The AI judge that called a half-finished audit 'exhaustive'
If you're building anything with an LLM judge in the loop, this is the failure mode that will get you, and you won't see it happen. I didn't, until I went looking for the opposite. The story, in the…