nah, this is just what LLM AIs do. You used to get this all the time when working with earlier versions like gpt2. Turns out when you really confuse an ai that predicts the next word, its goes off into its own little world, creating fragments of meaning that sound like lost parts of a larger conversation.