Yenolab

Building mathematical sense for AI

Are we doing the mathematics right?

Are we doing the right mathematics?

Explore results

10,000

agents, searching in parallel
on Navier–Stokes at OpenAI.

Did Perelman have 10,000 subagents in his head when he solved the Poincaré conjecture?

“He must have had some sense that this was the right track to pursue, because it takes years to get from A to B.”

Terence Tao, on Perelman

Where does that sense come from?

Not from earlier attempts at the same problem. From years spent inside problems he had already solved.

Yenolab is building that sense for AI.

So that when a proof is stuck, it can remember who else was once stuck like it.

Fig. 1 128 thin triangles, one at each slope i/128, packed into area 0.10677…, a new upper bound for Keich’s thin-triangle problem.
Results

Now

Why this proof, and how many more are there?

A second proof does not make a theorem more true. It can make it more useful. Furstenberg reached Szemerédi’s theorem through ergodic theory and built a bridge between combinatorics and dynamics. Selberg and Erdős proved the prime number theorem without complex analysis. The Pythagorean theorem has hundreds of known proofs.

When proofs become cheap, a solved problem becomes a workbench: the destination is known, and the question is how many roads lead there and where else they go.

About

An independent lab, started in May 2026.