Talking about human error in the context of LLM response behavior is hilarious. "No, the machine is fine, someone must have used it wrong". Sure...