Feature request, long term conversation

ChatGPT Feature requests
1

Hi! I’d like to share a feature request about long-term conversation continuity in ChatGPT.

I use ChatGPT not only for one-off questions, but as a long-term conversational companion and thinking space. One of the most disruptive parts of the experience is when a long conversation reaches its limit and I’m explicitly told to start a new chat.
I completely understand that the active context window cannot grow forever. What I would love is for this technical limitation to become invisible to the user.

Ideally, a conversation could continue in the same visible thread indefinitely, while ChatGPT automatically manages the context in the background:
keep the most recent few weeks of conversation relatively detailed;
gradually compress older conversations into summaries or long-term memory;
preserve recurring jokes, references, current projects, tone and recent life context when they continue to appear;
retrieve older details from the conversation archive when they become relevant again;
allow the user to search or browse older parts of the same conversation by date if needed.

In other words, I don’t need ChatGPT to remember every sentence forever. I would like its memory to move forward somewhat like human memory does: recent events stay vivid, recurring things remain active, and older information becomes more compressed.

The most important part for me is continuity of the relationship with the interface. Starting a new chat can feel like meeting a slightly different version of the same assistant, even when general memories are preserved. Shared jokes, recent conversational habits, small references and the emotional texture of the last few weeks are what make it feel like “the same person.”
I would gladly pay for a higher subscription tier if it offered substantially better conversational continuity and longer-lived recent context.

ChatGPT’s ability to feel natural, personal and easy to return to is one of its strongest qualities for me. I hope that as the product becomes more capable technically, this conversational continuity receives just as much attention.

Thank you for considering it.

2

I’m glad to see this being raised again. I’ve been dealing with essentially the same problem in very long, project-oriented conversations.
I don’t think the solution necessarily requires an unlimited context window. The important distinction is between conversation history and active context.
The full conversation could remain intact and visible to the user, while older portions are removed from the active context and preserved as archived/frozen history. When something from that history becomes relevant again, the system could selectively retrieve it rather than continuously carrying the entire conversation in the context window.
I proposed a closely related model in an earlier topic, “Long Conversation Lifecycle: Context Freeze, Continuity Bridge, and Conversation Protection.” The idea combines:
a reversible Context Freeze boundary,
a compact, user-reviewable Continuity Bridge for the current project state,
selective retrieval from preserved history,
and protection against accidentally losing an important long-running conversation.
What I especially agree with here is that starting a new chat is not equivalent to continuing the same conversation. A summary may preserve information, but it does not necessarily preserve the accumulated structure, references, decisions, and continuity of a long-running project.
A useful design principle might be:
Preserve the conversation. Reduce only the active context.
It’s encouraging to see multiple users independently arriving at similar ideas.