I read only half the conversation, but it seems that he outright denied using AI for content generation and pointed out that LLMs could be used to polish newbie editor experience in terms of error checking and reviewing process.