I tell it to stop that, but it still does that sort of thing every few prompts when it loses the immediate token context window. Sometimes it’s wider continuity

I tell it to stop that, but it still does that sort of thing every few prompts when it loses the immediate token context window. Sometimes it’s wider continuity [1] will keep it off that pedantic path longer, but I still have to occasionally put into prompts things like “don’t remind me about blah blah blah every time” [1] “continuity” is what it called the ability when, after about a half dozen prompts of interrogating it, I finally got it to acknowledge that it was still aware of something outside of its immediate context window (without me giving the direct knowledge with a leading question). I don’t know if it hallucinated the term “continuity” to account for its ability to look back further than the token context window or if that’s actually what the openai folks call it, but ChatGPT said it was limited, not everything was available, and there was some type of pruning based on perceived importance. I’ll try to dig up the precise wording and post in a comment to myself here, it’s late and I’m on a mobile device though and too tired and lazy to touchscreen my way through tons of chats histories. We need a global search of prior chat sessions.