When it Reset

中文EN

Public channels

Evidence ledger

Verified = a first-hand post (Codex leads, official accounts) or a status-page original. Reported = someone else retelling it. Full resets, banked credits, official schedules, and your personal window are four different things. We never log into your account.

WAITNo linked ingest yet. Briefing and signals come only from the Grokbot timer — this site does not ship fake samples.

✓ Verified 0Reported 0 0 items

When the next one might land

Official schedule first · historical gaps · conditional rates

Official schedules come first. Otherwise we use verified full-reset gaps for a median date and IQR window, then an empirical rate: of historical waits that already lasted this long, how many reset in the next 24/48 hours. Too few samples or noisy gaps → no date. Confidence is capped at 60%. The 24/48h rate is never the homepage hero.

Codex

samples 0/4

Not enough verified full resets to publish a date. Fill the ledger, or backfill linked historical posts.

Claude

samples 0/4

Not enough verified full resets to publish a date. Fill the ledger, or backfill linked historical posts.

How the model works

We borrow discrete-time survival (willtiboreset) and evidence lineage (codex-reset-forecaster), without community-noise weights and without putting 24/48h probability on the homepage. Official schedules are facts. Statistics use verified full resets only. Conditional rates are shown as k/n so you see the sample, not a marketing percentage.

  1. A verified official schedule is shown as a landing time, not as a probability.
  2. Verified full resets merge within 12 hours; we take the median gap and IQR window.
  3. Conditional rate: of historical gaps that already lasted this long, k/n reset in the next 24/48h. Hidden if n < 3.
  4. Confidence = (1 − IQR/median) × sample discount, capped at 60%. Too noisy → wait for a post.

Ledger

Newest first

No ledger rows yet.