So chain of thought is an awful experiment that doesn’t let you know how an AI reasons. Instead of admitting this, AI researchers anthropomorphize yet another test result and turn it into the model hiding their thought process from you. Whatever.