust. L2 Processing Fluency Typography, spacing, and color relationships. Does it feel effortless? L3 Perception Bias Anchoring, framing, social proof. Working for you or against you? L4 Decision Architecture Is the path to conversion clear, or is there friction? Each layer gets a score. Each finding gets a priority. Each prescription is specific enough to act on. But the part that changed how I work isn’t the first pass. It’s the second one. The Convergence Loop One pass gives you a score and a list of findings. Useful. But the real value comes from running it again after you fix things. First pass: 47. Twenty-two findings across four layers. The CTA competes with the nav. The font-size ramp breaks on mobile. The pricing section has three equal-weight options with no anchor. I fix the findings. Then I type /analyze again. Second pass: 68. Fourteen findings. The CTA issue is resolved. New finding: the hero image is fighting the headline for first-fixation. Didn’t notice that before because the CTA problem was louder. Fix. Run again. Third pass: 82. Seven findings, all medium or low priority. The score is climbing because each pass peels back a layer. Fixing the biggest violations reveals subtler ones underneath. The confidence increases because the analysis is looking at a cleaner surface each time. Fourth pass: 91. Three findings, all low. The design is converging. Not perfect. Converging. The gap between passes is narrowing, which is the signal that you’re approaching the ceiling for this particular design. That loop, score, fix, re-score, is the thing I couldn’t stop using. Not because I trust a number blindly. Because each pass forced me to look at the design through a specific cognitive lens, and the score gave me a way to tell if the changes actually moved the needle or just felt like they did. Where It Shows Up I use it at every decision point where the question is “will this work?” Plans. Before I build, I run the plan through PFD. Not the code. The plan itself. Does the information architecture make sense? Is the hierarchy going to create the right first impression? Will the user path feel obvious or will people have to think about where to click? If the plan scores low, building it is a waste. Fix the plan first. Impasses. I’m staring at a layout and something feels wrong. I can’t tell what. I run it. The report says L2 is at 38. The spacing rhythm in the feature grid is irregular, creating a processing fluency violation. That’s the thing I couldn’t name. Now I can fix it. QA checks. Before anything ships, it gets a PFD pass. This is the “did I miss something?” moment. Client dashboards, landing pages, book chapter layouts, pricing tables. If it’s going to be seen by someone, it gets scored. I’ve caught things on the QA pass that I’d been looking at for hours without seeing. Client work. When a client asks “why should we change this?” the report is the answer. Not “because I think so.” Because the L1 first-impression score is 44 and here are the three findings that explain why their bounce rate is what it is. Specific, scored, grounded in published cognitive science. It changed the conversation from opinion to measurement. From a Skill File to an API The skill file was a markdown document. A set of instructions that told the AI how to think like a PFD analyst. It encoded 15 years of design audits into a repeatable process: what to look at, in what order, how to score it, and how to write prescriptions that someone could actually act on. It worked inside my terminal. It couldn’t work anywhere else. The extraction started in early March. I pulled the methodology out of the skill file and into a Cloudflare Worker. Same five-layer analysis, same scoring, same prescription format. But now it was an API. Call it from anywhere. Send a URL, get a report back. The API went live at pfd-api.aurochs.agency. Same day, I pointed my CLI skill at the API instead of running it locally. The experience was identical. The difference: now anyone with an API key could run the same analysis I’d been running for months. From API to Product An API is useful if you already have a workflow to plug it into. Most people don’t. They need a place to go, paste a URL, and see a report. That’s Forge. Web app with auth, session management, and the convergence loop built into the UI. Scan a URL. See the scored report. Fix things. Scan again. Watch the score climb. Same workflow I’ve been running in my terminal, except now it’s a product anyone can use. Three tiers. Solo is free: one analysis a week, three layers. Professional is $39/month: unlimited analyses, all five layers, the convergence loop, API access, MCP integration for Claude Code and ChatGPT. Agency is $129/month: everything plus white-label reports and multi-client vaults. The free tier exists because I want people to feel the first pass. One scored report is usually enough to know whether this is useful to you. If the findings make you nod, you’ll want the convergence loop. If they don’t, it’s not for you. No pitch required. What It’s Not It’s not a replacement for user testing. Scores reflect heuristic analysis of perceptual factors. They’re informed by 55 research citations and 15 years of applied audits, but they’re not eye-tracking data. Pair it with real user testing for definitive product decisions. It’s not going to redesign your site for you. It tells you what’s wrong and why. The prescriptions are specific enough to act on, but acting on them is your job. It’s not vibes. Every finding maps to a perception layer. Every score has a methodology behind it. When the report says L2 is at 38, it means the processing fluency of your layout is creating measurable friction against how the human visual system processes information. That’s not an opinion. That’s a diagnostic. Why I’m Sharing It The honest answer: I built it for myself and it made my work measurably better. Clients got clearer reports. Designs converged faster. I stopped second-guessing layouts because I had a scoring framework that caught things I missed. Keeping it locked in my terminal would be the easy move. Extract it, charge for it, and let people find their own tools. But the methodology, Perception-First Design, is already public. The book chapters explaining the cognitive science are already published. The tool is the methodology applied. If I believe the methodology should be shared, the tool should be too. I use it every day. On every project. On client work, on my own products, on this article. If the question “will this work, and will this make sense to a real person?” matters to you, that’s what Forge answers. --- ## You Wouldn't Vibe Code a Car URL: https://aurochs.agency/writing/notes/you-wouldnt-vibe-code-a-car/ A friend sent me a link last week. "Just watch this." Sixty seconds of someone talking into a camera about a topic I could have read about in eight. I didn’t watch it. I almost never do. Reels, TikToks, YouTube Shorts, Instagram stories. I avoid them the way other people avoid spam calls. They frustrate me in a way I couldn’t articulate for a long time, a low-grade anger that shows up the moment someone sends a video and expects me to stop what I’m doing and hand over my full attention to find out if the payoff is worth it. I finally figured out what the anger is. It’s my brain doing cost-benefit analysis in real time and returning a negative number before I even hit play. The Meme In the early 2000s, the Motion Picture Association ran an anti-piracy ad. You’ve seen it. Dark screen, dramatic narration: "You wouldn’t steal a car. You wouldn’t steal a handbag." The message was dead serious. The internet turned it into a punchline immediately. "You wouldn’t download a car." The remix was better than the original because it exposed the absurdity of the comparison. Digital copying isn’t theft in the