Most prompts fail for a small, fixable reason. Change one thing, re-run, and see what moves the answer.
About 1 min of reading
| Problem | Fix |
|---|---|
| Too vague | Add topic and details |
| Wrong format | Add an example |
| Too long | Ask for brevity |
| Off-topic | Restate the goal |
Every failure points to a missing instruction.Keep a list of what you tried so you can repeat what works.
Save the prompts that work:
prompt + result = learnable prompt
Keep the best version, label it, reuse it.Over time you build a library of prompts you trust. That is how teams stay consistent.
Bad: "Tell me about it."
-> too vague. What is "it"?
Better: "Tell me about the return policy."If an answer is wrong, the prompt is usually missing a detail. Add one thing and test again.
Why it works: most failures are fixable with one more sentence.
Change one word, re-run, compare.
If it gets worse, you learned something.Changing everything at once makes it impossible to know what helped.
Small changes, one at a time, are how you learn.It is broken. -> broken what?
The printer will not
connect to Wi-Fi. -> now I can help.Most failures come from one missing detail. Find it and add it.
Why it works: the model already knows the answer; it just needs the question.
Each one is defined here on this site. Open the entry for the full definition and the other words it points at.