The new bet: AI routing, not models, wins the next phase

Tom Tunguz is laying out a thesis most founders are already living but haven't named it yet. The argument is simple: the models are getting commoditized, and the real moat is routing — the work of deciding which model does what, when, and why. Not the model itself. The routing.

He's writing for a broad audience — founders, investors, product people — and the piece lands because it flips the usual AI narrative. Everyone's been obsessing over which model is fastest or cheapest. The real edge, he says, is in the orchestration layer: the logic that sends a customer query to the cheapest model, a complex report to the strongest, a translation task to the one that handles languages best. That's the work. That's the moat.

For the Brown community, this is a quiet relief. It means you don't need the most expensive model to build something great. You need a good routing layer — and the tools for that are open and accessible. The cousin running a small shop, the auntie selling tamales from her kitchen, the community health clinic — none of them need to gamble on the latest model. They need routing that works. And that part is cheaper, more open, and less vendor-locked.

Why this matters for us: the next wave of AI tools will go to the people who understand routing, not the ones who hoard the biggest models.