This is a special post for quick takes by 849. Only they can create top-level comments. Comments here also appear on the Quick Takes page and All Posts page.
It seems that when I vibecode a tool for long enough (>month), I end up with a lot of tests that measure the performance of the tool in various situations. I am wondering if these tests are actually the best way to improve the tool or if the LLM is creating graders that they can optimize for as a kind of compulsion