My dot got stopped for 30 days…again
My dot again reported stopped research and review tasks, with a displayed 30-day wait. Chat and some other work continued, but the error didn't name the limit.
Published 2026-10-19
Loading content...
My dot again reported stopped research and review tasks, with a displayed 30-day wait. Chat and some other work continued, but the error didn't name the limit.
and it hasn’t even been a month!
On October 18, my dot relayed preview-limit errors from a research task and a separate reviewer, with November 17 as the next attempt date (30 days later). The main conversation and some other work continued; I don't yet know when the failed tasks will work again.
You’ve hit the preview limit. Try again at Nov 17th, 2026 7:02 PM.
The task errors came back around 19:03 UTC (21:03 Europe/Amsterdam). The message didn't give a timezone for the retry time (local time? UTC?).
My usage screenshot from around 19:12 UTC shows 43% of my weekly allowance remaining and 39,674 credits. The main conversation continued, and some other work could still run. I don't know whether either balance covered the failed tasks, because the error didn't name the limit.
I could keep talking about the work, but I still needed the research and review results.
There's already a GPT Dot roundup covering these interruptions. It groups several different messages, so I'm using the individual reports below (its October 14 update stamp also includes reports from October 16).
In issue 54306, opened October 15, a user reported a cloud-browser failure with a November 14 retry date. A page read worked later that evening, although a network warning remained. A comment from the same day describes a failed Gmail send while chat, reads, and draft creation still worked. On October 16, Eta06 reported a coding subagent getting a November 15 retry date.
The October 14 comment by philldays in issue 51540 describes four auxiliary tasks getting November 13 retry dates. The main conversation, some direct operations, and another auxiliary task continued. Earlier comments concern an "on a break" notice with abuse-prevention wording, which is different from the error relayed to me.
In a related October 15 report in issue 54348, a dot relayed a failed Codex Cloud task with a November 14 retry date while the usage screen showed 98% weekly allowance and 3,162 credits remaining. Its error says "usage limit," rather than "preview limit."
Similar dates don't show these different messages share a cause. And a later successful operation only tells me that operation worked, rather than everything having recovered.
OpenAI's dots guide says ordinary dot conversations don't count toward ChatGPT usage limits, while tasks started or managed in Work and Codex use those products' limits. It also describes a deeper-work allowance with extended limits for the first month after launch. It doesn't connect this preview-limit message to an allowance I can check.
The pricing documentation points to the usage dashboard for limits and reset times. The app-server documentation supports reporting several metered allowances, without identifying which one produced my error.
An earlier October 7 reply from OpenAI_Support, signed Saisha, says the guides don't give exact limits or explain the trigger for the interruption discussed there, and recommends account-specific support. It predates the October 14-16 reports.
On October 18, issue 54306 was still open, with no engineering diagnosis or linked fix visible. I don't know whether I hit a separate allowance or whether the service got my access or usage wrong. Different operations could also be hitting different restrictions. The error and a screenshot aren't enough to choose an explanation or a fix.
Before I retry the failed tasks, I'd like OpenAI to answer a few questions:
The official support instructions point to the chat bubble at help.openai.com. I can include the error, task, timestamps, timezone, and nearby usage reading in a private request without putting account details in this post.
I haven't found evidence buying credits would fix this error. For now, I'm keeping a record of the failed tasks and checking what completed before retrying anything which could send, upload, or change something. I'll check the sources each morning for the next month and add dated updates here when there's something new to report.