for some reason i got a bug where every time i want chatgbt to create another message the count part is broken, no matter how many different messages i tell chatgbt to create it can only pick two of them. to put it simply no matter how many messages it generates i can only scroll between “0” and “1” the rest of the messages i cant access at all.
did you find a way to fix it?
Unfortunately no, but there are some people with the same problem hoping that they will solve the problem. For now, all we can do is wait 
i also have the same problem and i can’t also edit my prompts
I have the exact same problem. It says that, and won’t let me see the previous versions, and I can’t edit the prompts unless I quickly press the little pencil button when it appears when I hit the refresh button. And it’s affecting all of the new chats I create.
In the pinned topic from yesterday that discusses this:
When they will fix this problem?! It’s beyond ridiculous with their 0/2 errors! 2 days passed since it started occurring! Does somebody knows a solution or manual ways to fix it?
Support posted two hours ago that the bug got reported internally: See Here
does anyone also have the problem that they can’t edit their first prompt in a new chat, which is anoying
I have the same problem too. Time for use to spam their emails until they have no choice but to acknowledge there’s still a problem.
Yes, I have the 0/X bug for at least 3 days the same
i hope they fix this really soon because i always need to make a new chat when i made a fault in my first prompt
I think they might’ve fixed it a bit. Because mine is back to normal now. I think since enough of us were experiencing this, they worked on it.
Subject: Persistent Message Loss & Corruption Since GPT-5 Rollout
We’ve conducted structured testing across multiple environments and can confirm consistent message instability following the GPT-5 model rollout. Testing environments include:
Android mobile (S10 and newer)
Ubuntu-based sandbox system
Desktop (browser and native app installations)
Voice-to-text input modes
Key Observations:
Threads display frequent corruption errors, such as:
“There was an error generating a response.”
“Message could not be loaded.”
Messages disappearing after thread exit.
Mobile-specific behavior:
Messages sent via mobile often fail to persist across devices.
Inconsistent visibility when switching between mobile and desktop.
Messages typed or dictated via voice may appear briefly, then vanish when the thread is reopened.
Desktop behavior:
Messages generally persist, but corrupted segments still occur.
Performance is noticeably better, but still affected.
System-wide lag:
Noticeable delay in loading threads and generating responses.
GPT-5 responses feel rehearsed, shallow, or incomplete compared to GPT-4o.
Recovery Behavior:
Switching back to GPT-4o (Legacy model) immediately improves performance.
Lost messages sometimes reappear when legacy mode is activated.
Logical coherence, speed, and stability all improve on GPT-4o.
We’re continuing to test across models, but initial results suggest the instability is tied directly to the GPT-5 rollout.
Has anyone else noticed message retention or corruption issues specifically tied to the GPT-5 model? Would appreciate any confirmations from other developers or testers.