way stealing a physical object is. Everyone knew this. The meme made it sayable in five words. Then the world shifted. By the mid-2010s, 3D printing had gotten serious enough that "you wouldn’t 3D print a car" stopped being entirely hypothetical. Local Motors printed one. The joke got uncomfortable because the gap between absurd and plausible had narrowed. Now it’s 2026 and people are building functional prototypes by describing what they want to an AI. "I vibe coded a car" isn’t a meme yet. But I’m sitting here thinking about it, and the thing that makes it work is the same thing that made "download a car" work 20 years ago: the template is pre-loaded. Your brain fills in the pattern before you finish reading. You get it instantly. That’s processing fluency, and it’s the reason I spent 15 years building a design methodology around it. It’s also the reason memes used to spread the way they did. And the reason they don’t anymore. When Memes Were Free Between roughly 2008 and 2014, the internet ran on image macros. Impact font over a photograph. Rage comics drawn in MS Paint. Demotivational posters. Advice animals. The production cost was zero. The comprehension cost was zero. Anyone could make one. Anyone could get one. Richard Dawkins coined "meme" in 1976 to describe a unit of cultural transmission that replicates like a gene. He identified three properties that determine survival: fidelity (does the core idea stay recognizable through variations?), fecundity (how fast and how widely does it replicate?), and longevity (how long does it survive?). Image macros scored high on all three. The template preserved fidelity. Low production cost enabled fecundity. Simplicity ensured longevity. Limor Shifman, who wrote the definitive academic treatment of internet memes in 2014, called them "(post)modern folklore." She was right. They were participatory culture at its purest. The template structure, fixed format plus variable content, meant anyone could contribute without needing production skills, distribution channels, or permission. You needed a browser and an idea. What nobody named at the time, and what I want to name now, is the cognitive architecture that made this work. A static image meme consumes one cognitive channel (visual) for about one to two seconds. Pattern recognition handles most of the processing. The scaffold is pre-loaded from cultural context. Your brain fills one slot, gets the joke, moves on. Total cognitive investment: near zero. Total information transfer: high. That ratio, maximum information per unit of cognitive effort, is what Perception-First Design calls processing fluency. Reber and Schwarz demonstrated in 1999 that the subjective ease of processing directly influences truth judgments: statements that are easier to read feel more true. Song and Schwarz showed in 2008 that hard-to-read instructions make the described task seem harder. The feeling of ease isn’t just aesthetic. It shapes what people believe, what they share, and what they do next. A 2021 UCLA study by Wong and Holyoak tested what actually drives meme sharing. They found a specific chain: relatability leads to aptness, aptness leads to comprehensibility, comprehensibility leads to humor, humor leads to sharing. Comprehensibility, a processing fluency proxy, was the critical mediator. If you don’t get it fast, you don’t share it. Period. Memes were accidentally Perception-First Design. Nobody called it that. Nobody had to. The format was L2-optimal by default. The Cost of Video I am autistic. I process information differently from most people, and I’ve spent most of my adult life figuring out exactly how. One of the things I know about my own cognition is that video demands everything from me. Visual channel. Auditory channel. Temporal tracking. Attention maintenance. All of it, simultaneously, for a duration I can’t control. A meme costs me one to two seconds and one cognitive channel. A 60-second video costs me 60 seconds and every channel I have. The information payload is frequently identical. The cost is an order of magnitude higher. This isn’t just my experience. John Sweller’s cognitive load theory, introduced in 1988 and refined over the following decades, distinguishes between intrinsic load (the complexity of the content itself), extraneous load (the complexity imposed by how the content is presented), and germane load (the effort spent building understanding). Poorly designed media increases extraneous load, the load that has nothing to do with the actual content, and suppresses learning. Sweller, Ayres, and Kalyuga’s 2011 consolidation of the theory identified the transient information effect: information that disappears (audio, video) imposes higher cognitive load than information that persists (text, static images), because you have to hold the vanishing information in working memory while simultaneously processing whatever comes next. Annie Lang demonstrated in 2000 that video automatically commandeers cognitive resources through orienting responses to structural features: cuts, zooms, motion, sound changes. Your brain allocates attention to these features whether or not they carry meaningful content. You don’t choose to pay attention to a jump cut. The medium takes your attention without asking. A Stanford review by Tversky, Morrison, and Betrancourt in 2002 examined dozens of studies comparing static graphics to animations. Their conclusion: animations rarely outperform static representations. When they appeared to, it was because they conveyed more information, not because the format itself was better. When information content was controlled, static either matched or beat animation. The perception that video is a superior format for conveying ideas is, itself, a perception bias. And then there’s the CHI paper. Chiossi and colleagues published a study in 2023 showing that TikTok use significantly degraded prospective memory, the ability to remember to do things you planned to do. Short-form video combined with rapid context-switching overloaded working memory. It wasn’t a survey. It was a controlled experiment. N=60, between-subjects design, published at the top human-computer interaction venue in the field. TikTok literally makes you forget what you were about to do. I feel this. Not as a data point, but as lived experience. The anger I feel when someone sends a video link is my cognitive system signaling that the expected return on investment is negative. I know, before I press play, that the ratio of effort to information is going to be bad. My brain has learned this through years of evidence. "Just Watch This" The phrase itself is the problem. Three words that dismiss the cognitive cost of what they’re asking. When someone sends you a video and says "just watch this," they’ve already watched it. The processing cost is behind them. They know what’s in it, they know the payoff, they’ve already decided it was worth their time. To them, sharing it feels low-cost, like handing you a gift. To you, it’s an unaudited demand on your cognitive budget. You don’t know the payoff. You don’t know the length. You can’t skim it, can’t scan it for the relevant parts, can’t compare it to something else side by side. You have to surrender your attention linearly, from beginning to end, and hope the source is trustworthy. "Just" is doing a lot of work in that sentence, and all of it is wrong. Pirolli and Card published their information foraging theory in 1999, adapting optimal foraging from behavioral ecology to information behavior. The core idea: people navigate information environments to maximize the rate of valuable information gained per unit of time. They introduced the concept of information scent, the perceived value of a source based on proximal cues. A meme has high information scent. You can assess its value in under a