I was talking with a colleague the other day about the particular ways Large Language Model (llm) output gets under our skin. Neither of us was making a grand claim about the technology; we were comparing notes on an ordinary work task. He was tired of being congratulated. The model would give him perfectly serviceable code and then congratulate him for asking such thoughtful questions to prompt that output. I, on the other hand, was tired of the overly explanatory tone: the long preamble, the insistence on explaining why every small thing mattered.