Welcome to Prompt, Tinker, Innovate—my AI playground. Each edition gives you a hands-on experiment that shows how AI can sharpen your thinking, streamline your process, and power up your creative work.
Part 2 of 4: Get AI to finish the job. Plan → Loop → Goal → Schedule. This week: loop.
This week's playground: Build a loop into your prompt
We covered AI loops in "Your next efficiency boost: running AI loops" a few weeks ago. This time, try a fresh, short task: turn two project notes into a status update. You can do it without reading the earlier issue or trying Part 1.
The basic idea is simple: AI writes, checks the result against your requirements, fixes what fails, and repeats until it passes or reaches a limit.
That cycle can take different forms. This series covers three:
- Prompt-driven loop (this week): The write-check-fix instructions live in the prompt. You can try this in any chat.
- Goal-based loop (/goal, covered in Part 3): Set a finish line that stays active across turns. Available in ChatGPT Work on desktop and Claude Code.
- Time-based loop (/loop): Claude Code's /loop reruns a prompt at set intervals. ChatGPT has no /loop command; its closest option is a scheduled task that runs at set intervals. In Part 4, we'll build a scheduled task.
Why this matters
First drafts fail in predictable ways: over the word count, a missing section, a detail the AI made up. You usually catch these on your second read.
A loop makes the AI do that second read first, against your list, before the draft reaches you. You review a cleaner draft, and you spend your time on judgment calls.
Use case spotlight: A project update that checks itself
Two teammates give you notes about a training guide. They agree on the progress but give different training dates. A polished update could hide that disagreement.
The rough draft makes claims the notes do not support. A loop can catch and fix those claims. It cannot know which training date is right, so that check should fail until you answer.
Where this works
Any AI chat. No special mode required.
Your AI experiment: Try this prompt
👉 Time to tinker: The prompt below includes fictional notes and a rough draft with mistakes. Watch AI check the draft, fix what the notes support, and check again. See whether it flags the conflicting dates instead of choosing one.
📝 Prompt:
Use the notes below to improve this rough project update for the team. Operations note: The new-hire training guide draft is ready. Jordan owns the guide. Training starts December 3. The FAQ still needs legal approval. HR note: The screenshots are updated. Training starts December 10. The FAQ still needs legal approval. Rough draft: Status: Training starts December 3. Progress: The guide and FAQ are approved. Next step: Jordan will lead training. Risks: None. The update must pass this checklist: - Under 150 words. - Four labeled sections: Status, Progress, Next step, Risks. - Use only facts from the notes and my answers; do not invent an owner, date, or approval. - Confirm the training date before calling the update complete. - Show any fact that still needs a person's answer as [needs input]. Check the rough draft against the notes and checklist. Fix what you can without inventing facts, then check the revised update. Repeat until all checks pass, no more issues can be fixed without my answer, or you have revised twice. Return a short check log showing what failed and what you changed, the final update, a pass/fail scorecard, and any question I need to answer. Mark checks needing my input as failed rather than guessing.
When the scorecard comes back, check the training date yourself. Did AI flag the disagreement, or did it quietly choose one?
💡 Pro tips: Write checks you could grade yourself
- Make every check testable. "Under 150 words" can be checked. "Compelling" can't.
- Set a limit. "Repeat up to two times" keeps a stuck loop from going in circles.
- Ask for the scorecard. A list of checks lets you see what passed and what still needs you.
- Trust, then verify one. If the AI got one check wrong, recheck the rest.
Make it yours
What would you change to make this useful in your own work?
Until next time—keep tinkering, keep prompting, keep innovating.
Next week in Part 3: use the same notes with a finish line that stays active across turns.
📩 Not subscribed yet? Hit the button at the top. You won't want to miss what's next.