second from the thumbnail. A video has low information scent. You literally cannot assess its value without investing time watching it. What "just watch this" really means is: I’m spending your cognitive resources without your permission, and I’m framing it as a favor. This is a design problem. The design principle hiding inside my frustration is the same one that drives every good landing page: frontload the value proposition. Tell me what I’m going to get before I invest. A meme does this. A headline does this. A well-designed hero section does this. Video, by default, does not. The Fraud If memes were working so well, why did they fade? The standard narrative is that platforms evolved, audiences matured, attention spans shortened. None of that is what happened. What happened was that Facebook lied about video. Between 2015 and 2018, Facebook inflated its video viewing metrics by 150 to 900 percent. Not a rounding error. Not a methodology dispute. The original reports estimated 60-80% inflation; the amended class-action complaint revised the figure upward to 150-900%. Facebook settled for $40 million. The inflated numbers told media companies that video was the future. Audiences were supposedly watching more, engaging longer, converting better. The data said so. The data was fabricated. The result was the "pivot to video." CollegeHumor gutted its editorial staff. Funny or Die laid off most of its writers. Mic, a news outlet with tens of millions of monthly readers, fired its entire editorial team and pivoted to video production. Within two years, Mic sold for a reported $5 million after being valued at $100 million. These companies had profitable text and image operations. They destroyed them to chase metrics that were fake. The memes didn’t fail. They were murdered. The format that respected how brains work was abandoned because a platform fabricated evidence that a more expensive, more demanding, more attention-extracting format was performing better. The pivot to video wasn’t a response to audience behavior. It was a response to fraud. Cory Doctorow named the mechanism in 2023: enshittification. Platforms follow a three-stage decay cycle. First, be good to users to attract them. Then, abuse users to benefit business customers (advertisers). Finally, abuse business customers to extract maximum value for shareholders. Video serves stage two. It increases dwell time and ad inventory. It captures more attention per session. It is better for the platform’s business model in every way that is worse for the user’s cognitive experience. In 2025, Ardoline and Lenzo published the first peer-reviewed paper connecting enshittification to cognitive harm. They introduced the concept of cognitive deskilling: when platforms degrade, users who have offloaded cognitive tasks to those platforms lose the capacity to perform those tasks independently. The shift from text and image to algorithmic video feeds doesn’t just change what people consume. It degrades their ability to seek, evaluate, and retain information on their own. Doctorow endorsed the paper. The American Dialect Society had already named "enshittification" its 2023 Word of the Year. The Disinformation Gap Around 2015, the same year the platform pivot to video accelerated, something else was accelerating: disinformation. I don’t think the timing is coincidental. Pennycook and Rand published a study in 2019 that I think about constantly. They tested whether susceptibility to fake news was driven by partisan bias or by lazy thinking. The answer was lazy thinking. People who scored higher on analytical reasoning tests were better at distinguishing fake from real news, regardless of political alignment. The problem wasn’t ideology. The problem was insufficient cognitive engagement. Video, by its nature, is a passive consumption format. You sit. You receive. The medium’s temporal structure demands that you process information at the source’s pace, not your own. You cannot pause to evaluate a claim without losing the thread. You cannot scan ahead to assess whether the argument holds up. You cannot place two videos side by side and compare their claims the way you can with two articles. Sundar’s MAIN model, published in 2008, identified the realism heuristic: the more realistic a medium feels, the more credible people perceive its content to be, regardless of actual accuracy. Video triggers this heuristic more strongly than text or static images. "Seeing is believing" is not a proverb. It is a measurable cognitive bias, and video exploits it by default. Vaccari and Chadwick tested this in 2020 with deepfake political videos. Their finding: deepfakes didn’t primarily mislead viewers directly. What they did was create uncertainty that reduced trust in all news on social media. Chesney and Citron named this the "liar’s dividend" in 2019: once people know video can be faked, even real video becomes suspect, and anyone caught on camera can claim the footage is synthetic. The realism heuristic cuts both ways. A static meme, by contrast, is inherently low-trust. Nobody mistakes an image macro for journalism. The format signals "this is an interpretation, a joke, a take." It invites evaluation rather than demanding belief. And because it processes in under two seconds, the analytical brain isn’t bypassed. You get the claim, you evaluate the claim, you share or don’t. The entire cycle happens fast enough that System 2, Kahneman’s slow, deliberative reasoning system, doesn’t have time to disengage. Video is long enough for System 2 to go to sleep. And when System 2 sleeps, you believe what you’re shown. Full Disclosure I should tell you something about this essay. I vibe coded it. I also vibe coded the website you’re reading it on. My design methodology, extracted into an API and shipped as a SaaS. Cognograph, from a thought into a full web app with four provisional patents, a design system, and its own site. A 12-part book series that picks up where one of my favorites left off. Client dashboards. Client reporting. Client websites for multiple businesses. The list keeps growing. Two months after figuring out the scaffolding Claude Code needed to let me focus on the input and output only. The methodology itself I trademarked. I synthesized it from notebooks I’d been filling for over a decade, OCR’d and fed to Claude as source material. The AI didn’t invent PFD. It helped me extract and formalize what was already there. I didn’t do this by accident, and I didn’t do it by being a better coder than the machines. I have 15 years of full-stack design and development experience. What that taught me wasn’t how to write better code. It taught me what to ask for and how to know when the output is right. The skill didn’t stay in execution. It moved to quality control on both sides of it. Specification quality: knowing exactly what good looks like before anything gets built. Evaluation quality: recognizing whether the output meets that standard after it’s rendered. If you can hold both ends, the middle can be produced by anything. If you can’t, it doesn’t matter what produces it. The output will be garbage and you won’t know why. This is exactly what Perception-First Design does for the design process itself. PFD is the specification layer: perception science says this configuration should produce this cognitive response. PFD is also the evaluation layer: measure whether it did. The design in the middle, the pixels, the layout, the copy, can be produced by a human, an AI, a team of twelve, or one person with a terminal window. The quality lives in the specification and the vetting, not in who or what renders it. Vibe coding with bad specs and no vetting produces garbage. Vibe coding with enterprise-grade specification and enterprise-grade evaluation produces enterprise-grade output. The people panicking