A confirmed, staff-announced, provider-wide replenishment of usage limits (a "hard reset"), or a redeemable "banked" reset offer. It is not your personal 5-hour or weekly rollover, and a green light never guarantees your own account is full.
We poll allowlisted official/staff accounts (for example OpenAI's @thsottiaux and Anthropic's @ClaudeDevs) on a schedule via a public timeline mirror, with a fallback source. Every confirmed event links to its original announcement so it can be audited.
Strict text rules must clear a usage-phrase floor and an all-users scope before a green light; catchphrases alone never qualify. Teasers, quotes, replies and negations are rejected. If a source goes stale it shows "source unhealthy", never "calm". A false green is treated as the highest-severity incident.
A deterministic Weibull renewal survival model is fitted to the intervals between confirmed hard resets. Given the time since the last reset, it reports the conditional probability of another public reset within 48 hours, with a jackknife uncertainty range and the sample size. Because the process is clustered and heavy-tailed, the probability is highest shortly after a reset and decays as a drought lengthens. The forecast is a demoted, explainable estimate — never a confirmation, never a notification, and it never turns a card green. With fewer than two confirmed hard resets it reports "insufficient data".
Zero login; the board only reads a public API and never accesses your AI accounts. We prefer "unknown" over a false green, and every confirmed event keeps a source link.
Independent utility. Not affiliated with OpenAI, Anthropic, xAI, Moonshot, z.ai, or Google. A confirmed public reset does not guarantee your personal quota is full.