Did your AI coding plan reset?
No resets today. Nothing announced.
Codex
Last reset 2 months ago
Reset
@theo @maria_rcks But is it true
Nov 20, 22:07 UTC
Banked reset
We shipped some good updates to dots today and the banked reset has been activated and should land in everyones accounts
Nov 12, 21:59 UTC
Banked reset
Thursday is feeling codexy to me.
Nov 11, 06:57 UTC
Reset
It’s been a while without resets.
Nov 11, 06:55 UTC
Claude Code
Last reset 3 months ago
GLM Coding Plan
Last reset 3 months ago
Higgsfield
Last reset 2 months ago
Latest signals
Every entry links to its source. Unconfirmed reports are labelled and never move a forecast.
- Nov 20, 22:07 UTC
@theo @maria_rcks But is it true
- Nov 12, 21:59 UTC
We shipped some good updates to dots today and the banked reset has been activated and should land in everyones accounts
We shipped some good updates to dots today and the banked reset has been activated and should land in everyones accounts sometime soon!
- Nov 11, 06:57 UTC
Thursday is feeling codexy to me.
Thursday is feeling codexy to me. Perhaps also a good time for a banked reset.
- Nov 11, 06:55 UTC
It’s been a while without resets.
It’s been a while without resets. Maybe I should come back out of the basement
- Nov 8, 05:59 UTC
@ForwardEditor We hesitated to embed into the codex app to make sense of the resets, but decided against it.
- Nov 5, 08:09 UTC
@fudgeposting Might be
- Nov 1, 17:15 UTC
@TokenGremlin I demand a banked reset
- Nov 1, 00:43 UTC
Banked 1/2
How a reset becomes a number
- 01
Watch
Official X accounts, status pages, release notes, docs and community threads are polled on a schedule.
- 02
Classify
A model reads each new item and decides whether it is a reset, banked reset, announced reset, boost, limit change or incident, with a confidence score.
- 03
Verify
Confident official reports publish on their own. Community reports stay unconfirmed until an official source or three independent reports back them up. Anything unsure waits in a review queue.
- 04
Forecast
Verified resets feed a memoryless cadence model. We backtest it every day and publish the score against a baseline.

- 1//// -- one item, two sources
- 2source: status.claude.com [official]
- 3item: "Elevated errors on Claude Code"
- 4kind: outage
- 5confidence: 92 // official, 75 or higher
- 6decision: verified // published automatically
- 8source: hacker news [community]
- 9item: "my weekly limit just reset"
- 10kind: reset
- 11confidence: 68 // community stops at unconfirmed
- 12decision: unconfirmed // never moves a forecast
Build on the data
Free JSON, an Atom feed and an MCP server for agents. No key needed. If you show the data, credit airesettracker.com with a link.
API docs
- 1{
- 2 "providers": [{
- 3 "id": "claude-code",
- 4 "forecast": {
- 5 "basis": "cadence",
- 6 "p24": 14, "p48": 26,
- 7 "band24": [7, 21]
- 8 },
- 9 "lastResetAt": "2026-10-02T09:37:00Z"
- 10 }]
- 11}
Questions
How do you know a reset really happened?
A reset is only called verified when an official channel says so and the classifier is confident, or when independent reports corroborate it. Every entry links to its source, and unconfirmed reports are labelled.
Why is there no forecast for some plans?
A forecast needs at least three completed reset intervals. Until then we show "collecting history" instead of inventing a number.
Does waiting longer make a reset more likely?
No. The model is memoryless: goodwill resets are irregular, not scheduled, so the 24h number stays flat while you wait.
Can you see my account or usage?
No. We never touch your accounts. Your personal clocks live in your browser’s local storage.
Is this affiliated with OpenAI, Anthropic, Z.ai or Higgsfield?
No. This is an independent community project. Product names are trademarks of their owners.