Signs of introspection in large language models | Not Hacker News!