More like annoying, as some of us will no doubt run into this self-lobotomization at some point and wonder why a GPT-6 model is behaving like GPT-2 all of a sudden