A Team Is a Bet. I'm Running a Portfolio.

There’s a strategic point hiding inside the solo-plus-AI conversation that gets lost in all the speed talk, and it’s the part that matters. The question was never solo versus team. It’s depth versus breadth.

A team is a concentration of force. That’s the correct structure when the target is known and the challenge is executing one hard, integrated thing at depth or scale. Big infrastructure, regulated systems, the deep moat. When that’s the job, concentrate, every time.

Solo plus AI is a distribution of force. It’s correct when the target is unknown and the challenge is discovery. You spread cheap bets and expect most of them to be wrong.

Put that way, the decision criteria almost write themselves. Choose depth when the target is known, the build is genuinely hard and integrated, the moat is doing one thing extremely well, or the domain demands redundancy and review. Choose breadth when the target is unknown, you’re hunting product-market fit, each experiment is cheap, and being wrong four times out of five is fine if the fifth one lands.

What changed isn’t that breadth suddenly beats depth. It’s that the price of running a broad portfolio collapsed. Running five simultaneous bets used to require capital and coordinated people, which meant only firms could do it. A venture studio was a business model precisely because portfolio-running took an organization. Now one person with domain depth runs the whole portfolio out of their own head. That’s the thing I’d want a reader to sit with: the portfolio strategy didn’t get better. It got cheap enough for individuals.

I’m running one now, several products in parallel, and the obvious pushback deserves a real answer: doesn’t running lots of things mean no focus? It would, if the bets were teams. Five teams means five versions of the coordination tax and a founder spread across all of them. Five solo streams is different in kind. Each product is one coherent thread in one head. I time-slice between them the way a single good engineer has always time-sliced between a feature, a bug, and a refactor. The focus objection assumes the old cost structure.

The caveat I have to own, because it’s real: scaling via multiple agents re-imports the team’s expensive part. Divergent context, integration work, keeping streams in sync. You become a team lead again, except the tax moved from human meetings to agent orchestration, from a shared calendar to your own review queue. A better place for it, but it doesn’t vanish, and pretending it does would be exactly the hype this argument is supposed to avoid. That trap gets its own treatment in Orchestrating Agents Without Becoming a Manager Again.

Teams aren’t dead. The default flipped. For the exploratory front end of building, one person running a portfolio is now the obvious move, and a team is what you reach for later, on purpose, once you know what you’re concentrating on. The part I haven’t settled is how you know the moment to switch. Concentrate too early and you’ve hired a team to execute a guess. Too late and you’re the bottleneck on the one bet that landed. The default flipped, but the timing is still judgment, and judgment is the thing nobody’s model hands you.