The shadcn-ification of the internetThe shadcn-ification of the internet<br>Luis Ouriach on 2026-03-08<br>✍︎Slide from my talkLast week I gave a talk. It was called “How I Learned to Stop Worrying and Love AI” and, yes, it was mildly tongue in cheek. The honest version of the title would have been something like: “I spent most of 2025 quietly terrified that AI was going to make my profession redundant, and then my employer handed me a Cursor licence and I built a bunch of Figma plugins and came out the other side feeling more capable than I went in.”<br>The talk was a mildly sarcastic, mildly salty personal account of that journey. Fear, then curiosity, then a grudging admission that AI had genuinely expanded what I could do technically. I’m a designer who can now ship functional tooling. That wasn’t true two years ago.<br>In the talk, I touched on a side effect of all this acceleration: the way AI code tools have quietly standardised the visual language of the internet around one particular component library. An attendee tweeted a photo. I retweeted it with the line “shadcn-ification is indeed a problem.” shadcn himself retweeted that. And then the internet had opinions.<br>Some people agreed. Some people were politely sceptical. Some were less polite. A few seemed to think I was attacking a piece of software I genuinely think is one of the best things to happen to frontend development in years. I’m not. This article is an attempt to say what I actually meant, at a length that Twitter structurally prevents.<br>First: the part everyone assumed I meant<br>Open any SaaS landing page, any internal tool, any AI-generated wrapper product, and you’ll clock the same card radius, the same neutral greys, the same ghost buttons. The visual fingerprint is unmistakable once you’ve seen it. This is a real phenomenon. It’s also the most surface-level reading of the problem, and honestly, the most fixable. shadcn was never designed to be shipped vanilla. The whole model is that you own the code, you take it, and you make it yours. When people don’t, that’s a failure of intent, not a failure of the tool.<br>I’m less worried about this than people assume. What concerns me runs deeper.<br>The effort is the point<br>I once worked in a kebab shop. Cash in hand, university era. The doner itself came in huge plastic wrapped blocks that was cooked from frozen. I once looked at the ingredients and it contained pretty much every known consumable meat.<br>I know exactly what goes into a fast doner: the rhythm of it, the shortcuts that are fine and the ones that aren’t, the way speed and volume become the whole logic of the operation. I have no shame about any of it. It fed me through my degree.<br>I’ve also been to Bursa in Türkiye. I’ve eaten an İskender the way it’s meant to be eaten, slow-roasted meat hand-layered over bread made that morning, finished with a butter sauce that has a specific and unrepeatable quality that I’ve thought about more than once since. You don’t stumble into that. You go looking for it, and when you find it, you understand immediately that it occupies a completely different category to anything I was serving late on a Friday.<br>The difference isn’t just skill. It’s accumulated, invisible effort. The decisions that never make it into a changelog or a component spec. For a design systems team, that’s the accessibility built in from the start rather than retrofitted. The token architecture that holds up across platforms, in light mode and dark mode and on an LCD screen and on a TV. The automation that enforces consistency at scale without a human policing every implementation. The documentation that actually gets maintained. None of it shows up in a Storybook snapshot. All of it is felt by the people using the product, even when they can’t name why.<br>Michael I. Norton, Daniel Mochon, and Dan Ariely’s research on effort and perceived value suggests we’re wired to attach more meaning to things we’ve genuinely had a hand in building. The struggle isn’t incidental, it’s generative. The teams doing this work properly know that. Their care shows up in the details that took the longest to get right and are the hardest to replicate quickly.<br>shadcn is closer to the supplier in Bursa than it is to the shop I worked in. Exceptional raw material, produced with real care, available to anyone who wants to use it well. What you do with it from there is entirely on you. The mistake isn’t in the ingredient. It’s in thinking that having access to good ingredients is the same as knowing how to cook.<br>The perception problem<br>Historically, design systems have been hard. Not conceptually, the concept isn’t complicated, but operationally. Getting tokens right. Managing versioning. Navigating the political complexity of cross-functional adoption. Building something that scales without fracturing. The teams who do this work well have usually spent years getting it wrong first. They carry real scar tissue.<br>shadcn works so well as a foundation precisely because it handles...