It's nearly unusable. Unnecessary, nit-picky corrections that derail the discussion. Too bad claude keeps rate limiting. The jokes really do write themselves. Pasted this thread into ChatGPT, and it says: But here’s the important distinction (this is the part Reddit doesn’t separate cleanly) What’s happening is not: the model thinks users are wrong by default It’s closer to: the model is biased toward making every response useful in a “review/analysis” framing sub...