Useful information. Clear boundaries.
WhenReset is an independent guide to AI usage resets and allowances.
How records enter the radar
We review official documentation, changelogs, status pages, and original provider announcements. Reviewed records carry their source, audience, event type, and the limits of what was established.
Older Codex history comes from the Codex Resets public API. Those entries remain third-party archive reports until independently reviewed.
Different clocks, different meanings
An announcement time, a tracker observation, an effective policy date, and a personal reset time are not interchangeable. We label them separately and show date-only precision when an exact time is unavailable.
A banked credit is not an automatic refill. A service incident or capacity increase does not count as a bonus usage reset.
Freshness you can inspect
The public archive and available provider status feeds are refreshed on request with a five-minute server cache. The overview checks again every minute while visible.
If a source fails, saved records remain visible with a refresh warning. We never replace an unavailable status with “operational.” Reviewed rules were last checked Sep 10, 2026, 06:26 AM UTC. New policy claims require editorial review.
History is not a forecast
We do not display a “chance of reset” without a calibrated model. Reset announcements depend on releases, promotions, capacity, and operational decisions we cannot see.
A past gap between resets does not schedule the next one. You can save your account’s own reset time instead.
Source directory
Independence & privacy
WhenReset is not affiliated with or endorsed by OpenAI, Anthropic, or Google. Product names identify the services being discussed. Personal timers stay on your device. We do not ask for API keys, passwords, or provider session tokens.
Read the privacy notes