Codex reset estimates and outcomes
Every labelled estimate of the next Codex usage reset, with its sources, what our rules changed, and how it was graded against @thsottiaux’s own posts.
Each row is one estimate we published, including the ones a newer estimate replaced. A hit means @thsottiaux announced a Codex reset inside our window; a miss means no such announcement inside it, including a reset that came before the window opened or after it closed; withdrawn means we pulled the estimate before it resolved, and it counts as not-hit. Rows are never deleted, rewritten or moved off the record.
Track record
Record: 3 hits · 10 misses · 2 withdrawn · hit rate 20% of 15 · median window 11 h
Excluding withdrawals: 23%
Last 90 days: 3 hits · 10 misses · 2 withdrawn (15 resolved)
A calendar-only guess with the same windows would have hit about 1.4 of these 15
| Estimated at (UTC) | Window (UTC) | Confidence | Outcome | Graded by | Signals | What the rules changed | Why |
|---|---|---|---|---|---|---|---|
| → | low | miss | window expired · no reset post |
| No changes | Whypost:2113492187770175603 names a reset; post:2113492502355562875 suggests delivery the next Pacific morning. Together they support retaining the active window, from Pacific midnight through the literal PST deadline. The reply's missing parent leaves its connection to the reset unclear, so confidence stays low. What would make it wrong: A retrieved parent shows that post:2113492502355562875 refers only to a feature release. · The window ends at 2026-10-23T18:00:00Z without a primary-source report of Codex reset delivery. codex-cli-unreported · 2026-09-15.codex.1 | |
| → | low | miss · superseded | window expired · no reset post |
| No changes | Whypost:2113492187770175603 names a reset; post:2113492502355562875 suggests timing on the next Pacific calendar day, before 10am. The window accommodates both Pacific daylight time and literal PST. The reply's missing parent leaves its connection to the reset unclear, and the scorecard supports caution. What would make it wrong: The reply's parent or a later post shows that the timing refers only to a product release. · A later post postpones the reset beyond 2026-10-23T18:00:00Z or cancels it. codex-cli-unreported · 2026-09-15.codex.1 | |
| → | low | miss | window expired · no reset post |
| No changes | Whypost:2111492673890824348 names a limit reset; post:2111492738025959922 suggests delivery today. Together they may support this window, allowing for the wording "EOD PST". The reply's parent is unavailable, so its connection to the reset and the Codex scope remain unclear. What would make it wrong: The missing parent identifies a different feature or product as the subject of the timing reply. · A later Tibo post postpones the reset beyond 2026-10-18T08:00:00Z or excludes Codex. codex-cli-unreported · 2026-09-15.codex.1 | |
| → | low | withdrawn | withdrawn by the model |
| No changes | Whypost:2109812360596865224 points to a reset by Pacific end of day, supported by post:2109812216212099456. The six-hour window brackets that deadline with some allowance for propagation. The reply's parent is unavailable, and “PST” leaves a daylight-saving ambiguity. Past estimate results also warrant caution. No published preview covers this period, and no new post contradicts the active estimate's cited signal. What would make it wrong: A later post postpones or cancels the reset described in post:2109812360596865224. · A later primary-source update places the reset outside 2026-10-13T05:00:00Z–2026-10-13T11:00:00Z. Withdrawal reason: post:2109873287853707513 says "It's done.", superseding the active estimate's premise that the announced reset remains ahead. Withdraw the future window rather than extend it to another reset. codex-cli-unreported · 2026-09-15.codex.1 | |
| → | medium | miss · superseded | window expired · no reset post |
| No changes | Whypost:2109812360596865224 points to a global reset by Pacific end of day, alongside post:2109812216212099456 saying "Day 8/ Reset". The window accommodates both Pacific daylight time and literal PST wording. The timing reply's parent is unavailable, which limits confidence. The published previews concern earlier periods. What would make it wrong: A later Tibo post postpones the global reset beyond 2026-10-13T08:00:00Z. · The missing parent or a later clarification places this reset outside Codex usage limits. codex-cli-unreported · 2026-09-15.codex.1 | |
| → | low | hit | @thsottiaux post ↗ · |
| No changes | Whypost:2109045392684032299 says "Global reset by EOD", while post:2109045220411351517 says "Day 6/ Reset". Together they support retaining the active window through the end of Saturday Pacific time. The reply's parent is unavailable and neither post explicitly names Codex, so product scope remains unclear. Earlier preview events concern past periods. What would make it wrong: A later post excludes Codex from this global reset or moves it beyond 2026-10-11T07:00:00Z. · The window ends at 2026-10-11T07:00:00Z without a Codex reset announcement or observable Codex limit replenishment. codex-cli-unreported · 2026-09-15.codex.1 | |
| → | low | hit · superseded | @thsottiaux post ↗ · |
| No changes | Whypost:2109045392684032299 says "Global reset by EOD." Together with "Day 6/ Reset" in post:2109045220411351517, this supports a window ending at Pacific midnight, 07:00 UTC. The reply's parent is unavailable and Codex scope is unstated, so confidence remains low. The published previews concern earlier periods. What would make it wrong: A later post postpones the reset beyond 2026-10-11T07:00:00Z or cancels it. · A later clarification excludes Codex from the global reset. codex-cli-unreported · 2026-09-15.codex.1 | |
