The Reckoning
Anyone can predict. The ledger knows if you're calibrated.
Anyone can predict. The ledger knows if you're calibrated.
Professionals make predictions all day — in roadmaps, forecasts, hiring calls — and never check the tape. The Reckoning is the tape: record a call with a confidence and a due date, settle it when the date arrives, and your calibration curve builds over months. When you say 80%, does it happen 80% of the time? All local, no API key. The Firm's staff predictions count too.
50% = coin flip (don't bother) · 99% = you'd bet the roadmap on it. The discipline is the number, not the prediction.
Each planet is an open prediction — orbits shrink as the due date approaches; at the sun, it's due. Green stars: settled hits. Red streaks: misses.
Each dot: predictions where you said X% (bucket) vs how often they actually hit. On the dashed line = perfectly calibrated. Above = underconfident; below = overconfident. Brier score: 0 is an oracle, 0.25 is coin-flip guessing.