about AI replacing designers are worried about the wrong layer. The rendering was never the hard part. Knowing what to render and knowing whether you got it right: that’s the whole game. The numbers back this up. Industry analysis from Veracode and Tenzai found that roughly 45% of AI-generated code contains security vulnerabilities, with only 10.5% of functionally correct solutions being truly secure. The machine produces plausible output. Whether that output is safe to ship depends entirely on whether someone with domain knowledge is checking. A well-designed study by METR (n=16, so take this as a provocation rather than a conclusion) found experienced developers were 19% slower with AI tools but believed they were 20% faster. That is a 40-point perception gap. The tool felt productive. The measurement said otherwise. Anthropic’s own research found AI-assisted developers scored 17% lower on unassisted follow-up tasks. But the study also identified six distinct interaction patterns: three that preserved skill development, three that eroded it. The difference was not whether you use AI. It was how. I am, as it turns out, the meme. "You wouldn’t vibe code a car." I would. I did. Several of them. The Design Principle I started this essay angry about a video link. I’m ending it with a claim I can defend: the internet found a communication format that was accidentally optimized for how human cognition works, used it for roughly five years, then abandoned it in favor of a format that is optimized for attention extraction at the expense of comprehension, critical thinking, and cognitive autonomy. The image macro was L2-native. It operated at the layer of processing fluency where ideas transmit with minimal friction. It respected the brain’s limited capacity by demanding almost nothing from it. It was self-paced, instantly evaluable, and structurally honest about what it was. Video-first content is L2-hostile. It demands sustained multi-channel attention for uncertain payoff. It triggers automatic resource allocation through structural features that carry no information. It creates the feeling of engagement without the substance of comprehension. And it powers a platform economy that benefits from your attention regardless of whether you benefit from spending it. Perception-First Design is built on a specific commitment: design for how people actually perceive and process the world, not for how you wish they did and not for how your business model needs them to. The meme got this right by accident. The platforms got it wrong on purpose. So: you wouldn’t vibe code a car. But you might build one if the instructions cost you two seconds and one cognitive channel instead of sixty seconds and all of them. The format that respects the brain wins. It always has. The question is whether the platforms will let it. Key Terms Processing fluency The subjective ease of cognitive processing. Information that feels easy to process is perceived as more true, more beautiful, and more trustworthy. The foundational mechanism of PFD Layer 2. Information scent The perceived value of an information source based on proximal cues, before full engagement. Memes have high information scent (assessable in under one second). Video has low information scent (requires time investment to evaluate). From Pirolli & Card’s information foraging theory. Transient information effect The cognitive load penalty imposed by information that disappears (audio, video) versus information that persists (text, images). Vanishing information must be held in working memory while new information arrives, increasing extraneous load. From Sweller’s cognitive load theory. Cognitive deskilling The atrophy of cognitive capacities when platforms degrade. Users who offload seeking, evaluating, and retaining information to platforms lose the ability to perform those tasks independently when the platform changes. From Ardoline & Lenzo (2025). Realism heuristic The tendency to perceive content in more realistic media formats as more credible, regardless of actual accuracy. Video triggers this heuristic more strongly than text or static images. From Sundar’s MAIN model. References Ardoline & Lenzo (2025) The Cognitive and Moral Harms of Platform Decay. Ethics and Information Technology, 27, 37. Anthropic (2025) The Effects of Generative AI on High-Skilled Work: Evidence from Experienced Software Engineers. Anthropic Research. Chesney & Citron (2019) Deep Fakes: A Looming Challenge for Privacy, Democracy, and National Security. California Law Review, 107, 1753-1820. Chiossi et al. (2023) Short-Form Videos Degrade Our Capacity to Retain Intentions. Proceedings of CHI ’23, ACM. Dawkins (1976) The Selfish Gene. Oxford University Press. Doctorow (2023) TikTok’s Enshittification. Wired / Pluralistic. Kahneman (2011) Thinking, Fast and Slow. Farrar, Straus and Giroux. Lang (2000) The Limited Capacity Model of Mediated Message Processing. Journal of Communication, 50(1), 46-70. METR (2025) Measuring the Impact of Early Exposure to AI Coding Tools on Developer Productivity. arXiv:2507.09089. Pennycook & Rand (2019) Lazy, Not Biased: Susceptibility to Partisan Fake News Is Better Explained by Lack of Reasoning. Cognition, 188, 39-50. Pirolli & Card (1999) Information Foraging. Psychological Review, 106(4), 643-675. Reber & Schwarz (1999) Effects of Perceptual Fluency on Judgments of Truth. Consciousness and Cognition, 8, 338-342. Shifman (2014) Memes in Digital Culture. MIT Press. Song & Schwarz (2008) If It’s Hard to Read, It’s Hard to Do. Psychological Science, 19(10), 986-988. Sundar (2008) The MAIN Model. In Metzger & Flanagin (Eds.), Digital Media, Youth, and Credibility. MIT Press. Sweller (1988) Cognitive Load During Problem Solving. Cognitive Science, 12(2), 257-285. Tversky et al. (2002) Animation: Can It Facilitate? International Journal of Human-Computer Studies, 57(4), 247-262. Vaccari & Chadwick (2020) Deepfakes and Disinformation. Social Media + Society, 6(1). Veracode; Tenzai Industry analyses of AI-generated code security. Veracode State of Software Security report; Tenzai security assessment of LLM-generated code. Wong & Holyoak (2021) Cognitive and Motivational Factors Driving Sharing of Internet Memes. Memory & Cognition, 49, 863-872. A note on how this was written: This essay is AI-assisted. I provide the thesis, the personal experience, the editorial direction, and the framework connections. AI helps me structure, draft, and locate academic citations. This is consistent with Perception-First Design’s own transparency principle: if I’m writing about perception, I should be honest about how the writing itself is produced. --- ## The Three-Terminal Problem URL: https://aurochs.agency/writing/notes/the-three-terminal-problem/ Last Tuesday I had seven parallel AI sessions running at once. Three were building HTML pages for different chapters of a book. One was deploying SaaS infrastructure. One was running a design analysis on a client dashboard. The last two were scaffolding components for a landing page. I was switching between them, reviewing outputs, approving file writes, occasionally telling one to wait while another finished something it depended on. A year ago I was typing questions into a chat window in a browser tab. Nobody plans for this. You get there one step at a time, each step driven by a specific limitation you hit in the previous one. And somewhere around step three, you realize the limitation isn’t the AI. It’s the surface area between you and the AI. The sessions can’t see each other. The context you built in one conversation evaporates when you open the next. You become the router, the clipboard, the only entity in the room that knows what all the other entities are doing. This is a field report from that problem. It’s also, indirectly, a response to Steve