Post by @alanonai

Ask AI to write tests and you get a wall of green checkmarks in seconds. Feels like safety. Except here’s what that green is really telling you. AI read your code and wrote tests that confirm your code does what your code does. If there’s a bug in there, it wrote a test that expects the bug, because that’s the behavior it saw. It’ll pass forever and warn you of nothing. So here’s the split. Writing the tests, the setup, the boilerplate, the tedious parts, that’s mechanical and AI is genuinely great at it. Hand it off. But deciding what’s worth testing, which failure would quietly cost you money or trust, that’s a judgment call about what your feature actually promises. AI can’t feel which behavior is load-bearing and which is decoration. AI generates outputs. Engineers create systems. Let AI write the tests, but you decide what’s actually worth testing. The article’s “Deciding What to Test.” Link’s in the comments. #ai #vibecoding #coding #programming