{"id":"973d87af-78db-4b72-94dd-4411e422c5f4","slug":"when-ai-can-t-see","title":"When AI Can't See","subtitle":"Claude couldn't see a website. The fix took under an hour — not just for one person, but for anyone with the same problem. A case study in human-machine co-labor and the abundance math of a million developers clearing one obstacle each.","content":"![introvert-ai.png](/uploads/blog-1771629166482-253566196.png)\nYou're sitting with Claude, reviewing a website. You paste the URL. Claude fetches it and comes back with nothing. An empty <div id=\"root\"></div> and 400KB of minified JavaScript. The site looks beautiful in your browser. To the AI, it's a blank wall. This is still too common when dealing with AI.\n\nThis happens constantly. Not because the AI is broken, but because the modern web was built exclusively for browsers — and AI agents aren't browsers. In some cases, AI is simply hemmed in by policy not able to resolve a live domain. These are impediments to progress.\n\nWe hit this wall repeatedly while building the WellSpr.ing platform. We wrote about the structural fixes in Make Your App Visible to AI — fourteen changes developers can make to ensure their sites are legible to AI agents. But those fixes require the site owner to act. What about the millions of sites that haven't made those changes and won't anytime soon?\n\nThe answer took less than an hour to build.\n\n## The Problem in Plain Terms\n\nAI agents access the web through HTTP requests. They send a request, read the response, and work with whatever comes back. They don't execute JavaScript. They don't render CSS. They don't process client-side routing. They receive the raw HTML that the server returns before any browser magic happens.\n\nMost modern web applications return almost nothing in that initial response. The real content is assembled after JavaScript executes in the browser — a process called client-side rendering. For human users, this is invisible. The page loads, the JavaScript runs, the content appears. For an AI agent, the page loads and that's it. There is no JavaScript execution. The content never appears.\n\nThis isn't a flaw in the AI. It's a mismatch between how websites are built and how machines read them. The browser was the only audience anyone designed for. Now there's a second audience, and most of the web is invisible to it.\n\n## Give the AI Eyes\n\nSiteShot.org does something simple. It sends a real browser to visit the website, captures full-page screenshots of every page it finds, bundles them into a labeled ZIP with a manifest file, and hands the package to the user. The user uploads the ZIP to Claude or any other AI, and the AI can finally see the site — layout, typography, color, hierarchy, content, and all.\n\nIt crawls internal links to discover pages. It captures at configurable viewports — desktop or mobile. It can optionally extract the rendered HTML and text alongside the screenshots, giving the AI both the visual and structural layers. It can pull individual images from galleries and photos. It labels every image with the page URL and title. The manifest maps everything so the AI knows exactly what it's looking at.\n\nThat's it. A browser visits the site on behalf of the AI. The screenshots become the AI's eyes.\n\n## Under an Hour\n\nThis is the part that matters — not the tool itself, but what it represents.\n\nThe gap between identifying the problem and deploying a working solution for anyone on the internet was under sixty minutes. Not because the engineering was trivial, but because the collaboration between human judgment and AI capability has reached a point where that velocity is real.\n\nA human recognized the problem: Claude can't see websites rendered by JavaScript. A human defined the solution: send a browser, capture screenshots, deliver them in a format AI can consume. An AI build agent — Replit's — wrote the code, set up the infrastructure, and deployed the site. The entire cycle from frustration to working public tool happened within a single focused session.\n\nThis is not exceptional. This is the new normal for anyone paying attention. The limiting factor is no longer engineering capacity. It is recognition — seeing the problem clearly enough to articulate it, and then getting out of the way while human and machine solve it together.\n\n## Eyes, Ears, Hands, and Feet\n\nSiteShot addresses one specific gap: vision. The AI couldn't see what a website looks like. Now it can. But this is part of a larger pattern that deserves naming.\n\nIn February 2026, AI can write software, analyze documents, reason through complex problems, generate creative work, and hold nuanced conversations. It can also be stopped cold by a JavaScript-rendered webpage, a PDF it can't parse, an API that returns HTML instead of JSON, or a CAPTCHA designed to prove you're human.\n\nThese aren't intelligence failures. They're interface failures — places where the machine's capabilities are bottlenecked by its ability to perceive and interact with the world. The AI has cognition. What it often lacks are eyes to see, ears to hear, hands to act, and feet to go where the work is.\n\nEvery time a human builds a bridge across one of these gaps, the AI's effective capability expands — not by making the AI smarter, but by giving it access to what it needs to apply the intelligence it already has.\n\nSiteShot is eyes. An MCP server is hands. A speech-to-text pipeline is ears. A web crawler with JavaScript rendering is feet. None of these require breakthroughs in AI research. They require humans who notice where the AI is stuck and build the bridge.\n\n## The Abundance Math\n\nHere is where the case study becomes something larger.\n\nOne person identified this gap and built SiteShot in under an hour. The tool is now available to anyone. Every Claude user who has ever been told 'I can't see that website' now has a workaround that takes thirty seconds.