When might the next public reset signal appear?
ResetPulse uses a transparent historical baseline to estimate whether another broad public reset signal could appear within 24 or 48 hours. It is not a prediction of your personal account window.
What the baseline uses
Confirmed, deduplicated broad-public full-refresh events, their published UTC times, and the last fourteen usable gaps. Plan-specific, announcement, ambiguous, and duplicate records stay out of the cadence history.
Elapsed-aware, not overdue-biased
The estimate first asks which historical gaps were still “at risk” after the current elapsed time, then blends that small sample with a three-gap short-horizon baseline. A stale source cannot turn into artificial certainty simply because the clock kept running.
What it does not use
It does not read your Codex account, private usage panel, plan, banked credits, or credentials. Your own client remains the source of truth for personal limits.
How to read confidence
Confidence reflects usable gaps and the number of historical intervals still at risk, not the chance that an individual account will reset. A small at-risk sample reduces confidence instead of being dressed up as certainty.
Why keep it separate?
A global goodwill reset and a rolling five-hour or weekly window are different events. Mixing them creates the exact false alarm this monitor is designed to prevent.
Backtest, not marketing
The live card shows classification accuracy and Brier score. Brier score measures squared probability error, so lower is better; both metrics remain descriptive until a larger sample exists.
See the live estimate and plan privately
The live homepage section shows the current 24-hour and 48-hour baseline. The same section includes a browser-only personal planner and calendar export for the time shown by your own Codex client.
Evidence boundary
An estimate is not an announcement, a guarantee, or proof that any account has changed. If the public evidence is delayed, ambiguous, or unavailable, the correct result is unknown. Read the methodology and data sources and limitations for the full boundary.