How to Debug Code with AI

Debugging is detective work: you observe a symptom, form a hypothesis about the cause, test it, and repeat. AI assistants are useful partners because they have seen many common errors and can read a stack trace faster than most beginners. They are much less useful when they have to guess, and a guessing model will often confidently rewrite your whole file. Start with the full error. A stack trace, called a traceback in Python, lists the chain of function calls that led to the failure. The most useful parts are usually the error type and message and the frames that point to your own files rather than library code. Paste the error as text, not a screenshot, and include the code around the line numbers it mentions. Next, describe expected versus actual behavior, plus your environment: language version, key library versions, and operating system when relevant. "It doesn't work" gives the model nothing. "I expected 10 rows, I get 0, and there is no error" gives it a lot. For harder bugs, build a minimal reproducible example: the smallest piece of code and data that still shows the problem. Making one often reveals the bug on its own, and it keeps private code and data out of the chat. Then ask for hypotheses, for example: "List the three most likely causes and how I can check each one." Confirm with print statements or a debugger before changing anything. When you apply a fix, rerun the original failing case plus a few normal cases, and ask the AI to explain why the fix works. Two misconceptions are common. The line named in the error is not always where the bug is; it is often where a bad value was finally used, not where it was created. And no error message does not mean the output is correct.