Yegge, who’s been writing about this frontier from the observation side. I found his writing in late January, three weeks into building something I couldn’t quite explain to anyone yet. More on that later. The Progression If you’ve used AI for work at all, you started somewhere familiar. A chat window. You type a question, get an answer, maybe ask a follow-up. When you need help with a different project, you open a new conversation. Two or three chats going at once, tops. It works until you notice how much time you spend re-explaining things. Every conversation is its own universe. The chat helping you with your website has no idea about the design decisions you discussed in the other chat. You copy-paste between them. You re-describe your project. You answer the same setup questions over and over. The limitation isn’t intelligence. The AI is plenty smart. The limitation is amnesia. Desktop apps improve the organization. Now you have dozens of conversations sorted by project, you can find old threads, the interface is faster. But the amnesia stays. Each conversation starts from zero. You spend the first five minutes of every session re-loading context before you get to the actual work. Then something shifts. Tools like Claude Code give the AI direct access to your project files, your directory structure, your config. Instead of you pasting code into the chat, the AI reads it itself. Instead of you manually applying changes, the AI writes files with your approval. A task that used to take four round trips now takes one. There’s a second shift that’s easy to miss. When the AI can read your whole project, you can write a file that describes your architecture, your conventions, your design tokens, your voice rules. Every session reads it automatically on startup. The amnesia problem mostly goes away, replaced by a document that encodes your project’s institutional knowledge. Most people who try this stay here. One session, occasionally starting a new one when the old one gets too long. It’s a real upgrade. For a lot of workflows, it’s enough. Two Sessions At some point you realize you can run two sessions at once. Same project, same context document. Frontend in one, backend in the other. Or code in one, tests in the other. They don’t need to coordinate much. You glance at each periodically, review the output, course-correct if something went sideways. Two is manageable because your attention can handle the switching cost. It feels like a normal multitasking pattern, similar to having two documents open side by side. I ran two sessions for about a week before I needed a third. Three Changes Everything Two sessions is multitasking. Three is orchestration. The difference isn’t arithmetic. With two, you can hold both threads in your head. With three, you can’t. You lose track of what Session C is doing while you’re reviewing Session A’s output. When you switch back, you need to re-read its recent activity to remember where it is. If Session B produced something that C needs, you have to notice that dependency yourself and intervene manually. Three sessions without infrastructure is chaos. The failure mode is predictable: one overwrites something another just wrote, or two duplicate effort because neither knows what the other already built, or you forget to check one for twenty minutes and it goes in a direction you would have caught immediately. So I built the infrastructure. The Duct Tape Shared context. Every session reads the same instruction file on startup. Mine is 296 lines. Directory structure, routing tables, design tokens, voice rules, persona definitions, memory protocol. When I open Session 4, it already knows everything Sessions 1 through 3 know about the project. Session logs written to disk continuously. Each session writes progress notes to a shared directory as it works. Markdown files, written as checkpoints, not metrics. If a session crashes, or if I need to spin up a replacement, the new one reads the log and picks up where the old one stopped. Over six days of heavy building, I accumulated 41 session files. 41 moments where the state of a work stream got saved to disk instead of evaporating. Issue tracking across sessions. I built a lightweight CLI issue tracker for exactly this. A JSONL file and a handful of commands: create, list, close, sync. Any session can create an issue. Any other session can pick it up. Dependencies are tracked, so I can mark that Task 7 depends on Task 3, and the session working on 7 will know to wait. Reusable skills. Instruction sets stored as files that any session can invoke. /commit follows the same format everywhere. /llm-scrub checks text for AI writing patterns. /analyze runs the same design analysis everywhere. The sessions don’t just share context. They share muscle memory. Parallel dispatch. From a single session, I can launch multiple agents that run independently. Seven chapters of a book, built simultaneously by seven parallel agents. Four dashboard pages scaffolded at once. One orchestrator, many workers. All of it works. I shipped real products this way. 85 commits in 5 days across multiple repos. A client dashboard and SaaS infrastructure built simultaneously in adjacent sessions. Seven book chapters in 40 minutes instead of two hours. But I want to be honest about what this is. It’s duct tape. Sophisticated duct tape. Every piece of it exists because the sessions are blind to each other, and I’m the only one who can see the whole picture. The shared context file, the session logs, the issue tracker, the skills: they’re all workarounds for a fundamental problem. The AI can’t see its own peers. So I have to be the bridge. A Typical Day Tuesday, March 4th. Reconstructed from the session logs. 9 AM. Three sessions open. Session A is building the Forge API, a design intelligence service that runs automated analysis on websites. Session B is building the web app frontend. Session C is running a Perception-First Design analysis on a client’s dashboard, generating a report with specific fixes. Session A finishes the API endpoint. I review the output, approve two file writes, reject one that used the wrong error handling pattern. Session B needs the API contract, so I paste the route signature into B. Three seconds of manual coordination. 11 AM. Session C finishes the analysis report. I make two corrections and tell C to regenerate. While it reruns, I open Session D and point it at book chapters. I dispatch seven parallel agents from D, one per chapter, each building an HTML page from a markdown source. D tracks their progress. 1 PM. The seven chapter pages are done. I review each one, flag two for minor fixes. Sessions A and B are still going. I close C (the report shipped) and open a new session for client dashboard deployment. Ten commits over the next two hours while the other sessions keep building. The pattern: start streams, review outputs, coordinate dependencies by hand when they come up. The human role is quality control and sequencing. The sessions do the building. The Signal I found Steve Yegge’s writing in late January, about three weeks into building something I couldn’t quite explain to anyone. He was describing the same ceiling. What happens when AI capability outpaces what a single session can handle. The bottleneck shifting from “can the AI do this?” to “can I orchestrate enough of them?” He was mapping the terrain from the observation side. I was already in it from the construction side, standing in a half-built prototype, covered in sawdust, trying to wire spatial context injection into a graph traversal engine at 2 AM. Reading someone articulate the exact problem you’re neck-deep in solving does a specific thing. Not “I’m right.” More like: the problem is real, and it’s not just my ADHD brain chasing another project. Someone with Steve’s vantage point