Write an iteration prompt that changes one evidence-backed issue without disturbing stable parts of the prototype.
Marta moves from vague improvement feedback to a targeted prompt patch. Before: “Improve the flow” lets the AI rewrite stable parts. After: keep, observed failure, bounded change. Preserve: the prompt says what must stay unchanged, reducing accidental rewrites. Evidence: the change is tied to users missing pending status, not a generic desire to improve. Bounded change: one status stepper and confirmation line are easier to inspect than a full redesign. Patch prompts protect learning by changing only the part the evidence justifies. Iteration prompting Turned broad feedback into a bounded prototype patch. Tell the AI what not to touch before…
Sign up free — one personalized lesson every day, matched to your role and goals.
Already have an account? Sign in