NUNX

AI tools · Coding

Use AI for coding

Get help without accepting insecure or untested code.

8 min read · General educational guidance
  1. Share the smallest reproducible problem, expected behavior, actual output and relevant versions.
  2. Ask for an explanation of the cause and a minimal patch, including tests and security assumptions.
  3. Review the diff, run tests and check official documentation before using unfamiliar packages or commands.

Practical example

Provide a failing test, the smallest related function, expected output and runtime version. Ask for the root cause, a minimal diff, a new regression test and any security assumption introduced by the patch.

Common mistakes to avoid

  • Running generated commands that delete or overwrite data without understanding them.
  • Adding a package the model named without checking the official registry and documentation.
  • Accepting code because it compiles without tests, review and threat consideration.

Trusted sources and next reading

AI can be confidently wrong. Verify important facts, calculations, citations and professional decisions.