Open research: how AI models actually learn

We run a public research programme on how AI models actually learn, using models small enough that anyone can reproduce every number on a laptop in minutes. We publish the results and the data behind them here, including the experiments that went against what we expected.

These pages are the records: the design, the numbers, the controls and the limits. Most of them are about a measurement that misled us before we caught it, which is the same class of failure that makes vendor benchmarks unreliable. If you would rather have the story than the record, the blog carries a plain-language version of most of these, written for a reader who has never seen a training curve.

Ongoing research. These are experimental results from active work, not settled conclusions. The numbers are what we measured and the methods are described so you can judge them, but the programme is still running: later experiments here have already overturned earlier readings more than once, and several pages record exactly that. Expect this section to change as the work moves forward.