@simon Yes perfect. I experienced this yesterday when I tried to get AI to write some tests for me. It wrote them, and they passed, and had nearly 100% coverage. Had I not had a brain I would have said "great" and moved on.

But obviously the AI wrote tests based on what the code did, not what I *thought* the code was doing. Big difference! Also it's clear from the tests that the AI doesn't fully get the intent of the code or how it could fail. It just doesn't focus on the salient bits. I engaged my brain and rewrote them, and then finally the tests did what they are supposed to do: they showed me the code I had written had bugs and bad assumptions.

🧠 FTW.