A ChatGPT user asks OpenAI for dependable ways to resume long projects

What the user is reporting

A post on OpenAI's community forum, headed as a formal product review, comes from someone using ChatGPT for large, long-running academic and professional work. Their main example is an Academic Refresher Handbook with multiple chapters, extensive research, repeated revisions, screenshots and document preparation.

As community.openai.com reported through this user's post, the trouble starts when work moves from one conversation to another. The author says ChatGPT is unreliable at finding, recovering and continuing earlier work, and that the problem goes beyond the ordinary kind of forgetting that chat assistants are known for.

This is a single user's experience. We have not seen an OpenAI response to the post, and nothing in it is an official finding.

The costs of losing the thread

The author lists what goes wrong when a project builds up a lot of history. These are their claims, listed here as they describe them:

• Earlier project context disappears.
• Work that was already finished gets repeated.
• Editorial decisions made earlier are applied inconsistently.
• Original source materials are hard to get back.
• Time goes into rebuilding progress that already existed.
• Trust in ChatGPT as a long-term collaborator drops.

Think of a human editor who forgot which corrections you had approved. You would keep re-explaining the same decisions, and you would stop handing over big jobs. That is the experience the post describes.

The six fixes, A to F

The user frames their requests as a framework for project continuity and recovery. In plain terms:

• A. Retrieval. A dependable way to find relevant past conversations, decisions, research and content across a user's history.
• B. Persistent context. Big projects keep their goals, instructions, progress and editorial decisions when work moves between conversations.
• C. Version history. Project-level versions, so users can find and restore earlier drafts, amendments and materials.
• D. Traceable decisions. A record of revisions, corrections and approved changes, so later work follows the same direction.
• E. Source recovery. Easier retrieval of the original uploaded documents, screenshots and research files tied to a project.
• F. Resumption. A structured way to restart a project, see what is done and what is outstanding, and continue from the last verified stage.

Version history, in particular, is a familiar idea from word processors and code tools: a saved trail of changes you can step back through. The author is asking for something similar for AI-assisted projects.

Feedback with no visible destination

The author says they sent the same formal review to OpenAI Support. Support acknowledged it but could not confirm that it would go to Product Management or Engineering. It also could not provide an escalation route or a reference number for tracking the feedback.

That gap is why the user took the issue to the public forum. They also asked other users with similar difficulties to add their experiences, and said the post is meant as product-quality feedback rather than a personal complaint.

The author does not dismiss existing features. They call Projects and the sidebar Search useful, but say the deeper need is for people to continue substantial work without manually rebuilding its history.

How to read this

Forum posts are good for spotting pain points that come up repeatedly, but each one is a single person's account. We cannot tell from this post how widespread the problem is, how often it happens, or whether OpenAI has anything planned. Nothing here confirms a product change.

It does fit a pattern in today's forum posts: another user asked for access to full chat history for project tracking. Taken together, they suggest people are using chats as project records and finding that role hard to sustain. Whatever OpenAI decides, the sensible move is to keep your own record of long projects rather than rely on the chat history alone.