Iteration is the core loop, and it works — for about three rounds.
Past that, something changes. Each new version is different rather than better: sentences get reshuffled, a working function gets rewritten in a style you didn’t ask for, a good paragraph gets replaced by a worse one that technically addresses your note. The is responding to “that’s not right” by producing change, because change is what the feedback asked for.
Meanwhile the version you liked best is three revisions back.
What to do
Notice the plateau. The signal is easy once you know it: you can no longer clearly say how this version is better than the last one. That’s the moment to stop.
Go back rather than forward. Rewinding to the version you preferred is almost always faster than revising toward it. See the rewind-dont-repair tip.
Change tactic instead of repeating it. If three rounds of “make it shorter” haven’t produced something shorter, a fourth won’t either. Try a different move:
- Show an example of what you want instead of describing it.
- Ask what’s blocking it — “you’ve tried three times and it’s still too long. What’s making that hard?”
- Start fresh with a better brief, now that you know what you actually wanted.
Keep the good version where it can’t be overwritten. Copy it into a file, or it. The cheapest insurance there is.
Why this is worth saying out loud
Nobody tells beginners that persistence has a downside. Everything about the interface encourages one more round, and stopping feels like giving up. It isn’t — recognising the plateau is a , and the people who look fastest at this tool are usually the ones who stop three rounds earlier than you do.