Shadcn made a prompting utility, said it worked great, and asked a simple distribution question: if you buy it, do you want the app, or the app plus its source?
The source won 83.1% to 16.9%. Out of 1,697 votes, the finished thing you can't change lost by a landslide to the finished thing you can.
Everyone read this as another data point for personal software. It's the opposite. Buried in that number is a distribution problem nobody's naming.
This week the question sharpens: if everything is copyable, what is still worth distributing?
Start with the story everyone agrees on. Code is no longer scarce. "Producing code is becoming the cheapest and least distinctive part of building software." Agents write it, test it, document it, iterate on it.
So the pitch goes: stop shipping apps, start shipping blocks. Shadcn put it plainly. "Instead of only distributing apps, you distribute small, well-built, well tested blocks that an agent can put together. For you." Personal software.
And then the load-bearing claim: "people will pay for small parts that work well together. your agents will look for them." The block becomes the paid unit. This is the bet everyone is excited about.
It's a trap. Watch the same logic that kills the app kill the block.
JOV.AI lays it out with a scene. Two business owners, both stopped buying software. One reportedly built a full property management system in two weeks — records, CRM, fleet, financials — instead of renting it for $200,000 a year. The other is systematically ripping out his rented stack.
Then the turn, and it's the sentence this whole piece hinges on. "A million people are building ERP systems. A million people are building project management tools. A million people are building CRMs... and none of them have a market. Because anyone who wants that software will just build their own."
A million people are building CRMs, and none of them have a market. Because anyone who wants that software will just build their own.
Read that carefully, because it does not stop at apps. The block seller has exactly the same problem as the app seller.
If assembly is trivial enough that a million operators can each build their own CRM, then re-assembly is trivial too. The moment I can ask an agent to snap your well-tested blocks into an app, I can ask it to generate blocks that do the same thing. The tested block is a commodity one layer down from the commodity app.
Shadcn's own analogy gives the game away. Shadcn/ui changed the unit of distribution from a packaged library to components you copy into your repo. It won precisely because the components were not scarce — they were yours to deform. Pebblebed reportedly put it this way: "One good idea almost immediately spawns an infinite number of shadcn ui clones."
That is the mechanism. The thing that made shadcn/ui spread is the same thing that makes a market in shadcn-style blocks impossible to defend. Copyability is not a feature you can sell twice.
So the block economy has no floor. When execution is free, whatever you distribute as raw material gets re-generated by the next person who wants it, and your price collapses to zero. The vibecoders racing to sell tested parts are building the next commodity layer down, and calling it a business.
Now look back at the poll. The 83.1% did not vote for blocks. They voted for a finished, opinionated thing — and then the right to change it.
The framing shadcn used is exact: option B is "buying the app but also a 'head start' that you can change to fit your workflow." Not parts. A head start. A pile of pre-made decisions someone already committed to, with the seams left open.
That is the unit that survives. Not the block, which anyone can re-generate. The seed — the source plus the taste plus the specific set of choices you would rather not remake from scratch.
Because here is what re-generation cannot cheaply reproduce: judgment. The blocks are interchangeable; the decisions about which blocks, in what arrangement, tuned to what use case, are not. That is what people paid for in the poll. They were paying for shadcn's decisions, delivered in editable form.
JOV.AI arrives at the same place from the other side, listing what survives when code is free. Product leadership only if "the complexity is real" — models, optimization, "a decade of algorithmic research," not a dashboard a weekend can approximate.
Customer intimacy wins: "anywhere the deliverable is judgment, accountability, or trusted expertise — with software as the delivery mechanism rather than the product." And two moats outside the framework entirely: regulatory posture and infrastructure position. A vibe-coded clone "might replicate the features. It won't replicate the compliance posture."
Every one of those is a seed, not a block. They are pre-committed positions — a domain model, a compliance stance, a set of hard-won defaults — that resist re-generation because the value was never in the code. It was in the decisions the code encodes.
Which is why "personal software" is misnamed as a distribution model. It is a distribution dissolver.
The whole personal-software thesis says the finished artifact stops mattering — software becomes a "living tool that isn't the same as anyone else's copy," a harness you deform. Pi's entire pitch is a list of what it didn't build: no MCP, no sub-agents, no plan mode, do it yourself. The artifact dissolves into the act of assembly.
But if the artifact dissolves, so does the unit you were going to sell. You cannot charge for a thing that is defined by being changed past recognition. The block-seller's market evaporates for the same reason the app-seller's did.
What does not dissolve is the starting point. The poll landed at 83.1% because a head start is the one thing re-assembly does not hand you for free — someone still had to decide what to start from.
So the map redraws around three units, and only one holds value. The app: dead, anyone builds their own. The block: dead on arrival, anyone re-generates it. The seed: durable, because it packages the decisions no one wants to remake — and it resists copying not through legal lock-in but because taste and posture and accountability don't fall out of a prompt.
Shadcn is half right and half selling the wrong thing. Right that the unit of distribution is fracturing. Wrong that value settles on the parts.
The 1,697 voters already told him where it settles. Not on the parts your agent looks for. On the seed you'd rather not plant twice.