A sourced, append-only record of discretionary AI quota resets — and two forecasts built on it. Covers March–July 2026. Published by Kingy AI.
Ledger covers March–July 2026 · last updated · the Codex figure is re-reckoned against your clock on load, so it is never the frozen build-time number. If a reset happened that this ledger has not recorded yet, the figure reads low, not high.
Uncalibrated. These are hand-set priors from the public event record, not fitted parameters, and none has been checked against an outcome. There is no accuracy score behind them yet. The Codex figure is derived from this ledger's own record of when resets actually happened and updates as time passes; the Claude Code figure comes from a stated baseline, because no supported channel exposes its usage state. How they are built · why they may be wrong.
Three other sites track Codex quota resets. They are good, they are faster than this one, and two of them have been running longer. Here is what each does, checked by opening them on 2026-07-27 — and the one thing this project does that none of them offers.
Will Codex Reset? — a single-page verdict on the next 48 hours.
Codex Reset Monitor — a timeline of confirmed resets plus 24h/48h estimates.
Built by @wong2__ — the longest event list of the three, with interval statistics.
Entry counts are deliberately absent as a differentiator: they move daily, and on the date checked two reads of one tracker disagreed about its own total. The rows below are structural, so they stay true between checks.
| willcodexquotareset.com | codexreset.org | codex-resets.com | this project | |
|---|---|---|---|---|
| Forecasts the next reset | ✓ | ✓ | ✗ | ✓ two models, labelled uncalibrated |
| Lists past events | ✗ | ✓ | ✓ | ✓ 28 current, 2026-03-13 → 2026-07-21 |
| Covers Anthropic Claude Code | ✗ | ✗ | ✗ | ✓ 9 records |
| Links a source per event | ✗ | ✓ | ✓ | ✓ 62 sources |
| Archived copies of those sources | ✗ | ✗ | ✗ | ✓ 61 of 62 (98.4%) |
| Confidence graded per event | ✗ | ✗ | ✗ | ✓ 3 confirmed, 14 probable, 11 reported |
| Corrections visible | ✗ | ✗ | ✗ | ✓ 19 superseded records, still on the page |
| Machine-readable download | ✗ | ✗ | ✗ | ✓ JSON and CSV, CC BY 4.0 |
| Method published with its parameters | ◐ | ◐ | ◐ | ✓ every weight and half-life, generated from the code |
✓ offered · ◐ described but not published in full · ✗ not offered
Stated plainly, because a comparison with no losing rows is an advertisement.
Not speed, and not the number of events — we lose the first outright and do not clearly win the second. It is this:
Every record here can be checked, and when one of ours was wrong you can see that it was wrong and what replaced it.
None of the three grades its confidence, and none shows a correction history, so an error on those sites leaves no trace. Errors do happen — including here. This ledger carries 19 superseded records, and both halves of every correction stay on the page precisely so you can see what we got wrong. That, plus two things nobody else offers at all: Claude Code coverage and a download.
All three sites were fetched and read on 2026-07-27; the ledger figures in the right-hand column are computed from the records as at 2026-07-26. These are live products and their pages change — if a row here is wrong, it is wrong on our side, and the correction belongs in the issue tracker. Nothing on this page is affiliated with OpenAI or Anthropic.