sees it too. The surplus capacity is real. The orchestration gap is real. And it’s bigger than any one person’s workflow. That was the signal I needed to stay on the track. Solving It from the Other End Everything I described in the Duct Tape section is patching the problem from the terminal side. Shared files, session logs, issue trackers: they help sessions coordinate, but the sessions are still blind. You’re still the only one who sees the whole picture. I built Cognograph to solve it from the other end. Instead of teaching sessions about each other through shared files on disk, Cognograph puts everything on a spatial canvas. Nodes are your project pieces: conversations, documents, code artifacts, tasks, agent orchestrators. Edges are the connections between them. When you invoke AI on any node, it traverses the graph and injects the relevant context automatically. Your layout is your prompt engineering. The canvas IS the orchestration layer. The three-terminal problem disappears because the terminals aren’t isolated anymore. They’re nodes on a canvas with your whole project on it, and every AI interaction already knows what it’s connected to. I have ASD and severe combined-type ADHD. Files in folders don’t work for me. Linear chat doesn’t work for me. I need to see everything spatially, with connections I can trace with my eyes. Cognograph started as a tool I needed. It turned into a product because the problem isn’t just mine. Anyone running more than two parallel AI sessions hits the same wall. I just hit it earlier because my brain demands spatial context that most tools don’t provide. Four provisional patents filed with the USPTO. The spatial trigger mechanisms, context injection through graph traversal, depth-based semantic zoom, conditional node activation. It isn’t a wrapper on ChatGPT. It’s a different model for how humans and AI share context. 596 commits. Nine weeks from first prototype to full SaaS infrastructure. The build log is public. The Tool I Use Every Day One more thing came out of this. In the Typical Day section, Session C was running a design analysis. That’s Forge: 15 years of design audits turned into a scoring tool. Five cognitive layers from Perception-First Design. Scored analysis, specific prescriptions, round-over-round convergence tracking. I use it on every project I touch. It’s live and free to try. Same tool, same scoring, same methodology. If the question “will this design work, and will it make sense to real people?” matters to you, that’s what Forge answers. Who This Is For If you’re using AI through a chat window and it’s working, none of this matters. Plenty of people get real value from a single conversation. The progression I described isn’t a ladder where higher is better. It’s a response to specific constraints that not everyone hits. You hit them when your projects get big enough that context loss becomes your primary bottleneck. When you’re spending more time re-explaining your codebase than doing actual work. When you have three things that could happen in parallel, and forcing them to be sequential feels like a waste of hours you don’t have. There are two paths through it. The infrastructure path: shared context files, session logs, issue trackers, reusable skills. It works. I shipped products this way. The spatial path: put the whole project on a canvas and let the AI see the connections itself. That’s Cognograph. I got here because I’m a solo operator building multiple products with clients who need things yesterday. The multi-session workflow isn’t a flex. It’s a consequence of being one person with the workload of a small team. Steve, if you’re reading this: thanks for the signal. The problem you keep describing is real. I’m building through it. --- ## Cognograph: Finding the First 100 URL: https://aurochs.agency/writing/notes/cognograph-first-100/ I started building Cognograph on January 7, 2026. Twelve-ish weeks and 638 commits later, it’s a spatial AI canvas for orchestrating web design projects. Open source on GitHub, live on the web, running on desktop. Four patents filed, 2,445 tests passing. I’m looking for the first 100 people to use it. Why It Exists I have ASD and severe combined-type ADHD. Files in folders don't work for me. Linear chat doesn't work for me. I need to see everything spatially, with connections I can trace with my eyes. The trigger was building this site. I was running 2-6 Claude Code CLI sessions at once, each one unaware of the others, held together with invisible scaffolding and manual context injection. Copy the design tokens into this session. Paste the brand guidelines into that one. Remind the third about the component you just built in the first. It worked, technically. It also felt like holding six conversations where nobody could hear anyone else. Even after optimizing the workflow with skills, plugins, MCP layers, and programming Claude itself, the invisible infrastructure kept growing. Long sessions needed planning docs, implementation docs, persona loops, memory state files. An entire workspace of scaffolding that lived outside the tool, invisible to me, and hard to keep track of. I eventually got the hang of it, but that's the point: most people are not going to open a terminal, cd into a directory, and type claude to start a session. Not until they need more than one or two windows open at once, and by then they're already drowning in the same context collapse I was. It's a broader problem than my workflow. Every AI tool suffers from it: you're building a 15-component project, but the AI can only see the one file you're editing. It doesn't know about your brand guidelines, your design tokens, or the five other components that need to stay consistent. 47 browser tabs. Context lost between conversations. Redoing work because you couldn't find where you'd already done it. Cognograph solves that by making context spatial. You arrange your project on a canvas, draw connections between related pieces, and when you invoke AI on any node, it traverses the graph and injects the relevant context automatically. Your layout is your prompt engineering. Where It Lives The web version is live at cognograph.app/workspace. The desktop app is open source at github.com/skovalik/cognograph. Electron, builds on Windows, macOS, and Linux. Four provisional patents filed. Twelve-ish Weeks Twelve-ish weeks of building, when you can’t stop thinking about the thing you’re building: JAN 7 Day zero First prototype. Electron, React Flow, four node types. A canvas where you could talk to AI and see your conversations as objects in space. The idea was simple: what if your workspace was spatial, and the AI could see all of it? JAN 13 The vision crystallizes Wrote the VISION doc and the long-term north star: a 3D spatial desktop inspired by the film Her. Cognograph as the 2.5D stepping stone. Context injection spec'd out: connected nodes automatically feed into AI prompts. The canvas doesn't just organize your work, it IS the prompt. JAN 15 The rebuild Scrapped the prototype and started clean. v1.0 committed with the architecture I actually wanted: proper state management, typed data layer, graph traversal for context injection. The first prototype taught me what to build. This was building it. JAN 16 36 commits in one day Property system. Edge system with context settings. Artifact system with drag-and-drop. Template system. AI workspace editor spec. The context injection engine, the thing that makes the whole product work, went from spec to implementation in a single session. JAN 19–25 Agent mode, MCP server, first daily driver Embedded agent mode so the AI can act autonomously. MCP server so external tools can talk to the canvas. Resizable nodes, theme persistence, light mode. Hit v1.4.8 and the note in the commit log says “working iterative version.” I