Generative models collapsed the cost of the hardest part of building software. The bottleneck moved. We think it moved to taste, specificity, and distribution — and almost no one is structured to take advantage.
The last twenty years of software optimized for scale. The shape of a successful company was: one product, sold to as many people as possible, for as long as possible. The unit economics rewarded breadth. Generality was a feature, not a defect.
That shape worked because building software was hard, and so the cost of starting a second product was almost always higher than the cost of stretching the first one to cover more ground. Big companies got bigger; the next product was always a feature in the existing one.
Generative models broke that arithmetic. The expensive part of software is no longer the building. It's deciding what to build, for whom, and getting it in front of them. Three things that don't scale.
§ 01What the shift actually changed
It used to be that the smartest person on a team spent their week shipping the third quarter of a product roadmap. Now that work is a weekend. The bottleneck moved upstream — to taste in choosing the problem, specificity in defining it, and patience in distributing the result.
None of those things are getting cheaper. If anything, they're getting more valuable, because the floor of "things that technically work" is now extremely close to the ceiling of "things you'd actually use".
The expensive part of software is no longer the building. It's deciding what to build, for whom, and getting it in front of them.
§ 02The magazine analogy
A magazine is not a platform. It picks a narrow audience, makes a finite, fully-finished thing, and ships it. The next issue is a different finite, fully-finished thing. The publisher is the persistent entity; the issues come and go.
We think the next decade of software looks more like that than like the platform decade behind us. Small focused tools, made by small teams, sold to people who want exactly that thing. The persistent entity is the studio. The tools come and go under it.
That's not a thesis about scale. Some of these tools will be huge. It's a thesis about shape — about what gets rewarded.
§ 03Why we are structured this way
We are two people. We will probably grow to four. We will not grow to forty.
We don't take outside money for an individual project, because the math of single-project capital — the dilution, the timeline pressure, the "your one good thing has to also be huge" — is exactly wrong for what we want to do. We hold the parent and let the projects exit.
This is unusual. It is also, we think, what good operators of small things will look like in five years.