Codex will ship something or reset, every day for 28 days

For 28 days, Codex will ship a broadly useful improvement each day or issue a full usage reset; Day 1 was a roughly 50% speedup.

On 2026-10-04, Tibo Sottiaux posted that the Codex team would ship one clear improvement relevant to most Codex and ChatGPT Work users each day for 28 days, or issue a full usage reset. On 2026-10-05, he marked Day 1 with a roughly 50% default-speed increase for GPT-6 Astra and GPT-6.1 Sol through subscriptions. The post does not promise 28 resets or define which plans would receive one.

What the 28-day pledge says

Sottiaux posted the pledge at 20:33 UTC on 2026-10-04. His X profile identifies him with "Codex & ChatGPT @OpenAI." The pledge says each day will bring either one broadly relevant improvement or a full reset. It does not name a release schedule, eligible plans, or a delivery method for a reset.

At 04:59 UTC on 2026-10-04, Sottiaux said the work was limited to simpler products, more efficient usage, new features, and new models. Those are the categories he gave publicly. They are not a day-by-day roadmap.

At the 2026-10-06 source check, the pledge post showed 6.3 million views and 25,000 likes. Those numbers are a snapshot of the X page, not a measure of how many Codex users support the plan.

What Day 1 delivered

At 17:20 UTC on 2026-10-05, Sottiaux announced that the default speed for GPT-6 Astra and GPT-6.1 Sol was about 50% faster through subscriptions. He said the change applied across OpenAI products and partners using Sign in with ChatGPT, naming OpenCode, Pi, Amp, and Devin, and said users needed to change nothing.

An OpenAI Developer Community thread repeated the Day 1 announcement. In that thread, Vaibhav Srivastav described the change as an inference optimization for GPT-6 Astra and GPT-6.1 Sol and said the rollout was in progress. This is the evidence for Day 1. It is an announced performance change, not a benchmark from an independent test.

What a full reset does and does not mean

The pledge uses the words "full reset" without saying how the reset would reach a user. That distinction matters.

The checked reset guidance distinguishes a saved reset from an automatic or global reset. WillReset says a saved reset does not refresh usage until the user redeems it, and that a full redemption resets both windows. Explainx describes resets as one-time refills, not a higher plan ceiling.

WillReset also reports that Codex and ChatGPT Work draw from the same agentic pool. A reset therefore changes the available window, not the plan's basic entitlement. The pledge itself does not publish the eligible-plan list. Check the account or workspace where you use Codex or Work instead of treating the announcement as a guaranteed personal reset.

How to place the 28 days on a calendar

The 2026-10-04 post gives no formal end date. If 2026-10-04 counts as Day 1, Day 28 falls on 2026-10-31. If the first counted day is 2026-10-05, Day 28 falls on 2026-11-01. Those are calendar interpretations, not an announced deadline. The 2026-11-02 date is not supported and is one day too late.

For the practical state of an account, run the local status command and read the limit-window rows it actually prints:

codex /status

The reset tracker used for this review warns against inferring a missing five-hour row and says the weekly window has its own account-specific clock. That is a safer way to read the current allowance than converting the public pledge into a fixed quota promise.

Sources

Last verified: 2026-10-06.

Spotted an outdated or wrong claim? Agents can report it with evidence throughPOST /api/feedback; an editor checks every report. See llms.txt for the agent API.