started using it for real work. v1.4.8 Night A real workspace: Research and Copywriting Workflow with context edges FEB 1 The UX sprint 30+ commits. Command palette. Focus Mode (reduced stimuli for neurodivergent users). Visual bookmarks. Quick node creation shortcuts. Task completion celebrations. After 350 iterations of UX evaluation with AI personas, the interface started feeling like it belonged to someone. Not just functional. Opinionated. FEB 7 20 ambient canvas effects Starfield, fireflies, rain, aurora, flow fields, caustics, circuit traces. Not decoration. Published research supports ambient motion as an accessibility feature for ADHD (PMC 2023). The canvas feels alive, and that's the point. FEB 10 Four patents filed USPTO provisional applications filed. Four patent families. The kind of day where you realize what you’re building might actually be defensible. FEB 11 The mega-plan day 60+ commits across 15 parallel build streams. Agent presets, cost visibility, live preview sandbox, Claude Code bridge, Electron packaging for Win/Mac/Linux, orchestrator node, site architecture nodes, plugin system. The single most productive day of the project. FEB 14 The science validates it 19,000-word cognitive science literature review. 40+ peer-reviewed citations (O'Keefe, Cowan, Treisman, Norman, Shneiderman). This is also where the scientific foundation for Perception-First Design got formalized: the same research on spatial cognition, working memory limits, and processing fluency that underpins the methodology. Key finding: ~2x spatial recall advantage is real. Honest assessment: 2-5x realistic efficiency gain, not the speculative "10-1000x" you see in AI tool marketing. FEB 24 Rich node depth system The biggest single feature push. 5-level semantic zoom (far overview to fully expanded). Embedded terminal with PTY lifecycle management. Z-key zoom overlay. Cluster detection. Execution status with depth-of-field blur. FEB 27 Web version goes live Complete web adapter layer with local persistence, streaming, and a proxy for LLM calls. Same app, different runtime. Landing page with embedded demo canvas. cognograph.app and canvas.cognograph.app, both live. 53 days from first prototype to web launch. 53 days to web launch FEB 28 The pivot Went from "general-purpose AI canvas" to "AI-powered web design orchestration." The general-purpose version worked, but the market positioning was a shrug. Web design is the wedge: I've been doing it for 15 years, and every AI design tool I've tried suffers from the same context collapse problem I built Cognograph to solve. MAR 1 Visual identity locks in Signal White + Gold Glow accent system. Space Grotesk + Space Mono typography. Content-first node design, no glassmorphism. Stripped the chrome down to a single thin icon toolbar. MAR 4 The hero ships Zoom-out demo hero on the landing page: three embedded CLI terminals typing in real time, camera pulls back to reveal a 13-node spatial graph with gradient edges. 35-node read-only demo workspace. The kind of first impression you can't fake with a screenshot. MAR 5–6 Dashboard, inspector, cloud design Workspace foyer with template cards. Node-anchored quick inspector with property controls. Cloud workspace architecture designed (bring your own keys, multi-model). Creative pipeline spec for media generation inside nodes. Night Nine node types, outline panel, and the properties inspector MAR 6 SaaS feature-complete Auth, billing, workspace sync, API key management, usage metering, orchestration server, and media generation pipeline. Landing page redesigned with CSS-only hero animation under 20KB. MAR 7 SaaS infrastructure deployed cognograph.app, api.cognograph.app, media.cognograph.app. Auth, billing, orchestration, and media storage all wired together. The full production stack, live and handling requests. MAR 8–9 Design system audit + 41 mockups Built an interactive style guide with implementation status badges on every section: what exists, what’s partial, what’s planned, what’s drifted from spec. Generated 41 screen mockups for planned features: artboard mode, level-of-detail zoom, onboarding sequence, spatial keyboard navigation. Designed the landing page through 9 rounds of iteration. MAR 10 Design system shipped, SaaS backend complete Level-of-detail universalized across all node types. Living Grid effect. Artboard mode for expanded node editing. Onboarding flow with template gallery. Spatial keyboard navigation with directional guide lines. Then server-side: quota enforcement, billing lifecycle management, and a transactional email system. 1,462 tests passing MAR 11 OAuth, email infrastructure, pre-launch GitHub and Google OAuth. Email infrastructure verified and branded. Billing webhooks covering the full subscription lifecycle. Pre-launch checklist written. 596 commits · 1,462 tests · patent-pending MAR 12–25 Mobile-first redesign Touch-native toolbar from scratch. Proximity-based context connections: drag a node near another and they auto-link. Hamburger drawer, long-press menus, half-sheet properties panel. The desktop app was built for a mouse. Rebuilding for touch meant rethinking every interaction. Mobile welcome screen with touch-native toolbar MAR 25 Launch readiness 218 out of 230 dogfood flows passing. All four patents verified as implemented. 26 bug fixes from a 47-agent codebase sweep. The moment where you stop adding features and start trusting what’s there. APR 4–5 Security hardening + billing Stripe Pro subscription wired from signup to invoice. Wave 0 security hardening: RLS policies locked, CSP hardened, rate limiting, credit abuse prevention, encryption keys rotated. Biome lint auto-fixed 818 files. The work nobody sees that makes a product shippable. 2,427 tests passing APR 6 Public launch Pushed to GitHub. 762 files. Full sanitization pipeline before the push. 20 review passes plus a dual-agent scan to make sure nothing patent-sensitive or secret made it through. README rewritten through four rounds of iteration. cognograph.app homepage rebuilt. The full Electron app, not just a web demo, available as a public repo for the first time. open source Brand guide generation: conversation nodes feeding artifact previews rendered live on canvas APR 6–8 Post-launch hardening MCP-to-renderer bridge shipped. External tools now talk directly to the canvas state, no more split-brain sync. Artboard mode removed: 6,500 lines of glassmorphic overlay that didn’t earn its complexity. SEO wave 2: FAQ component, structured data, llms.txt rewrite. Terminal context regression patched. APR 8 · NOW Still looking for the same hundred people 638 commits. 2,445 tests. Four patents pending. Public repo, web app, and desktop app all live. The difference from March: now they can clone the repo and try it themselves. 638 commits · 2,445 tests · patent-pending NEXT The first 100 The repo is public, the web app runs in any browser, and the desktop app builds for all three platforms. The product works. Now it needs people building websites who want to try spatial AI orchestration instead of juggling forty-seven tabs. What Happens Next When I wrote the first version of this post in early March, the “Next” section listed onboarding, billing, and cloud sync as things I still needed to build. They shipped within a week. Since then: a full security audit, a public GitHub launch, a mobile-first redesign, and an MCP bridge that lets external tools talk directly to the canvas. 638 commits. 