| → | low | hit · superseded | @thsottiaux post ↗ · |
| No changes | Whypost:2109045392684032299 says "Global reset by EOD." Read in Pacific time, this points to a window ending at 2026-10-11T07:00:00Z. post:2109045220411351517 adds "Day 6/ Reset". The reply's parent is missing, and neither post explicitly names Codex. Those gaps and the scorecard's six misses support low confidence. Earlier preview events concern past periods. What would make it wrong: A later post says the global reset excludes Codex. · A later post postpones the reset beyond 2026-10-11T07:00:00Z. · The window ends without a Codex reset announcement or observable usage-limit replenishment. codex-cli-unreported · 2026-09-15.codex.1 | |
| → | medium | miss | window expired · no reset post |
| No changes | Whypost:2103637477760311522 explicitly commits to a paid-user Codex reset after service recovery; incident:01M3DCNWMW57HYK8FJ5FBFPA39:resolved provides recovery context. This supports retaining the remaining active window. Timing is inferred from the immediate recovery context, with no deadline stated; the past estimate misses also warrant caution. What would make it wrong: A later Tibo post postpones the paid-user Codex reset beyond 2026-09-26T12:00:00Z. · A subsequent rollout report places application of the usage-limit changes before this window starts or after it ends. codex-cli-unreported · 2026-09-15.codex.1 | |
| → | medium | miss · superseded | window expired · no reset post |
| No changes | Whypost:2103637477760311522 explicitly promises a Codex usage-limit reset following service recovery. The resolved incident:01M3DCNWMW57HYK8FJ5FBFPA39:resolved supports that recovery context. Together they suggest a near-term makegood within this window, but the post gives no execution time. The existing estimate remains supported; its start is moved forward to meet the snapshot window constraint. What would make it wrong: A later Tibo post postpones the reset beyond 2026-09-26T12:00:00Z or cancels it. · A later primary-source update places execution before 2026-09-26T01:00:00Z. codex-cli-unreported · 2026-09-15.codex.1 | |
| → | medium | miss · superseded | window expired · no reset post |
| No changes | Whypost:2103637477760311522 explicitly commits to a reset for paid Codex users following service recovery. The resolved incident:01M3DCNWMW57HYK8FJ5FBFPA39:resolved supports that recovery context. This suggests a near-term makegood; the window is an inference from the immediate recovery wording, not a stated schedule. What would make it wrong: A later Tibo post postpones the paid Codex reset until after 2026-09-26T12:00:00Z. · A later Tibo post retracts the reset commitment for paid Codex users. codex-cli-unreported · 2026-09-15.codex.1 | |
| → | medium | withdrawn | withdrawn by the model |
| No changes | Whypost:2102254445082116335 explicitly names Tuesday, supporting a window covering the remainder of Tuesday Pacific time. post:2101920928070562029 mentions an earlier hour that has passed, but its missing parent prevents treating it as a reset deadline. The supplied evidence contains no later cancellation or reset announcement for this period. What would make it wrong: A later Tibo post cancels the Tuesday reset or moves it to another day. · Tuesday Pacific time ends without a primary-source reset announcement. Withdrawal reason: post:2102463847714247142 specifies banked credit, superseding the active estimate's interpretation of post:2102254445082116335 as a timed usage-limit reset. No separate reset timing is provided. codex-cli-unreported · 2026-09-15.codex.1 | |
| → | medium | miss · superseded | window expired · no reset post |
| No changes | Whypost:2102254445082116335 explicitly ties a reset to Tuesday, supporting a window covering the remaining Pacific Tuesday. post:2101920928070562029 lacks its parent, so its reference to an hour does not justify a narrower window. No published preview covers this period, and no new post contradicts the active estimate's supporting statement. What would make it wrong: A later Tibo post postpones the reset beyond Pacific Tuesday or cancels it. · The window ends without a primary-source reset announcement. codex-cli-unreported · 2026-09-15.codex.1 | |
| → | medium | miss · superseded | window expired · no reset post |
| No changes | Whypost:2102254445082116335 supports retaining the existing window covering Tuesday in Pacific time. It does not explicitly name Codex. The missing parent of post:2101920928070562029 leaves its timing context unclear, so it does not justify narrowing the window. What would make it wrong: Tibo posts a cancellation or moves the reset beyond Tuesday Pacific time. · A later post identifies the promised reset as applying only to a product other than Codex. · No Codex reset announcement appears by 2026-09-23T07:00:00Z. codex-cli-unreported · 2026-09-15.codex.1 | |
| → | medium | miss · superseded | window expired · no reset post |
| Confidence lowered from high to medium: no cited post states a coming reset in explicit words. | Whypost:2102254445082116335 explicitly names Tuesday for a reset. This points to the full Pacific Tuesday window shown here in UTC; the post gives no specific hour. No published preview covers this period. What would make it wrong: A later Tibo post postpones or cancels the Tuesday reset. · The window ends at 2026-09-23T07:00:00Z without a Codex reset announcement. codex-cli-unreported · 2026-09-15.codex.1 |
Estimates are our reading of public signals. Only @thsottiaux’s own words count as an announcement, and only your account shows your reset time.