Hi there,
Just a thought to put out there for chatbot continuity. I would love if ChatGPT conversations could feel like different rooms in one house. I separate chats because I want different projects and subjects organized, not because I want a different conversational partner in each one. I don’t need perfect or word-for-word recall. I want enough shared context that when I move between chats, ChatGPT can recognize relevant things we’ve already discussed and continue naturally without making me re-explain them. The room should determine what we’re working on—not erase who we’ve been talking to. I understand that the memory feature can retain basic context i.e. personality, projects and whatnot but there is no recall from something still being worked on that needs to be moved over. This is most noticeable, after filling a chat to its capacity and having to start a new one. I hope this makes sense.
This “different rooms in one house” framing is exactly right.
I think the missing piece is not just better memory, but a persistent and structured project state.
Memory can preserve who the user is and their long-term preferences. Project state should preserve where an ongoing piece of work currently stands: what has been decided, what changed, what is still open, what was rejected, and what the next step is.
Today, when moving between chats, the assistant often has to reconstruct that state from old conversations. Ideally, chats would just be different views into the same underlying project, rather than separate containers that each have to rebuild the current state from history.
Memory and Projects can help carry context between chats: https://help.openai.com/en/articles/10169521. Thanks for explaining the remaining gap—we’ll pass along the structured project-state and smoother continuity request. No timeline to share.