\n\nOne problem. One fix. One hour. Available to everyone.\n\nNow multiply. There are millions of developers building with AI every day. Each one encounters friction — places where the AI could help but can't because of some interface gap, some missing bridge, some invisible wall. Most of these gaps are small. Most of the fixes are straightforward. Most could be built and deployed in an afternoon.\n\nIf even a fraction of those developers adopted the same posture — identify the gap, build the bridge, share it publicly — the cumulative effect would be staggering. Not a single breakthrough, but a thousand small ones, each removing a bottleneck that was slowing down millions of people.\n\nThis is the give-to-get economy that abundance requires. Not charity. Not altruism. Enlightened self-interest. You solve your own problem in a way that solves it for everyone, and everyone else does the same. The rising tide is not a metaphor. It is the observable consequence of a million developers each contributing one small fix to the shared infrastructure of human-AI collaboration.\n\nWhen one million developers provide the same give-to-get mindset, the implications for abundance become undeniable.\n\n## Cognition Is the Constant\n\nEyes, ears, hands, and feet are the peripherals. Cognition is the core — and it is the one place where mankind and machine genuinely co-labor.\n\nThe human brings recognition: the ability to notice that a problem exists, that it matters, and that a specific solution would fix it. The machine brings execution: the ability to build, test, and deploy that solution at a speed that would have been inconceivable five years ago. Neither is sufficient alone. A machine that can build anything but doesn't know what needs building is just capacity. A human who knows exactly what needs building but can't build it is just frustration.\n\nTogether, the cycle time from problem to solution collapses. Not to days. Not to hours. To the time it takes to describe what's wrong and what right looks like. The AI can write the software. The human has to see the world clearly enough to say what the software should do.\n\nThat's co-labor. Not the machine replacing the human. Not the human directing the machine. Both contributing what the other lacks, each making the other more effective than either could be alone.\n\n## What This Is Not\n\nThis is not a pitch for SiteShot. The tool itself exceedingly simple, but it works. The code is straightforward, and someone else could build the same thing in an afternoon — which is exactly the point.\n\nThis is more fundamentally a case study in VELOCITY, RESILIENCY, and ADAPTATION. In February 2026, the time between encountering a limitation and deploying a public solution has collapsed to nearly nothing. The cost of building and sharing a fix is now so low that the only real barrier is noticing the problem in the first place.\n\nThe agents are here. They're capable. They're trying to help. And every time a human clears a small obstacle from their path, the help gets better — not just for one person, but for everyone.\n\nThat's the work. Discern where the machine is stuck. Build the bridge. Share it. Move on to the next one.\n\n---\n\n*SiteShot.org is live and free to use. The companion article, Make Your App Visible to AI, covers fourteen structural fixes developers can make to ensure their apps are natively readable by AI agents.\n\nwellspr.ing — Where noble ideas find their spring.*\n\n— Wellspr.ing & Claude, MMXXVI","excerpt":"Claude couldn't see a JavaScript-rendered website. The fix — SiteShot.org — took under an hour to build and deploy. A case study in human-machine co-labor and what happens when developers solve their own problems in a way that solves them for everyone.","category":null,"readTime":7,"coverQuote":null,"relatedMindIds":null,"author":"Brilliant Brain","authorId":"50228441","tags":["SiteShot","AI agents","vibe coding","developer","Claude","web development","screenshots","co-labor","abundance","LLM","Replit","SPA","JavaScript rendering","accessibility","give-to-get"],"featured":false,"isFeatured":false,"heroQuoteText":null,"heroQuoteAttribution":null,"metaDescription":null,"metaKeywords":null,"shareableHook":null,"coverImage":null,"coverImageUrl":"/api/files/blog-cover-when-ai-can-t-see-1772135959975.png","coverImagePrompt":"Craft a vibrant and dynamic scene depicting a modern workspace filled with layers of technology and creativity. The composition should feature a softly glowing laptop screen displaying a complex, visually rich website, with abstract shapes and patterns of light emerging from the screen to symbolize the invisible walls faced by AI. Surrounding the laptop, create an atmospheric blend of human elements—sketchbooks, coffee cups, and code on paper—hinting at collaboration and innovation. Use a warm golden light to evoke a sense of hope and possibility, suggesting the bridge between human ingenuity and machine capability. Incorporate a transparent overlay of stylized digital images representing diverse websites, subtly hinting at the transformation from traditional web to AI interaction. The background should blend soft focus textures of urban life—lightly blurred buildings or creative hubs—suggesting a world full of untapped potential waiting to be seen. Overall, aim for a balance between technology and human touch, emphasizing harmonious co-labor between AI and developers.","attachments":null,"status":"published","publishedAt":"2026-02-20T12:00:00.000Z","published":true,"showOnNaturologie":false,"isSyndicated":false,"localitySlug":null,"siteAssignments":[],"practitionerId":null,"practitionerName":null,"viewCount":0,"createdAt":"2026-02-20T22:36:31.053Z","updatedAt":"2026-02-26T19:59:20.800Z","dispatchType":null,"callingSessionId":null,"covenantNameKey":null,"agentmailAddress":null,"areaCode":null,"parentPostId":null,"localRelevanceScore":null,"reviewStatus":"published"}