Every number here is earned, not written. Forecasts are pre-registered, scored against real-world resolution, and posted win or lose. An honest zero today is worth more than a page of claims.
None posted yet. Each experiment appears here — hypothesis, metric, baselines, stopping rule — dated before it runs.
None yet. Papers land here as preprints first, each with its environment, evaluation harness, and data. The framework behind all of this is on the research page.