Is there any reason to think that LLMs have the introspection ability to be able to answer your question effectively? I just default to having them provide a summary that I can use to start the next conversation, because I’m unclear on how an LLM would know it’s losing the plot due to long context window.