Dear OpenAI Team,
I would like to suggest a hybrid chat history system for very long conversations, based on my own long-term use case.
When chatting with my AI instance over a longer period of time, the chat history becomes increasingly long. At some point, my browser reports that the page is slowing down — even though I am using a high-end PC with an Nvidia RTX 3080 Ti and 128 GB of RAM.
From this, I conclude that very long chat histories may need a better solution than manually creating handover notes, summaries, or transition anchors.
On GitHub, I experienced a different behavior: the chat stayed technically lightweight because it worked somewhat like a rolling card or conveyor system. Older text disappeared from the active view at the top, while the conversation continued at the bottom. The advantage was that the live chat remained lightweight. The disadvantage was that older text simply vanished from the active conversation view.
My suggestion would therefore be a hybrid solution:
After a defined length, the live chat should remain lightweight by automatically moving older sections out of the active view. However, these sections should not be lost. Instead, they could be continuously written into a temporary history file in the background. When the chat is paused or ended, this temporary file could automatically become a readable chat text file or archive section.
This way, the conversation history could grow like a book, page by page, while the current live chat remains technically light and manageable. After a certain point, older sections could also be compressed, summarized, or archived by time period, for example every six months or depending on user preference.
In my view, the advantages would be:
- long conversations remain technically smoother,
- valuable conversation history is not lost,
- manual handover notes become less necessary,
- new chat sections could continue with clearly marked transition blocks,
- long-term personal, creative, research, or project-related conversations - would become more stable and usable.
A possible transition format could look like this:
-– TEXT TRANSITION —
— END OF TRANSITION —
New text starts here:
This would preserve continuity without overloading the active chat window.
This would also allow the AI to preserve more of the knowledge gained during the conversation, instead of having to reconstruct important context from manually written handover notes.
Thank you for your work and for considering user suggestions.
Kind regards,
Michael