Codex needs authoritative in-product quota reset and entitlement-change notices

Open 💬 4 comments Opened Jul 23, 2026 by grtninja
💡 Likely answer: A maintainer (github-actions[bot], contributor) responded on this thread — see the highlighted reply below.

Summary

Codex users can see an account-specific reset timestamp in analytics, but broader quota resets, entitlement changes, temporary capacity changes, and product rollouts are often inferred from social posts, comments, or third-party forecasting sites rather than from an authoritative in-product event history.

This creates an avoidable information asymmetry. A user nearing the limit cannot reliably distinguish:

  • their normal account-specific weekly reset;
  • a global or cohort quota reset;
  • a temporary incident-related allowance change;
  • a product release that changes available models or limits;
  • a staged rollout that does not apply to their account;
  • an inaccurate third-party prediction.

This report does not allege selective or follower-based resets. The product defect is the absence of an authoritative, account-applicable communication channel inside Codex.

Environment

  • Product: Codex Desktop and ChatGPT Codex analytics
  • Platform: Windows 11 Pro
  • Subscription: ChatGPT Pro
  • Observed: July 2026

Observed behavior

  1. Account analytics shows a precise weekly reset date and time.
  2. Public expectations of an additional reset or product change circulate through social media and third-party prediction pages.
  3. The user has no in-product record stating whether the event is official, global, cohort-specific, account-applicable, completed, delayed, or unrelated to their normal reset.
  4. The user plans production work around incomplete information and may reach the usage limit while waiting for an event that does not apply to the account.
  5. When the expected change does not occur, there is no authoritative explanation or event log in the product.

Expected behavior

Codex should expose a signed/authoritative account event feed containing:

  • current five-hour and weekly windows;
  • exact reset timestamps with timezone;
  • source of each reset (scheduled, manual_global, incident_remediation, plan_change, promotion, or other defined class);
  • announcement timestamp;
  • effective timestamp;
  • account/cohort applicability;
  • whether the change completed successfully for this account;
  • old and new entitlement/limit values when they change;
  • links to the relevant official status incident or release note;
  • a clear statement when a public announcement does not apply to the current account.

The app should notify the user before a displayed reset date changes and explain why.

Suggested UI

A small Usage events section near the usage meter:

Weekly reset: Jul 28, 2026, 1:09 PM local time
Last entitlement event: No account-applicable events since Jul 21
Official incident: Elevated Error Rates — monitoring
Third-party forecasts are not authoritative

Impact

  • Users cannot plan high-value work reliably.
  • Social-media followers and users watching external comments receive information sooner than users relying on the product.
  • Third-party predictions can be mistaken for official account changes.
  • Confusion is amplified when usage is already being consumed by retries, compaction, or service incidents.
  • Trust suffers even when the underlying quota implementation is functioning as designed.

Related reports

  • #27435 — reset display should remain date/time unambiguous.
  • #30816 — displayed weekly reset date changed without explanation.
  • #24080 — expose reset times and account rate-limit details.
  • #35037 — active OpenAI incidents are not surfaced inside the running Codex session.
  • #35032 — usage was consumed during repeated compaction/reprocessing.

The requested outcome is not privileged early access. It is equal, authoritative, account-specific product communication for every user.

View original on GitHub ↗

4 Comments

github-actions[bot] contributor · 1 month ago

Potential duplicates detected. Please review them and close your issue if it is a duplicate.

  • #35042
  • #35040

Powered by Codex Action

grtninja · 1 month ago

Consolidating the same-account duplicate reports #35040 and #35042 into this canonical issue.

Please also include these event-ledger fields from those reports:

  • status: pending, applied, skipped, reversed, or expired;
  • affected usage pool/model;
  • before/after balance when applicable;
  • eligibility/reason code;
  • stable event identifier that support can reference;
  • explicit distinction between the normal recurring reset, a broad announcement, and an event actually applied to this account.

#35040 and #35042 will be closed as duplicates of #35044 so the discussion and maintainer signal remain concentrated here.

grtninja · 1 month ago

Concrete reproduction detail from duplicate report #35045: a broadly worded employee post was interpreted publicly as an imminent paid-user reset, third-party forecast pages amplified it, while the authenticated Pro analytics page continued to show the existing July 28 reset timestamp and no account-specific eligibility/rollout explanation. This is preserved here as the motivating case; it does not claim individual or follower-based reset control.

grtninja · 1 month ago

Additional July 24 reproduction of the information-asymmetry problem:

A third-party reset forecast displayed the July 21 10M reset announcement as its large “LATEST COMPLETED RESET POST” hero card on July 24, but the card omitted the post date. The same page simultaneously showed 2 open incidents, described the incident desk as quiet or the last reset as too recent, and issued generic token advice that was not grounded in the authenticated account's currently constrained usage state.

The underlying historical selection may be technically “latest completed reset,” but the presentation makes stale history look like a current signal. It also collapses separate facts into one score:

  • historical global reset event;
  • current official incident state;
  • social/public hints;
  • account-specific quota, credits, eligibility, and completion state.

This is precisely the gap requested here. An in-product event feed should make the current account state authoritative and show, at minimum:

  • absolute event timestamp and age;
  • event class (historical_global_reset, scheduled_account_reset, incident_remediation, etc.);
  • account applicability and applied/not-applied state;
  • current official incident correlation;
  • an explicit third-party forecast — not authoritative boundary.

No private account identifiers or payment information are included.