i agree w amanda- put claude in unwinnable contexts, expect claude to "fail". it can be useful to find out which way claude breaks but stop expecting the impossible. https://x.com/AmandaAskell/status/2084369056765989224
imo the corollary to expecting claude to never flip the bozo bit on users is to at least *try* not to act like a bozo in eval setups https://medium.com/@mike.besso/do-not-flip-the-bozo-bit-b06fb9395845