Month-long unresolved ChatGPT continuity and retrieval failure — cases #13584959 and #14252673 — requesting human review
For more than a month, I have been trying to get help from OpenAI Support for a persistent ChatGPT continuity and retrieval problem that is making long-running work effectively unusable.
I am a ChatGPT Plus subscriber and use Projects and long-running conversations extensively. Beginning around August 21, 2026, continuity that had previously worked reliably began failing.
The problem is not simply that old material falls outside the model’s active context window.
What I am experiencing includes:
- Conversations abruptly reaching a hard stop with the message that the conversation is too long to continue.
- These hard stops occurring at inconsistent intervals, including one recent conversation that lasted only a few days.
- After a hard stop, information from the previous conversation remains stored in my account, but a new conversation cannot reliably retrieve or incorporate that information.
- Retrieval across and inside same conversations and Projects has also become unreliable.
- In some cases, ChatGPT can retrieve fragments of an earlier conversation but not the complete conversation.
- I am now having to request repeated account data exports because the export is the only dependable way I have found to preserve and regain full access to conversations after these hard stops.
This has caused me to close projects that depended on continuity because repeatedly rebuilding context after every hard stop defeats the purpose of the project.
I have reported this through OpenAI Support under cases #13584959 and #14252673.
Unfortunately, the responses have repeatedly failed to address the actual problem. At one point I was asked to provide information relating to Codex usage even though I do not use Codex and had never reported a Codex problem. I have repeatedly clarified the issue and requested escalation.
I have also specifically asked whether a human being has reviewed my case. I have not received a clear answer.
I am not asking for routine troubleshooting instructions about how Projects or Memory work. I have already gone through those steps extensively.
I am asking for a human OpenAI Support representative to review the existing case history and the evidence already provided, and to escalate this to the appropriate ChatGPT product or engineering team if necessary.
In particular, I would like OpenAI to help answer:
- What determines when a ChatGPT text conversation is declared “too long to continue”?
- Is there a documented maximum conversation length, and if so, how is it measured?
- Why are my conversations reaching hard stops at apparently inconsistent lengths?
- Why does information that remains stored and visible in my account become unreliable or inaccessible to ChatGPT after these stops?
- Why has continuity and retrieval behavior on my account changed so significantly since approximately August 21?
- Can someone examine the affected conversations and account behavior directly rather than continuing to provide generic troubleshooting responses?
I can provide screenshots, dates of multiple hard stops, conversation histories, data exports, and the prior support correspondence.
I would be very grateful if @OpenAI_Support or another OpenAI staff member could locate cases #13584959 and #14252673 and arrange an actual human review of the underlying issue.