"Feature Request: Improve Project Memory, Context Retention, and Performance in ChatGPT"

Hello OpenAI Team,

I have been using OpenAI products, especially ChatGPT, for almost three years, and I have also used the premium version of ChatGPT during the past year. I must say that the experience has been truly impressive. The ability to complete complex tasks through a single prompt and get assistance across different workflows has made my work much easier and more efficient.

However, I would like to share some feedback regarding the user experience, especially when working on long-term projects.

One issue I have noticed is that when a conversation becomes very detailed and reaches a deeper level, ChatGPT sometimes starts losing track of important project details, previous instructions, and decisions that were already discussed. This can lead to inaccurate responses, misunderstandings, and extra effort for the user to explain the same information again.

Another challenge is that a single long conversation becomes slow over time. As chats grow larger, loading, scrolling, and continuing the conversation can become frustrating due to delays and performance issues.

I believe the project feature has great potential, but there is still room for improvement. For example, if a user creates a project named “XYZ” and provides detailed instructions, goals, preferences, and working methods inside that project, the AI should maintain a strong understanding of that project across all new chats created within the same project. Currently, starting a new conversation inside the same project can sometimes feel like starting from zero, where important context is missing and the discussion can move in the wrong direction.

My suggestion is that OpenAI should improve project-level memory and context management so that:

  • Project instructions and goals remain consistent across all conversations inside a project.
  • Important decisions and previous discussions are automatically remembered within the project.
  • Long conversations can be optimized without affecting speed and performance.
  • Users do not need to repeatedly explain the same background information.

Despite these challenges, I want to highlight that ChatGPT provides one of the best AI experiences available today. The technology, capabilities, and overall impact are truly remarkable. Improving these few areas would make the experience even more powerful and create a smoother workflow for users who rely on ChatGPT for serious projects and daily work.

I kindly request the OpenAI leadership and development team to consider these improvements and work towards making project-based conversations more intelligent, consistent, and user-friendly.

Thank you for building such an incredible product.

Have you enabled local memory on your machine? I saw such an option on the ChatGPT app (running on Ubuntu 2604.1), and I enabled it. ChatGPT seems to keep track of details from prior discussions spread of many chats. Caveat: I am a casual, non-professional user.

Thank you for your response. I think there might be some confusion regarding my feedback. I am not referring to the local memory option available in the ChatGPT app or device-level memory settings.

I am talking specifically about using ChatGPT through a web browser, especially when working on larger projects. Even when a project is created and detailed instructions, goals, and context are provided, starting a new chat inside the same project does not always maintain the complete understanding of previous discussions. The AI sometimes loses important details, decisions, or the direction of the project.

My concern is more about improving project-level continuity and context management in the browser experience, so users can work on long-term projects across multiple chats without needing to repeat the same information again and again.

I appreciate your point about memory, but the issue I am highlighting is related to maintaining consistency and context within projects, not local memory on a device.

Thanks! Sorry, I misunderstood. You are a more advanced user than I am .

No worries, thank you for your response. I appreciate your feedback too.

I’ve run into something very similar with longer research projects.

At first I also expected Projects to solve most of the continuity problem. I started keeping a kind of master research ledger in the Project files so new chats had a shared reference.

That helped, but as the research kept changing, maintaining the current state became another problem — I had to keep updating the shared file whenever important decisions, directions, or open questions changed.

Eventually I started keeping that evolving state outside the individual chats instead — Git for some projects, Google Drive/shared documents for others — and letting the AI help maintain it there. That made starting fresh chats much easier because there was a current external state to recover from, rather than relying only on conversation memory.

Have you tried anything like a master project document or external state as a workaround, or are you mainly relying on Project memory at the moment?

Yes, I understand your point. Keeping a master document is a good workaround, but I believe ChatGPT Projects should handle this automatically so users can focus more on their work instead of managing context manually.

Not that ideal. In a ChatGpt project, all new threads/branches can read same project source files. however, the Chats can’t update the project source file. It means, if one chat does something and needs to update masterledge and tried to share with other chats, you still needs to manually copy that update files to the project source files. That could exhausting especially when you have several threads working at the same time and making good progress. So I tried to use git to help, it has version control and conflict solving abilities, which make things much easier.

You can use some memory tools as an external data source. That way, they keep the memory, and you can plug it into any tool.

Projects can use shared instructions, files and conversation context: https://help.openai.com/en/articles/10169521. We understand the request is for more reliable automatic continuity and smoother long chats. We’ll pass that feedback along; no timeline to share.