@uncreativetom really? I thought (at least OpenAI) had an actual blog post on it somewhere and proper mechanisms such that it was actually less bad than a lot of people thought by some meaningful amount… I’m still looking for it