And you just train LLM's on things that people wrote with understanding in response to things they read without understanding.