the best model in existence today, in their own product, confidently answers math questions wrong rather than just calling a tool. and we're all okay with this.

one form of LLM psychosis i've seen is a belief that LLM's are like more or less past the point of making these kind of mistakes.