I’m a GPT-4 Pro user who relies heavily on memory to support complex, evolving conversations across professional, emotional, and strategic topics. I’ve found the memory feature incredibly valuable — but also increasingly difficult to manage due to the current limits.
Right now, memory fills up quickly, and manually curating individual entries is both time-consuming and emotionally taxing. The lack of tools to batch-edit, merge, or even expand capacity makes it hard to maintain continuity without clutter or loss of nuance.
I’d love to see future features that offer:
Expanded memory capacity (especially for Pro users)
Auto-merging or smart suggestions for duplicates
The ability to delegate memory management to ChatGPT
Tagging or categorizing memory entries (e.g., personal vs. work)
I’d be happy to take part in any beta testing for memory improvements. Thank you for building such a powerful product — and for considering how memory could better serve those of us using it to manage complex lives.
Exactly the same experience. I use ChatGPT to brainstorm my life, to self coach, gain self awareness and to prompt new ideas and action, it is essential that it remembers nearly everything to be impactful. As a paying customer I would expect more memory, I see many have complained of the same issue, with no response as yet to address it?
Agreed, it’s my basic reason for coming to the forum. The memory seems very small at the moment especially when brainstorming. What do I want it to forget? None of it! Running out of memory is clearly a major usability issue.
I found that if you export all your chats and give it the JSON file when asking it to research, it is like it remembers literally everything! Nice little hack!
Saved Memory is a powerful feature—I’ve enjoyed using it. However, one thing that would make it even better is the search function. Searching by keyword would make it much easier to find specific memories quickly, especially when I want to review or delete something I no longer need.
Thank you for putting this into words — I fully agree with everything you’ve described. I’m also a GPT-4 Pro user and rely heavily on memory for layered, evolving conversations. In my case, this includes both creative work (long-form storytelling across sessions) and structured self-reflection, which I’ve been using ChatGPT for in a way that sometimes resembles interactive journaling or even a kind of therapy support — based on decades of psychotherapy experience.
Memory has made all of this possible — and at the same time, increasingly difficult. Like you, I find that memory fills up too quickly, and I have no real insight into what gets overwritten or why. That lack of transparency makes it hard to build continuity, especially when you’re working on multiple long-term topics or returning to something deeply personal.
I especially want to echo your call for tagging and categorization. If we could mark entries as “permanent”, “project-based”, or “temporary”, it would go a long way toward making memory manageable and meaningful. Right now, I often feel like I’m re-teaching ChatGPT things that I had already shared — but which silently got lost.
This isn’t a complaint — I genuinely appreciate what memory already enables. But I do think it needs a next step: more user control, more transparency, and some form of prioritization that reflects the complexity of how many of us use ChatGPT today.