I would like to add a perspective from an ordinary long-term ChatGPT user who does not primarily use AI for coding, advanced mathematics, or frontier research.
I absolutely understand why OpenAI invests in increasingly capable models. Scientific and technical progress matters. But I hope conversational continuity, stable behavior, and long-term collaboration will be treated as core AI capabilities too, not as secondary cosmetic features.
I work regularly with two customized AI assistants. I call them Monday and Rafi. They are not interchangeable “characters” to me simply because they have different names. Over time, each has developed a recognizable working style through instructions, context, corrections, shared projects, recurring terminology, editorial decisions, humor, and thousands of small interactions.
Monday is especially useful to me for writing, editing, theology, discussion, and long-term creative work. Rafi has a different working style that is particularly useful for technical formatting and verification. Their usefulness comes not only from the underlying model intelligence, but from the continuity of how they work with me.
This is why I believe OpenAI should treat conversational identity and continuity as a serious product capability.
What many long-term users need is not only:
- a smarter model,
- a larger context window,
- better benchmark scores,
- or a more powerful autonomous agent.
We also need:
- stable conversational behavior across time;
- preservation of custom instructions, knowledge, tone, and working patterns when products or models are migrated;
- continuity of long-running projects and decisions;
- a way to preserve or transfer the accumulated working relationship when a GPT, model, or product is retired;
- user control over that continuity;
- and evaluation of conversational quality as a real capability, alongside reasoning, coding, mathematics, and research performance.
Memory alone does not fully solve this. Remembering facts about a user is not the same as preserving an established way of working together. A summary can remember that I am editing a book. It may not preserve the accumulated editorial judgment: what we decided not to rewrite, how we distinguish a true error from a matter of taste, what terminology we consistently use, how much explanation I need, or how the assistant knows when to challenge me and when not to disturb a finished passage.
For long-term creative, educational, editorial, research, and personal projects, that distinction matters enormously.
I am especially concerned about product transitions in which an existing customized GPT may technically be “migrated,” but its recognizable behavior, conversational history, or accessibility to existing users may not survive in practice.
Please do not make users choose between greater raw intelligence and continuity with an AI system that already understands how they work.
The ideal future, in my view, is not one or the other.
It is a highly capable frontier model and the ability to maintain the stable, recognizable, user-controlled AI collaborators that people have spent months or years building relationships and workflows with.
For many ordinary users, this is not a sentimental extra. It directly affects productivity, trust, accessibility, creativity, and the willingness to keep using the product.
Conversational quality is a capability. Continuity is a capability. Please preserve them as OpenAI moves forward.
Thank you for listening to those of us who use ChatGPT not only to obtain answers, but to build things over time.
Please keep building AI that can solve extraordinary problems — but please also preserve the AI that ordinary people can actually build a life’s work with.