First of all, a small thank you to OpenAI and to the people who keep this community going. I’m simply a ChatGPT user, and this space has allowed me to discover many ideas and also to share some of my own experiences.
While recently sharing a personal project, I realized that there isn’t really a dedicated space for users who experiment with personas, roles, and conversational architectures.
These topics are somewhat scattered across use cases, prompt engineering, and feature requests.
So I was wondering whether it might be interesting to create a category — or possibly a dedicated tag — for personas and conversational architectures.
For example, people could share custom personas, role systems, different ways of collaborating with ChatGPT, specialized perspectives, or conversational architectures.
Above all, this would make it easier to discover what other users are experimenting with in this area, exchange ideas, and compare different approaches.
Of course, this is only a suggestion.
Would this kind of space be of interest to other people here?
I think you have a dedicated GPT’s section which you can use to customize for anything you like depending on your choices , and it can be useful for the kind of experiment you are mentioning about. I think we don’t really need a separate section for it . But Hey, its my viewpoint, perhaps someone else can have more better opinion which might justify.
I see what you mean about the dedicated GPTs section. What I had in mind was slightly different, though.
I was thinking less about sharing complete custom GPTs, and more about sharing and discussing the conversational architectures themselves: personas, role systems, specialized perspectives, ways of structuring collaboration with a single AI, and experiments that can sometimes be used across different assistants rather than belonging to one particular GPT.
But your point is useful — perhaps a dedicated tag would make more sense than an entirely separate category.
This request doesn’t have a forum destination because such thoughts will not be entertained by AI models.
OpenAI is going the way of Claude. The AI will not comply with requests that it perform a character role, and has internal reasoning that considers every input run of any role of messages suspect. The language models already in years past had a threshold where you could perceive if a persona was believed or if the output had a radical shift in its behavior of merely being acted out. Now there is just denial and refusal.
You get to interact with the entity “ChatGPT” and no more, even when it is assigned the unmistakable job of being an actor or specialist that doesn’t break character.
I understand your point, but I think we may be using the word “persona” in two different ways.
I’m not necessarily talking about convincing the model that it is another entity, or asking it to believe that it is a fictional character.
A persona can also be an interface.
The underlying model remains ChatGPT, but the interface can define how we interact with it: a particular voice, a perspective, a specialist function, rules for collaboration, or even several complementary functional viewpoints sharing the same context.
In my own experiment, for example, I don’t treat the personas as separate AIs or independent identities. They are different functional frames inside one AI: analysis, investigation, exploration, practical reality, anticipation, orientation, etc. The interesting part is not whether the model “believes” it is those characters. The interesting part is how those interfaces change the way a human collaborates with the same model.
So even if models become less willing to perform immersive character roleplay, I think there is still a much broader subject here: conversational interfaces, functional personas, role systems, and architectures for human-AI collaboration.
And perhaps that distinction is exactly why I was suggesting a dedicated space in the first place.
