Debugging should not become a therapy session

Debugging starts with reproduction. If you can reliably reproduce the issue, you are already halfway there. From there, you can form a hypothesis, test it, and work towards a fix. LLMs can be a real catalyst in that process. Most of us remember the first time we described a bug to an LLM, and it pointed straight at the line that caused it. It felt like magic.

With LLMs, another pattern occurs. The LLM is convinced it found the bug. It explains the cause with confidence. You ask it to fix the issue, only to find out the fix does not work. You tell the model it failed; it apologises, then suggests the next plausible fix.

The model listens patiently; debugging becomes a therapy session. It validates the problem. It gives you a fluent, supportive, plausible answer. It feels like progress because the conversation keeps moving. But real debugging is not a conversation. It is a loop: form a hypothesis, gather evidence, test the hypothesis, draw a conclusion.

The model never gets tired or annoyed. It will always support you as you try one more thing. That makes it tempting to keep talking instead of doing the uncomfortable work: reading logs, checking traces, reproducing edge cases, reviewing recent changes, and proving what actually happened.

Use the LLM to do that grunt work faster. Let it help you inspect logs, compare diffs, list possible causes, and sharpen your hypothesis. Do not start fixing just because the answer sounds plausible. Start fixing when you have evidence.