Ask the Right Question

The What/Where/Why/How Pattern

• Specificity is everything — 'Fix this bug' gets generic suggestions, 'Fix the null pointer in handleSubmit (login.js:47) when user.email is undefined' gets working code.
• The pattern — four elements that transform prompts
  • › What — what's broken or what you need
  • › Where — file path, line number, component name
  • › Why — user impact, dependency, urgency
  • › How — suggested fix direction or approach
• Include error messages verbatim — copy-paste the stack trace. AI is remarkably good at parsing error output.
• You don't need all four elements every time — but the more context you give, the better the response.

Bad Prompts vs Good Prompts

1. 1 Side-by-side comparison
   • › 'Fix this bug' → 'Fix the null pointer in handleSubmit (login.js:47) when user.email is undefined — add a null check before the API call'
   • › 'Make it faster' → 'The user list page loads in 3s. Profile the component renders, identify unnecessary re-renders, and memoize the expensive computations'
   • › 'Refactor the code' → 'Refactor the auth middleware to extract JWT validation into a separate utility — we need it in the WebSocket handler too'
2. 2 Next time you get a bad AI response, re-ask with the file path, line number, and expected behavior — compare the difference
3. 3 Always copy-paste error messages and stack traces verbatim into your prompt instead of paraphrasing them