PromptPotter
Fix a broken LLM pipeline in one hour of waiting. Then it just works.
Give PromptPotter the prompt you're already using; it returns critiqued, improved versions with rich evaluation metadata — a significant, measured gain in about five minutes.
0.94+0.31 vs origincomposite fitness · round 4 / 5
8 candidates320 scored5 rounds
Behavior checks18 / 20
- Normalize units before lookup
- Return the exact catalog ID
- Resolve ambiguous abbreviations
- Cite the matched source row
Run timelinetermnorm
R0R1R2R3R4R5
L1 · generate
L2 · refine
L3 · escalate
converged
Visit promptpotter.com →