Zum Hauptinhalt springen

Prompting Basics

The quality of AI output depends almost entirely on the quality of your prompt. A vague question gets a vague answer. A precise, contextual prompt gets something you can actually use.

Be specific about what you want​

Avoid open-ended requests. The more constraints you give, the more useful the output.

Weak:

Write a function to parse dates.

Strong:

Write a TypeScript function that parses date strings in the formats
"YYYY-MM-DD" and "DD.MM.YYYY" and returns a Date object. Throw a
TypeError if the input doesn't match either format.

Provide context​

AI models don't know your codebase. Give them the relevant pieces:

  • The language and framework you're using
  • Existing interfaces or types the output must conform to
  • Any constraints (performance, bundle size, no external dependencies)
I'm using React 19 with TypeScript. I have this existing hook:

[paste your hook here]

Add a reset() method that restores all state to its initial values.

Specify the output format​

Tell the model exactly what you want back:

  • "Return only the function, no explanation"
  • "Show the diff, not the full file"
  • "Give me three alternatives with trade-offs explained"

Use examples​

Examples are one of the most reliable ways to shape output. Show the model a sample input and the expected output:

Convert these function names to snake_case:

getUserById → get_user_by_id
fetchAllProducts → fetch_all_products
parseISODate → ?

Iterate, don't restart​

If the first response isn't right, refine it in the same conversation. The model retains context from your previous messages, so use it:

  • "That's correct, but use const instead of let"
  • "Avoid the any type; use the User interface I showed above"
  • "Rewrite the error handling using a Result type instead of try/catch"