About Empirical AI

A small research lab studying world models — and a public, verified survey of the field.

The survey half: world models became the most interesting corner of AI while nobody was maintaining a map of it. Claims in this space are unusually noisy — press releases inflate parameter counts, "open" models turn out to be downscaled or hosted-only, benchmarks disagree. So we adopted a rule: every factual claim on this site is verified against primary sources (papers, model cards, license texts) before it ships, and sections that haven't cleared that bar sayin progress rather than pretending.

The lab half: we run small, reproducible world-model experiments and publish what actually happened — including the failures and the compute bills. The name is the thesis: empirical. See research.

Contact

General: hello@empirical.world
Research & survey corrections:research@empirical.world

Empirical AI is a 37digital project by Matthew Bouchard.