Why we run a studio, not a startup.

The standard advice is to pick one thing. Focus is treated as a moral virtue, and the single-product startup as its natural form. We chose the other structure — one studio, several independent ventures — and the choice was not a compromise. For a small technical team working in fast-moving markets, the studio is the higher-percentage bet.

Surface area beats conviction

A single-product company is a single hypothesis about the future, held for years, defended against all evidence. A studio is a portfolio of smaller hypotheses, each cheap to test and cheap to retire. In markets being rearranged by AI every quarter, the scarce resource is not conviction — it is contact with reality. More independent products means more surface area touching real demand, and faster, cheaper learning about where the demand actually is.

The discipline is that each venture must be genuinely independent: its own name, its own site, its own path to users, killable without ceremony. A studio that ships "features of the mothership" under different logos is a startup with extra branding overhead.

Infrastructure is the moat, products are the probes

What makes the portfolio cheap to extend is that the expensive parts are shared. Our data intelligence layer — the extraction infrastructure behind the AgentX catalog — was built once and now powers every data-shaped product we ship. BestCrawler turned the catalog into an editorial surface. The next venture inherits all of it on day one: the infrastructure, the distribution playbook, the machine-readable site pattern we described in The Machine-Readable Web.

Each launch makes the next one cheaper. That is the studio's compounding loop, and it is the part a single-product company structurally cannot have: their second product is a distraction; ours is the business model.

Small teams changed the math

The studio model used to require a studio-sized headcount, which is why most venture studios were funds with builders attached. AI-assisted engineering collapsed that constraint. A team of a few people can now carry the operational load that needed departments — which means the binding constraint moved from hands to judgment: which probes to launch, which to feed, which to kill. Structure follows constraint. When judgment is the bottleneck, the portfolio is the correct shape.

What we optimize for

Not valuation on a single curve, but durable, cash-generating products — each small enough to be excellent, independent enough to survive its siblings' mistakes, and connected enough to inherit the studio's accumulated advantages. Ventures in formation are listed openly on our constellation page; they launch when they are real, not before.

← All notes