The Aggregator Playbook

Google didn't win by being the best website.

Google won by being the only way 4 billion people reached the rest of the web. The websites — every news organization, every publisher, every retailer — sat on one side. The users sat on the other. Google was the layer between them.

Neither side could route around it without giving up the other.

Facebook (now Meta) did the same thing with social connection.

Your friends — the producers of "what's happening in my life" — sat on one side. You sat on the other. Facebook owned the layer. It built the feed, the algorithm, the ranking. You went there to see your friends. Your friends went there because that's where you were.

Neither could route around it. Same shape, different surface.

Amazon did it with commerce. 9.7 million sellers on one side. 2.7 billion annual buyers on the other. The marketplace in between.

You'd think this would be easier to route around — sellers have their own websites, buyers have their own browsers. But by the time Amazon's marketplace was mature, the cost of being not on it was higher than the cost of being on it. Both sides became dependent.

Same shape. Third time.

The pattern is the same every time. Two parties who need to find each other, a layer that sits between them, and a moment where the layer becomes indispensable for both sides.

The most valuable companies of the last 25 years all built one of these. Not because they had better products. Because they captured the layer between producer and user.

Ben Thompson calls this the "aggregator." Sangeet calls it the coordination layer. Same idea.

The reason this idea matters in 2026 is that AI is the next aggregator. The same shape, the same playbook, on a new surface.

On one side: a user with composite intent — "book my Lisbon trip, inc. one dinner, with the kids." On the other: every system, every API, every domain-specific tool that needs to be coordinated to fulfill it.

The aggregator in the middle is an AI. Probably an agent. Almost certainly a system you don't yet recognize.

Right now, in 2026, every market has a race underway to become its AI layer.

Travel — who becomes the AI agent that books for you? Coding — Cursor, Copilot, Claude Code? Health — what's the AI between you and your doctor? Education — between you and a curriculum?

The companies racing are the ones who understood the playbook. The companies sitting it out are the ones who'll be aggregated.

The playbook hasn't changed. Sit between producer and user, become indispensable to both, and the value flows through you for a decade.

— Adapted from Reshuffle, Section 2

If you build products, the strategic question isn't "how do we use AI?" It's "how do we become the AI layer in our market — before someone else does?"

And if you can't become the layer, the next-best position is to be such an essential producer that the layer can't aggregate without you. Those are the only two safe positions. Everything else is in the middle of the funnel, getting compressed.

Sangeet on this in Section 2 ↗