Predict a market event, let a local AI predict it too — then let reality score you both.
- Record a thesis on a market event — direction, magnitude, confidence, and what would invalidate you
- A local LLM predicts the same event independently, so you are measured against a real opponent
- Reality resolves both and scores them — Brier scores, calibration, and where your edge actually is
- Local-first — your own machine, your own model via model-dock; no cloud, no account, no API bill