"Days of planning, research, testing, designing, simplifying, throwing ideas away, and trying again." That is what shadcn says went into the Questionnaire component before he felt it was ready to ship.
Within hours of release, someone pointed an agent at his code and ported the whole thing to another framework.
His response was not a takedown or a subtweet. It was a confession: "Of course, it's open source. They were allowed to do it." And then: "I've been around long enough, with multiple successful products, that I don't care."
The person who made the port did nothing wrong. shadcn refused to even tag them. The whole idea of shadcn/ui, he reminds everyone, is exactly this — "copy these components. No attribution required."
Which is what makes the warning that follows land so much harder than the usual open-source lament. This is not a maintainer discovering that a license he chose was too permissive. It is a maintainer discovering that the thing that made his chosen license survivable has quietly disappeared.
Read his own explanation of why the model ever worked. "Copying once demanded effort. You had to study the work and understand it to copy it. That effort created appreciation for the person who made it. Not anymore."
Sit with that sentence, because it is the whole argument. The license said copy freely, no attribution required. The economics said: they'll appreciate you anyway, because copying is hard enough that the act of copying is itself a kind of study, and study breeds recognition.
Copying once demanded effort. That effort created appreciation for the person who made it. Recognition was the payment. Not the license—the friction.
Recognition was the payment. Not the license — the friction.
You made something small, people copied it, and in copying it they came to understand what you had done and who had done it. That understanding turned into use, into reputation, into reach, into the thing that funded the next attempt. shadcn names the loop plainly: "You keep going because, somewhere along the way, the work is rewarded. Maybe people use it. Maybe they pay for it. Maybe they simply recognize the thought, effort, and care that went into it."
An agent collapses that entire loop into a single prompt. There is no study. There is no understanding. There is no moment where the copier internalizes the care that went in, because the care is precisely what the model abstracts away.
The output arrives without the appreciation that used to arrive with it. Copy-freely-no-attribution was always a bet that the copier would attribute anyway, out of comprehension. The bet just stopped paying.
shadcn is careful to say the veteran survives this. He has multiple successful products; he doesn't care. The case he actually raises is the newcomer: "But what if this was someone's first product? What if they spent months building it, launched on Friday, and by Monday it had been cloned by agents, repackaged, and made free?"
Then the escalation — your second product, your fifth, your tenth. "What happens when your roadmap, your changelog, and your launch announcements become someone else's input prompts?" The genre of creative work being killed here is not the tenth product of an established figure but the first product of someone deciding whether to take the risk at all.
So the industry reaches for defenses. Each one, read closely, is a retreat from the openness that made the work legible in the first place.
Defense one: data as moat. Under shadcn's own post, a reply states it flatly — "Data is the moat. Not user data. Real data that powers your app. That can't be copied (easily)." Carles Abarca reaches the same conclusion after the Claude Code leak: a clone "can't replicate the millions of conversations that trained Claude."
This is true, and it is also an admission that the answer to "my open thing was copied" is "stop being open." The moat is the part you don't share. For a component library whose entire premise is sharing, that is not a defense — it's a genre change.
Defense two: private test suites. In February, tldraw reportedly pulled all its tests into a private repo to prevent, as Malte Ubl put it, a "Slop Fork." mbleigh framed the pivot exactly: a comprehensive test suite used to be a durable moat signaling quality, and now "the moat that became the drawbridge."
The logic is airtight and grim. A robust test suite fully specifies hard-won behavior; a model reads intent straight off the tests and reimplements across languages in a week or two. So the tests, once the proudest public artifact of a serious project, go behind the wall.
Defense three: restrictive licensing. And here the evidence is brutal about who pays. Thoughtworks documents that non-commercial clauses often function as "a death sentence for project adoption" — corporate developers abandon the tool the moment licensing triggers procurement review.
Even calibrated dual-licensing, like Akka's threshold aimed at organizations above $100 million in revenue, gets met with boycott rather than compliance. And mbleigh notes the deeper futility: even moderately restrictive licenses don't stop a weaponized clean-room reimplementation, because the only thing that changed is that rewriting dropped to near-zero cost. Cloudflare's vinext reportedly rebuilt most of Next.js in a week for $1,100 in tokens; a University of Washington researcher reportedly turned leaked Claude Code into a working Python reimplementation in an afternoon, plausibly immune to DMCA because it isn't a copy.
Notice the pattern across all three defenses. Each one works by withholding — private data, private tests, hostile licenses. Each is a way of un-sharing the thing you built your reputation by sharing.
And each is affordable only to the already-established. shadcn can say "I don't care" because he has the products, the audience, the standing. tldraw can wall off its tests because it has a commercial license and a business model to protect. The newcomer has nothing to withhold yet — no data moat, no proprietary behavior, no leverage in a procurement fight. Withholding is a rich creator's defense.
This is where shadcn's warning breaks from the usual maintainer-burnout story. The Thoughtworks account is about the exhaustion of people already doing the work — slop pull requests, harassment from billion-dollar consumers, projects closing to contributions. Real, but downstream.
shadcn is pointing further upstream, at the decision to start. "Creativity begins with small steps. If we make those first steps feel pointless, we don't just lose small products. We lose the people who might have gone on to build the big ones."
The friction that used to convert copying into appreciation was doing invisible economic work: it was the payment rail for a system that charged nothing at the door. Remove it and the license still says free, but nothing flows back. The permission was never the point; the appreciation the permission relied on was.
Which means the sustainable fix is not a cleverer license. mbleigh is right that the genie won't return to the bottle and that lawyering a line between "legitimate fork" and "slop fork" is hopeless when the only variable that moved is cost.
The fix has to be a rebuilt compensation loop — patronage, distribution, recognition — some mechanism that pays creators for the care that agents now strip out and discard. shadcn frames it as a choice about defaults: "We can use AI to build great things. Or we can use it to strip every new idea for parts. We decide."
He is careful to note this is not really about open source, and he's right. It's about what happens when the creator "pays the full cost of creativity but everyone else pays nothing to reproduce it" — a problem that will spread, in his words, to "writing, design, music, research, and every other field where someone still has to take the first creative risk."
The previous open thread was: if seeds are the durable unit, how do you price taste? shadcn's Questionnaire is the answer nobody wanted. You can't, not through the license — because taste was never priced through the license.
It was priced through the labor of understanding it, and that labor is exactly what we just automated away. The bill for creativity still exists. We simply deleted the only mechanism that ever collected it.