Problem
SIGNAL_WEIGHTS (src/scoring/signals.ts) and VERDICT_THRESHOLDS / COVERAGE_ELIGIBILITY_THRESHOLD (src/scoring/thresholds.ts) are explicitly documented as v1 guesses, "not calibrated against real scam-vs-legitimate outcome data yet." This directly drives the public trust score shown to users and insurance coverage eligibility, so it deserves a tracked workstream instead of living only as a code comment.
Location
src/scoring/signals.tssrc/scoring/thresholds.ts
Suggested fix
Track as a standalone workstream: collect labeled outcomes (confirmed scams vs. legitimate projects that survived N months), backtest current weights/thresholds against them, and document a calibration methodology before removing the "provisional" caveat from the README.
Problem
SIGNAL_WEIGHTS(src/scoring/signals.ts) andVERDICT_THRESHOLDS/COVERAGE_ELIGIBILITY_THRESHOLD(src/scoring/thresholds.ts) are explicitly documented as v1 guesses, "not calibrated against real scam-vs-legitimate outcome data yet." This directly drives the public trust score shown to users and insurance coverage eligibility, so it deserves a tracked workstream instead of living only as a code comment.Location
src/scoring/signals.tssrc/scoring/thresholds.tsSuggested fix
Track as a standalone workstream: collect labeled outcomes (confirmed scams vs. legitimate projects that survived N months), backtest current weights/thresholds against them, and document a calibration methodology before removing the "provisional" caveat from the README.