I would like to suggest improving the ability to maintain long-running conversations, especially for users who actively work with ChatGPT as a collaborative thinking and feedback partner .
There is an important benefit here not only for the user, but also for OpenAI.
In many of my conversations, I do not simply ask ChatGPT for information and accept the answer. I actively correct it, provide real-world observations, test its assumptions, point out mistakes, and introduce technical details that may not have been obvious to the model. In other words, I use the conversation as a continuous feedback loop to help improve the quality of the interaction.
I do this voluntarily and without compensation because I believe that these corrections and discussions can ultimately make AI more useful for everyone.
The problem is that this kind of learning process requires time and continuity. A technical discussion can develop over many hours or days. We may start with an assumption, discover that it is wrong, perform a physical measurement, revise the hypothesis, test another solution, and eventually reach a working result. When the conversation is interrupted by a usage limit, a large part of that collaborative process is broken.
This is therefore not simply a request for “more messages.”
I would suggest creating a persistent long-running conversation or project mode in which ChatGPT could maintain a compact working context containing the important facts, corrections, discoveries, rejected hypotheses, and conclusions from the discussion.
Such a system could benefit both sides. The user would be able to continue complex work without repeatedly rebuilding the context, while OpenAI would receive more sustained, high-quality human feedback from users who actively test, correct, and challenge the model in real-world situations.
In my view, users who spend significant time correcting and teaching the model are effectively contributing valuable free human feedback. Giving these interactions enough continuity would allow that effort to produce much more useful results.
The goal should not simply be to make conversations longer. The goal should be to allow meaningful human-AI collaboration to continue long enough for both the user and the system to reach better results.