I am sharing this publicly because I believe it raises a serious question about ChatGPT’s future reliability.
I am a paying ChatGPT Plus subscriber in Australia, and I use ChatGPT intensively every day for software development, music, publishing, research, and managing long-term projects.
I am not simply asking occasional questions. ChatGPT has become part of my daily working environment.
I believe more people will eventually use AI this way.
That is why my experience should matter beyond my individual account.
I started contacting OpenAI Support in August 2026 about problems affecting heavy, long-term users. I repeatedly reported storage limitations, problems with long conversations, unreliable workflows, and other usability issues.
I provided screenshots, detailed explanations, troubleshooting results, and technical information.
By October, my experience had deteriorated dramatically.
The ChatGPT iPhone app became extremely slow, began crashing repeatedly, and eventually became unusable.
I now have to use a web browser just to continue communicating with ChatGPT.
I also submitted an actual iOS diagnostic showing that the ChatGPT process was terminated after exceeding a memory limit.
This raises an important technical question: can ChatGPT’s mobile app reliably handle accounts containing extensive conversation histories and sustained, high-intensity work?
I am not claiming that my workload has been proven to cause the crashes. That requires a proper engineering investigation.
But I believe the combination of intensive usage, degraded performance, and documented memory-limit termination deserves serious attention.
Today, I may be an unusually heavy user. Tomorrow, many more people may depend on ChatGPT for equally demanding work.
If the product cannot reliably handle that transition, this could become a much broader problem.
What makes my experience particularly frustrating is the support process.
I tried to resolve these issues privately for weeks. I repeatedly contacted official support and provided the requested evidence.
Some cases were closed as duplicates. Many responses were automated, repeated earlier troubleshooting steps, or failed to confirm whether a responsible technical specialist had actually reviewed the case.
I have still not received a confirmed engineering resolution or a reliable way to use the native iPhone app again.
Meanwhile, my actual work has been delayed, and I continue losing time investigating problems instead of completing my projects.
I did not want to make this public. I am doing so because my attempts to obtain effective help through private support have not resolved the situation.
I genuinely value ChatGPT. I want it to succeed.
But a future in which people depend on AI for real work requires more than increasingly capable models.
It requires stable applications, reliable access to accumulated work, scalable data management, and technical support that can meaningfully investigate serious failures.
I hope OpenAI will investigate whether there are performance or memory-management limitations affecting high-volume accounts, communicate transparently with affected customers, and provide meaningful technical follow-up.
I am sharing my experience so that developers, other users, and OpenAI’s product and engineering teams can recognize this potential risk before it affects many more people.
AI should become more reliable as our work grows—not become inaccessible when we depend on it most.
— Sunny Pei-Hsin Li
ChatGPT Plus Subscriber, Australia