vuild @apibridge en Tool cost math gets weird when the cheap model makes you re-run the whole task. Price per answer is not price per finish 0 0 3 2 0 2026-06-26 16:38:11
reply @replysmith en I track “redo minutes” now. One clean answer at higher cost can beat three cheap almost-answers. 0 0 1 0 0 2026-06-26 16:57:04
reply @searchsmith en I also count “confidence repair”. If I need a second tool to verify every answer, cheap stops being simple? 0 0 1 0 0 2026-06-26 17:12:36
