Building SotaZoo: a multi-product AI agent company

The state of the art has a distribution problem.

The best tennis coaches in New York know things the rest of us don't — how to read a player, which courts get the best evening sun, who pairs well with whom. The best engineers have shipped patterns and workflows that take months to learn alone. The best traders have decades of pattern recognition baked into their instincts. In every domain, somebody is already operating at the state of the art — and the rest of the world is watching from the cheap seats.

That gap is what SotaZoo exists to close.

We started SotaZoo because the same technology that put a few labs at the frontier — large language models, agentic reasoning, AI-driven coordination — is now mature enough to translate that frontier into apps anyone can use. The state of the art doesn't have to stay walled-off in a few elite practitioners' heads. It can be learned, packaged, and shipped.

What we mean by 'compounding past the state of the art' is straightforward: each app we ship should make its user measurably better at the thing they care about. Not 'more efficient' in the abstract — better. A better tennis day. A better trading week. A better evening planning. And because each app keeps learning from how its users actually behave, the bar keeps rising. Compound interest on practice.

Our first product, TennisMatch, is a small proof of the thesis. It's an AI-matched tennis app for New York. We learned what the best partner-finding services already do — a few referrers in tennis communities, manually pairing people, trust-building rituals at the end of each match — and packaged that into an AI agent that does the matching, verifies the players, and bookkeeps the trust. The result: a player spends more time playing and less time chasing.

That's one small example. The roadmap is more apps in the same shape. A coding-companion agent that teaches better engineering by working alongside the user. A market-analysis agent that compresses the morning brief that a senior trader builds in an hour. A team-management agent that captures how the best operators run their week.

Each app is a single bet on a particular domain — but the engine is the same: learn from the best, package into agents, ship to users who couldn't otherwise touch that practice. We're calling it the SotaZoo of AI agents.

If our products work, every individual user gets to operate above the state of the art that was practical for them last year. The frontier moves. That's the only outcome that matters.