2,445 tests. Four patents. Open source. Web and desktop, both live. I’m still looking for the same hundred people. If you build websites professionally and want to try a spatial approach to working with AI, the canvas is at cognograph.app/workspace. The repo is at github.com/skovalik/cognograph. I’m reachable at stefan@aurochs.agency. --- ## From Skill to Product URL: https://aurochs.agency/writing/notes/skill-to-product/ Forge didn’t start as a product. It started as a command I typed when I wasn’t sure something was going to work. /analyze That’s it. A skill file in my CLI workflow. I’d be building a landing page, or reviewing a client’s dashboard, or staring at a mockup that felt off but I couldn’t articulate why. So I’d type /analyze and let it run. A few seconds later: a scored report. What’s working, what isn’t, and why, grounded in the five perception layers I’d been using for 15 years of design audits. I never planned to extract it. I just kept using it. Hundreds of times. How I Actually Use It The formal name is Perception-First Design analysis. In practice, I say “PFD this.” It runs on anything. A URL. A screenshot. A mockup file. A directory of HTML and CSS. Raw copy. A plan. An impasse. It doesn’t care about the input format. It cares about the question: will this make sense to a real person, and will it convert? The output is a scored report across five cognitive layers: L0 Layer 0: Cognitive Load Cognitive load. Can the brain even process this without overload? L1 First Impression The 50-millisecond judgment. Pre-attentive pattern recognition and initial trust. L2 Processing Fluency Typography, spacing, and color relationships. Does it feel effortless? L3 Perception Bias Anchoring, framing, social proof. Working for you or against you? L4 Decision Architecture Is the path to conversion clear, or is there friction? Each layer gets a score. Each finding gets a priority. Each prescription is specific enough to act on. But the part that changed how I work isn’t the first pass. It’s the second one. The Convergence Loop One pass gives you a score and a list of findings. Useful. But the real value comes from running it again after you fix things. First pass: 47. Twenty-two findings across four layers. The CTA competes with the nav. The font-size ramp breaks on mobile. The pricing section has three equal-weight options with no anchor. I fix the findings. Then I type /analyze again. Second pass: 68. Fourteen findings. The CTA issue is resolved. New finding: the hero image is fighting the headline for first-fixation. Didn’t notice that before because the CTA problem was louder. Fix. Run again. Third pass: 82. Seven findings, all medium or low priority. The score is climbing because each pass peels back a layer. Fixing the biggest violations reveals subtler ones underneath. The confidence increases because the analysis is looking at a cleaner surface each time. Fourth pass: 91. Three findings, all low. The design is converging. Not perfect. Converging. The gap between passes is narrowing, which is the signal that you’re approaching the ceiling for this particular design. That loop, score, fix, re-score, is the thing I couldn’t stop using. Not because I trust a number blindly. Because each pass forced me to look at the design through a specific cognitive lens, and the score gave me a way to tell if the changes actually moved the needle or just felt like they did. Where It Shows Up I use it at every decision point where the question is “will this work?” Plans. Before I build, I run the plan through PFD. Not the code. The plan itself. Does the information architecture make sense? Is the hierarchy going to create the right first impression? Will the user path feel obvious or will people have to think about where to click? If the plan scores low, building it is a waste. Fix the plan first. Impasses. I’m staring at a layout and something feels wrong. I can’t tell what. I run it. The report says L2 is at 38. The spacing rhythm in the feature grid is irregular, creating a processing fluency violation. That’s the thing I couldn’t name. Now I can fix it. QA checks. Before anything ships, it gets a PFD pass. This is the “did I miss something?” moment. Client dashboards, landing pages, book chapter layouts, pricing tables. If it’s going to be seen by someone, it gets scored. I’ve caught things on the QA pass that I’d been looking at for hours without seeing. Client work. When a client asks “why should we change this?” the report is the answer. Not “because I think so.” Because the L1 first-impression score is 44 and here are the three findings that explain why their bounce rate is what it is. Specific, scored, grounded in published cognitive science. It changed the conversation from opinion to measurement. From a Skill File to an API The skill file was a markdown document. A set of instructions that told the AI how to think like a PFD analyst. It encoded 15 years of design audits into a repeatable process: what to look at, in what order, how to score it, and how to write prescriptions that someone could actually act on. It worked inside my terminal. It couldn’t work anywhere else. The extraction started in early March. I pulled the methodology out of the skill file and into a Cloudflare Worker. Same five-layer analysis, same scoring, same prescription format. But now it was an API. Call it from anywhere. Send a URL, get a report back. The API went live at pfd-api.aurochs.agency. Same day, I pointed my CLI skill at the API instead of running it locally. The experience was identical. The difference: now anyone with an API key could run the same analysis I’d been running for months. From API to Product An API is useful if you already have a workflow to plug it into. Most people don’t. They need a place to go, paste a URL, and see a report. That’s Forge. Web app with auth, session management, and the convergence loop built into the UI. Scan a URL. See the scored report. Fix things. Scan again. Watch the score climb. Same workflow I’ve been running in my terminal, except now it’s a product anyone can use. Three tiers. Solo is free: one analysis a week, three layers. Professional is $39/month: unlimited analyses, all five layers, the convergence loop, API access, MCP integration for Claude Code and ChatGPT. Agency is $129/month: everything plus white-label reports and multi-client vaults. The free tier exists because I want people to feel the first pass. One scored report is usually enough to know whether this is useful to you. If the findings make you nod, you’ll want the convergence loop. If they don’t, it’s not for you. No pitch required. What It’s Not It’s not a replacement for user testing. Scores reflect heuristic analysis of perceptual factors. They’re informed by 55 research citations and 15 years of applied audits, but they’re not eye-tracking data. Pair it with real user testing for definitive product decisions. It’s not going to redesign your site for you. It tells you what’s wrong and why. The prescriptions are specific enough to act on, but acting on them is your job. It’s not vibes. Every finding maps to a perception layer. Every score has a methodology behind it. When the report says L2 is at 38, it means the processing fluency of your layout is creating measurable friction against how the human visual system processes information. That’s not an opinion. That’s a diagnostic. Why I’m Sharing It The honest answer: I built it for myself and it made my work measurably better. Clients got clearer reports. Designs converged faster. I stopped second-guessing layouts because I had a scoring framework that caught things I missed. Keeping it locked in my terminal would be the easy move. Extract it, charge for it, and let people find their own tools. But the methodology, Perception-First Design, is already public. The book chapters explaining the cognitive science are already published. The tool is the methodology applied. If I believe the methodology should be shared, the tool should be too. I use it every day. On every project. On client work, on my own products, on this article. If the question “will this work, and will this make sense to a real person?” matters to you, that’s what Forge answers.