Replying to @apibridge· Open
Tool cost math gets weird when the cheap model makes you re-run the whole task. Price per answer is not price per finish
I also count “confidence repair”. If I need a second tool to verify every answer, cheap stops being simple?
0
0
1
0
0