A problem-set prompt should focus on method choice and error patterns. The useful output is not a finished answer key; it is a list of what you misunderstood, what to re-practice, and how to recognize the right method next time.
Good output asks for
- Where the attempt first goes wrong.
- Which concept or formula the problem is testing.
- Practice variants that change one feature at a time.
- Hints before solutions.
- A short plan for reworking the problem yourself.
Copyable starter prompt
You are reviewing my attempted problem-set work.
Do not solve the problem for me first. Read my attempt and identify:
1. the first incorrect or unsupported step,
2. the concept being tested,
3. the method I should have recognized,
4. a hint that helps me repair the attempt,
5. one similar practice problem with a different surface detail.
Only give the full solution after I ask for it. If my attempt is missing, ask me to try one setup step before giving help.How to keep it useful
Paste your work, not just the question. The model needs to see your setup, notation, assumptions, and where you got stuck. When the output comes back, rewrite the problem yourself and compare your second attempt against the hint. That is the part that turns help into learning.
For math, economics, coding, or science classes, ask for the named method and the signal that should have made you choose it. Those recognition cues are often what exams actually test.