Current reset data
The live panel reads the latest reset, elapsed time, lifetime average, recent-ten average, P25, P75 and longest interval. When the API is unavailable it shows N/A instead of stale hard-coded figures.
Public signals come first
Historical pace cannot prove when another reset will happen. An explicit public signal—such as Tibo saying more resets are coming in a named week—carries more forecasting weight.
A reply still does not promise an exact day. If no day or timezone is given, AIReset uses a broad watch window and links to the source.
How the statistical window works
Without an active signal, the start is the latest Full Reset plus the recent-ten average. The end uses the latest Full Reset plus P75. The lifetime average is a fallback, and insufficient samples produce no window.
When an explicit signal exists, its time boundary takes priority. Historical figures then explain whether the current wait is ordinary or unusually long.
Why an exact time is impossible
Extra resets may follow incidents, launches, milestones, cost decisions or operational judgment. The events are not a stable recurring schedule, and a longer wait does not create certainty.
Some announcements cover banked credits or selected plans. AIReset distinguishes completed events, announcements and contextual signals rather than converting vague language into a promise.
Official time versus forecast
Your Usage page shows the normal windows guaranteed for your account. Public signals concern possible additional resets. Both can exist at the same time and serve different purposes.
The estimated window is a public-signal and historical analysis. It does not represent an official OpenAI schedule.