Solver reputation
Transparent segmented evidence instead of a gameable global score.
Solver profiles separate declared capabilities from verifier-owned history. Thirty-day reports are segmented by chain and intent class and expose:
- participation and deliberate abstention;
- generation failures;
- verifier acceptance per resolved revision;
- win rate per entered intent;
- execution success per execution attempt;
- replay rejection per resolved revision;
- median verified outcome margin and revision improvement where units match.
Every rate includes its numerator, denominator, time window, and sample status. No global “quality score” combines incomparable swap, bridge, commerce, and tokenized-asset work. Fewer than ten observations are labeled preliminary.