Public Codex reset appears to have been applied immediately instead of banked

What issue are you seeing?

The most recent public Codex rate-limit reset appears to have been applied as an immediate hard reset, rather than being granted as a banked reset that users could choose when to spend.

This conflicts with the reset-banking messaging OpenAI has been giving users.

OpenAI announced banked resets as a way to save Codex rate-limit resets and use them later:

https://x.com/OpenAI/status/2065225362544726371

Tibo also described reset banking in the same terms:

https://x.com/thsottiaux/status/2065468501750649006

But the most recent reset announcement said:

https://x.com/thsottiaux/status/2066956441173323943

That appears to have reset live Codex limits directly for many users instead of adding a banked reset.

Related user reports:

• https://x.com/argofowl/status/2067372183916331205
• https://x.com/mabanham/status/2067382513929179532

If the reset was supposed to become banked, this is a bug and affected users should receive the banked reset.

If the reset was intentionally applied only as an immediate hard reset, then the public messaging was materially misleading. Users were told that resets could be saved and used on their own time; an immediate reset is not equivalent, especially for users whose quota was already near a natural reset or who were not ready to use Codex at that moment.

What steps can reproduce the bug?

1. Have an active Codex subscription with reset banking available or expected.
2. Observe the public announcement that Codex rate limits would be reset across all plans.
3. Wait for the announced reset.
4. Observe that the current quota window is reset immediately.
5. Check whether a banked reset was added for later use.

What is the expected behavior?

Publicly announced Codex resets should follow the reset-banking model users were told to expect:

• Add a banked reset that users can spend later, or
• Clearly state in the announcement that the reset is an immediate hard reset and will not be banked.

What is the actual behavior?

The reset appears to have been applied immediately rather than added as a banked reset.

There is also no reliable user-visible reset status/count to verify whether a reset was granted, consumed, expired, or never issued.

Why this matters

The difference is material.

An immediate reset can be wasted if:

• A user is asleep or away.
• A user is near their normal reset time.
• A user deliberately saved work for later because they understood resets could be banked.
• A user already spent an existing banked reset expecting future public resets to be banked.

Please either retroactively grant the banked reset, or clarify that public/manual Codex reset announcements are not covered by reset banking.