I need this to do serial fiction with ChatGPT
We have to be careful not to confuse ChatGPT’s ability to reference saved memory with its lack of ability to access chat content. As far as I understand, it cannot access separate chats, even within Projects. It can, however, pull from saved memory, providing insights and content.
It can reference information within the project, however you need to add that into the instruction. Here is the response it gave me when asked about its functionality within Projects:
“I can only access and use the information provided in the current conversation’s instructions or context. If there’s any specific information you want me to relate to or understand, it needs to be included in the project’s instructions or explicitly mentioned here.”
It can also access data that is uploaded to the project space. So lets say you were developing a story, and wanted it to refer to you previous work to extrapolate on. In this case you would summarise that work in its specific “Chat”, save it as a word doc or PDF and upload it to the project. However you will need to ask it to refer to that document in the prompt.
I just started using projects today, and for me it seems to be able to cross-reference information from one chat to another. What I loved about the project was that I able to organize various chats all referencing a group of files I uploaded to the project. For me it seemed that they were able to track what progress I made in other chats within each individual one.
To Tyrywey (original post), and you’re specific example. You may have some success with it if you created a project named “Health Tracking” or “Fitness Convos” or something like that. And if you have files documenting diet, or tracking your workout, then upload those files to the project. Then any chat you create within the project workspace can individually access and view information on any of those files. You may have tried this already, but if not it may work a little closer to what you are requesting. (I’ll attach a screenshot if it helps explain what I am trying to say lol)
But overall I agree, I feel like all chats should be able to cross-reference each other and access info from other chats. It would make the user experience as a whole far more seamless and smooth.
Yay, big progress! You can now use different models without having to move a chat out of the project folder.
1 Like
I agree completely I thought the entire point was to be able to cross reference within the project folder itself. Before beginning a project within a project folder, I actually asked Chet GPT about it and it told me that it would be able to reference chats/threads within the same project folder, unlike Threads in the normal section. As far as I can tell, you can’t do this or I just haven’t figured it out seems. I’m not alone…
2 Likes
I wholeheartedly support this. It’s absurd that ChatGPT has LLMs that are literally billions of records long, with access to search all of humanity’s digital information. With deep research, we’re talking billions, if not trillions of data points that it can reach out to and catalog answers. However, it can’t look locally in my own chats to consolidate past history and information, so I have to repeat myself. I can also look for trends over time, but it’s absolutely insane that it doesn’t have this feature.
As a point of comparison, Perplexity’s Spaces feature allows for cross-referencing data across threads and files within a shared ecosystem, enabling collaborative analysis and trend identification. Since I’m a pro user on that platform, I’ll keep my Spaces and projects dedicated to Perplexity and use ChatGPT for more simple requests that don’t require cross-referencing in my catalog.
I’ve found a workaround for this issue in ChatGPT quite a while ago. Whenever I go into ChatGPT, I search for chats on the same topic I want to discuss. Then, I go back into that chat and continue the conversation, allowing it to reference the history in the chat. For example, I’ve been suffering from shingles for the past four weeks and have been cataloging my progress, generating doctor’s appointments notes, and other relevant information. I’ve done everything in one master chat over the past four weeks, and it’s been incredibly valuable.
1 Like
Now they do have memory and Reference chat history across chats, but…
It’s at the account level, not the project level, which I don’t think is ideal.
I would appreciate it if OpenAI could have project-level memory or project-level Reference chat history.
3 Likes
I find this prompt useful (looks like a “hidden feature”):
now memorize <all important parts | something specific | etc> (I want to refer it later in other chats and you must accurately retrieve the sufficient context so you can accurately reconstruct this knowledge). Memorize all and sufficient required memories in order to do so and then list them in the response.
You can simply ask it to use them in another chat or reconstruct the context based on them (use the memory as a buffer to transfer information between chats 
)
Challenges I faced were having different projects with different files and projects not related, and during chat, the LLM was ready with a file from another project. That is insane.
Problem
Chats are ephemeral working memory, while project documents are long-term knowledge. There is no native way to promote finalized decisions or content from a chat into project knowledge.
Why this matters
Projects become fragmented. Users must manually copy and paste important decisions, specs, or summaries into documents, which discourages maintaining a single source of truth.
Proposed solution
Allow users to highlight a portion of a chat and choose:
-
Save as new project document
-
Append to existing document
-
Auto summarize before saving
This would function similarly to committing changes in source control.
Who benefits
-
Product and engineering teams
-
Startup founders
-
Researchers
-
Long-running project users
-
Anyone using Projects seriously