I’m a longtime, paying ChatGPT user who uses the platform extensively for creative writing, interactive storytelling, character development, and maintaining multiple long-running fictional universes.
I love what ChatGPT makes possible creatively, but there are two persistent problems that significantly interfere with my experience: abrupt conversation limits and unreliable long-term continuity.
1. Please give us a visible conversation-length meter or advance warning.
I’ve had multiple lengthy, ongoing story conversations abruptly reach their maximum length without any warning.
In these instances, the assistant’s final response appeared on my screen, but after navigating away or refreshing, that response disappeared permanently from the visible conversation history.
These aren’t disposable conversations. They’re ongoing stories involving months of character development, established relationships, and complex continuity.
A simple indicator showing approximately how close a conversation is to its limit would make an enormous difference.
Ideally, I’d love to see:
-
A visible conversation-capacity indicator or approximate percentage.
-
Warnings at milestones such as 75%, 90%, and 95%.
-
An option to initiate a new conversation while preserving the current conversation’s context.
-
A reliable way to ensure the final response is saved before a conversation becomes unavailable.
I understand that conversation limits are necessary. What I find frustrating is reaching them without warning or an opportunity to prepare.
2. Please improve continuity across long-running conversations and Projects.
I maintain multiple interconnected stories, some with alternate timelines and different versions of the same characters.
These stories have extensive established histories, including relationships, character development, timelines, locations, and even the whereabouts and significance of individual objects.
Unfortunately, I frequently have to interrupt my writing to correct forgotten details, inconsistent character behavior, and continuity errors.
Sometimes correcting one behavior causes the assistant to overcorrect in the opposite direction.
Projects should be an ideal solution, but my previous experiences using them for interconnected stories were frustrating because continuity between conversations was not sufficiently reliable.
I’d love improvements that allow users to:
-
Establish authoritative canon documents that take priority over exploratory conversations.
-
Reliably reference specific details from other conversations within a Project.
-
Track character relationships, timelines, locations, and important objects across sessions.
-
Distinguish alternate timelines and discarded ideas from established canon.
-
Preserve character voices and relationship development without repeatedly resetting established dynamics.
-
Seamlessly continue a story in a new conversation without having to reconstruct its history manually.
3. Please make conversation preservation and handoffs easier.
I’ve had to develop my own methods for preserving lengthy conversations, including exporting account data, extracting conversation records through Chrome Developer Tools, converting transcripts into readable archives, and manually preparing continuity documents for new chats.
These methods work, but they require significant time and effort.
I’ve even found myself using my limited ChatGPT Work allowance primarily to preserve continuity rather than to create new material.
An official feature to archive a complete conversation, export it in a readable format, and carry its context into a new chat would be invaluable.
Why this matters
ChatGPT has become an extraordinary creative tool, especially for people developing long-form fiction and interactive stories.
But the more invested we become in these worlds, the more disruptive continuity failures and abrupt conversation endings become.
I don’t expect perfect memory or unlimited conversation lengths. I’m asking for better tools to manage the limitations that exist.
I want to spend my time writing stories, not repeatedly reconstructing them.
I genuinely hope these improvements are being considered, because they would make an enormous difference for users like me who have built extensive creative projects around ChatGPT.
Thank you