Get a better prompt, with proof. In about five minutes you get an improved prompt and a clear before-and-after score, so you can see exactly how much it gained.
Every round, PromptPotter writes a fresh batch of prompt variations, tries each one, and quietly drops the weak ones early — so it never wastes time or budget chasing a version that won’t pan out.
round 3 of 5 · 5 versions tried v1 step-by-step 94% ✓ best v2 worked example 88% v3 plain & short 71% dropped early v4 add an example — dropped early
Every version is graded against your own examples of a good answer. You don’t get a hunch — you get a number, before and after, so you can see exactly how much it gained.
// graded against your own examples · 250 checked your original 63% correct best new prompt 94% correct +31% still missed 15 of 250
After each round it writes a short note on what helped and what to try next, then uses that note to write the next batch. No knobs to turn, no jargon — it just keeps getting warmer.
what the Potter noticed this round "Asking for step-by-step reasoning fixed most of the multi-step questions. Shorter wording made answers vaguer — drop it. Next round: keep step-by-step, add one worked example."
PromptPotter automates the tedious work of tuning AI prompts — the same prompt-evolution approach behind Google’s AlphaEvolve, pointed at the prompts you use every day. Powerful underneath, simple enough for anyone on top.
A short, plain-language walk-through — no jargon, no setup headaches.
See how far the Potter moves the score across public benchmarks.
The four-layer loop, the API, and the source — the whole thing is open.
“It fixed a six-month-stuck pipeline in an afternoon. Then it just worked.”
— Maintainer dogfooding an internal pipeline, sprint 14
Free to start. Hand over a prompt and see the difference in minutes.
