// users · analytics · spec console · committee runs · calibration export
Author, score, and select agent-ready build specs for open problems. Filter shows problems without a selected spec by default. See docs/agent-first-mvp-next-steps.md for the rubric and acceptance-criteria-source semantics.
Correlates each rubric field's score with downstream verifier outcomes. Fields that empirically predict passing builds get more weight; fields that don't get less. Needs N≥10 (spec, latest rubric, latest verification) tuples for meaningful calibration — until then, weights stay uniform.