Large language models are very good at telling us we are right. They smooth the rough edges off half-baked ideas, turn vague thoughts into polished paragraphs, and, in the process, can quietly ...