Dear Magicians,

There have always been trade-offs between big and small organizations. Large teams solve problems. Small teams invent problems worth solving. That was the finding from Dashun Wang’s 2019 research at Northwestern — a clean division of labor between institutional scale and scrappy ingenuity.

That was 2019. Before the collapse of writing costs.

Today, any addition to a one-person team is likely to be named ChatGPT or Claude. As Wang recently noted, a risk of tiny teams that rely on AI is that “now all of a sudden they look a lot more similar, because they, in some sense, have collaborated with the same person” — who, of course, is not actually a person.

Think about that. The edge you thought a small, nimble team had was cognitive diversity. Different brains seeing around different cognitive corners. But if the second brain in your two-person team is Claude, and the second brain in my team is also Claude, and the second brain in every ambitious solo operator’s team is Claude — we’ve outsourced our diversity to the same model trained on the same data with the same optimization objectives.

The appearance of a small team persists. The actual advantage — heterogeneity of perspective — has been replaced by homogeneity dressed up as collaboration.

This isn’t about AI being bad. It’s about the economics of outsourcing cognition. When the marginal cost of an additional perspective drops to zero, and that perspective is trained on consensus, the incentive structure flips. You’re not assembling diverse viewpoints anymore. You’re assembling redundancy with different latency.

Spotify discovered something similar a decade ago. Their shuffle algorithm was too random — users heard clusters and streaks and assumed it was broken. The fix: make it less random so it felt more random. They called it “smart shuffle.” A deliberate departure from statistical randomness in the service of perceptual randomness.

We’re doing the same thing with our teams. We’ve smart-shuffled our collaborators. It feels like diversity. It performs like consensus.

The teams that will keep an actual edge are the ones that stay uncomfortably human — that deliberately defend the friction, the slowness, the genuine disagreement that comes from people who see differently. Not because they’re plugged into the same model. Because they’ve lived different lives.

Guard your team’s weirdness. It’s about to be your only advantage.

Until next time, have a M.A.G.I.C. Week,

Brian

P.S. Read all my Musings on Substack

P.P.S. 📷Check out Ad-free episodes on Patreon: patreon.com/drbriankeating