GP is making a joke about speaking to oneself really just being the human version of Chain of Thought, which in my understanding is an architecural decision in LLMs to have it write out intermediate steps in problem solving and evaluate the validity of them as it goes.