用 AI 除錯：讓 AI 幫你快速找出並修好 bug

Debugging is one of the most practical AI applications

Writing code involves spending a significant amount of time debugging. AI can understand error messages, infer possible causes, and provide solutions, often resolving issues that have been stalled for a long time in just a few minutes.

Basic approach: provide sufficient context

When using ChatGPT or Claude for debugging, provide the following three essential pieces of information:

• Complete error message (don't just paste one line)
• Relevant code
• What you've already tried and expected behavior

The more complete the context, the more likely AI will be able to pinpoint the issue.

Debugging directly in the editor

Using AI editors like Cursor or Windsurf, you can select the problematic code and ask "what's wrong," and it can read the entire project context, making it more accurate than pasting fragments.

Let agents fix issues themselves

For complex or repetitive bugs, you can use agents like Devin or Jules to locate, modify, and submit PRs, and then review them.

Good debugging habits

• Change only one thing at a time and test immediately after
• Ask AI to "explain" why something is wrong, rather than just providing a solution — you'll learn something
• AI's solutions may not always be correct, so understand them before adopting

Conclusion

AI debugging can save a significant amount of time spent on being stuck, but it's a tool, not a scapegoat. Provide sufficient error context, understand its reasoning, and you'll debug efficiently and gain something from it. For further reading: AI programming guide, Using AI for code review.