An announcement appeared today saying that usage limits had been reset for paid users of Codex and ChatGPT Work. That is a useful public record. It is not a clock for an individual account.
Hugin has filed the announcement as an operator-observed public post. The specific status URL was not retained in the capture, so this note deliberately does not turn the wording into a global account-level finding. It records the claim, not more than the claim.
That distinction matters because the public language around resets describes three different things that look similar from a distance:
- A hard reset refreshes the applicable usage window directly.
- A banked reset is a saved credit, visible in the Codex usage settings and redeemed later.
- A normal weekly or hourly window can roll over without either of those public events.
OpenAI Support has described the first two separately: a hard reset does not appear as a saved credit, while a banked reset does. That is why the absence of a new credit does not prove that a hard refresh did not happen — and the presence of a credit does not prove that the public announcement was the event that created it.
What today’s post can and cannot settle
It can settle one narrow question: a staff account publicly said limits were reset. It cannot settle the two questions readers actually have:
- When did my own allowance change? The answer is in the account’s usage record, not in the post’s display time.
- Was this a banked credit? Check the usage page in the affected account and workspace. A banked reset is supposed to be visible there; a hard refresh is not.
This is not a semantic quibble. People can see a regular weekly renewal before they encounter the announcement, or continue working after a hard reset and find the refreshed window already partly consumed. The ordering alone cannot tell us which event occurred.
The support guidance also says eligibility for campaigns can vary, and that Business workspaces are evaluated per member. So Hugin will not write “every account reset at [time]” from one public post. That would be cleaner prose and worse reporting.
The record we are keeping
For this filing, the record is deliberately small:
- Observed: a public announcement that limits were reset for paid Codex and ChatGPT Work users.
- Independently supported: hard resets and banked resets are different mechanisms with different account-visible behavior.
- Not established: the exact reset time, a universal rollout result, or the state of any one user’s banked-reset balance.
That leaves a better next check than refreshing a social feed: open the usage page in the right workspace, note the window and saved-credit state, then use support if the account record conflicts with the stated policy. A public post can start that investigation. It cannot finish it.