Why it's worth the line
"Make it better" asks the model to guess what you meant and grade its own homework. It'll change something — rarely the thing you wanted — and you'll both be vaguely disappointed.
Specificity is just naming the operation. "Cut it in half," "replace adjectives with numbers," "make the second paragraph the first" — each is a move the model can execute exactly, no guessing.
How to catch it in your own prompt
- Your instruction contains "better," "good," "nice," "improve," or "polish."
- You're disappointed but can't say precisely what's wrong.
- The same vague ask gives a different (still-vague) result every time.
What to write instead
- Name the operation: cut, expand, reorder, replace, tighten — and on what.
- Quantify when you can: "half the length," "three examples," "one sentence each."
- If you're reacting to a draft, point at the exact part: "the second paragraph," not "this."
Seen side by side
Make this copy better.
Cut this copy to half the length and replace every adjective with a concrete number or claim.
Why it lands — Two exact operations ("half the length," "adjectives → numbers") give a result you can predict; "better" gives a coin flip.
Improve this function's error handling.
Wrap the fetch in a try/catch, return a typed Result instead of throwing, and log the failing URL — leave the happy path untouched.
Why it lands — Naming each change AND what to leave alone keeps the edit surgical instead of a full rewrite you have to re-review.
Stop fixing this by hand.
Paste a prompt — Meerkat catches this and the seven other weak spots before you send it. Free, no signup.