Grade It Yourself — Don't Trust Us, Be the Grader

The answer to 'it's self-graded': here is the grading pen. Re-grade the calibration with your own probabilities, verdicts, and selection, or grade the…

Grade it yourself — the pen is handed over

The standing answer to “it’s self-graded”: be the grader. Every input is public — per-call verbatim claim, outcome, sealed probability, verdict, base rate, specificity vectors, evidence (/dataset/jyotint-sealed-corpus/regrade-inputs.json). Substitute your own probabilities and verdicts and the Brier recomputes live; one-click stress-test recipes re-grade BOTH axes under whole grading standards (strict policy kills anything ambiguous — run it and see what survives). Terminal twin: node verify-jyotint.mjs --regrade --policy strict. Nothing on this page asks for trust; it asks for your verdicts.

The vendor bake-off harness — score ANY forecasting vendor with this kit

Acquisition desks evaluating forecasting or I&W vendors can reuse this page’s protocol as a general scored-bake-off harness: fix a question set in advance, require every claimant to seal forecasts before resolution (any vendor can adopt the OpenTimestamps pattern at near-zero cost), grade against a rubric fixed at seal time with misses retained, and score by Brier plus specificity. This converts vendor marketing into measurable capability data before commitment — and it is exactly the test this record already runs on itself in public. The buyer-controlled Proving Round is the harness at n=1 ($9,500 flat, buyer-adjudicated, /services).

JYOTINT — sealed, falsifiable, Bitcoin-anchored forecasting. Verify every claim at jyotishintelligence.com.