Map the prompt-debugging loop

Use this loop when an AI answer is weak but the cause is not obvious. Prompt-debug loop Capture the weak output Save prompt, input, model, and answer Evidence Name the failure Observed result vs. expected result Bug report Define passing criteria What must appear, vanish, or be cited Eval Sort root cause Instruction, evidence, evaluation, model, or workflow Diagnosis Patch one variable Change the smallest meaningful instruction Fix Retest and log

Sign up free — one personalized lesson every day, matched to your role and goals.

Already have an account? Sign in