Test Bench

Run the raw rant and the compiled prompt on the same model. Compare the two outputs.

Raw rantCompiled prompt

Next: prove it works

Run the prompt here. A/B shows rant vs Magic Polish on the same model.