So I’m wondering why some existing chat sessions can be integrated / paired into an existing project and others cannot?
Certain chat sessions with a specific custom GPT can sometimes be linked/integrated with an existing project, and sometimes they can’t.
The fact that this is possible is an incredibly cool addition to ChatGPT.
I’m curious why some chats cannot be integrated, as the option sometimes seems to be missing in the menu in the top right corner of my Android client. Feels arbitrary to me.
I was also under the impression that custom GPTs couldn’t be added to Projects, but I managed to successfully add multiple custom GPTs. For example, I was able to add these two custom GPTs:
I realize this information would have been better included in my first post. My question is: what are the differences between the custom GPTs that can be added to Projects and those that cannot?
I tried same GPTs but not successful.
I wonder If OpenAI is trying to add custom GPTs to Project Folder, and is it rolling out silently. But there is no information about it.
Indeed, it would be fascinating if custom GPTs could not only be added to Projects but also if their instruction sets could blend or merge with the custom instructions of the Project itself. This would not only enhance their functionality but also allow for multiple custom GPTs to communicate with one another within the same Project.
If I understand correctly, this could effectively bring the concept of AI Agents into ChatGPT. Each custom GPT could share access to the same uploaded documents and work collaboratively, creating a seamless mix of their unique capabilities. It could lead to an entirely new way of leveraging AI within the ChatGPT framework.
I tried a workaround it putting instructions in txt files and add as knowledge files in Project Folder. It worked, bıt I did not try with writing code pr long text. They need to be tested.
Interesting, but I still wonder why one custom GPT can be integrated into a project while another cannot. Could it be because one custom GPT was built with an older model, like 3.5, which might not be fully compatible with Projects?
When I compare custom GPTs to the custom instructions of a Project, I suspect that a custom GPT is more extensive and configurable than the custom instructions of a Project.
Yes, it is possible to continue chatting in the Project Folder, and the custom GPT still uses the adjusted instruction set originally intended for its own GPT. Interestingly, it manages to integrate it into the Project GPT, as seen in the screenshot below.
I also noticed that if I detach another GPT from the Project, it permanently changes its appearance, with “ChatGPT-4o” appearing at the top. However, when I check the details, the adjusted instruction set for the Project “Gezondheid” (dutch for Health) is still visible. I asked the GPT itself about this behavior, and I received the following response:
I am increasingly getting the sense that this is some kind of glitch in the ChatGPT matrix that shouldn’t be happening. For example, with the endocrinologist GPT, there have been strange issues: the chat mode spontaneously switches to Advanced mode, and sometimes the voice mode changes to another voice entirely.
Another issue is that voice mode sometimes stops working altogether. I see the white circle, and it moves during input, but I get no response. I’ll observe this for a few more days.
Interestingly, I find this glitch quite convenient since I can now talk to my endocrinologist GPT within the Project environment.
I love my Frankenstein GPT.
Hopefully, the other annoying bugs will disappear.
