Pro tip: I stopped trusting AI code helpers to write my unit tests in Denver
Everyone at my Denver meetup keeps saying Copilot and Cursor write better tests than they do, and I flat out disagree. I tried letting Claude draft tests for a payments module back in March and it kept mocking things that did not need mocking, so I spent 3 hours fixing tests that should have taken 40 minutes. Now I write the assertions by hand first and only use AI to brainstorm weird edge cases. Anyone else find these tools better at ideas than actual finished code?