Tibo (Tibo) Sottiaux, head of Codex and ChatGPT Work at OpenAI, promised on X on October 5, 2026, that for 28 days his team will either ship a clear improvement for most Codex and ChatGPT Work users or announce a full quota reset. The statement came after recent Codex outages and amid the release of Anthropic Claude Opus 5.5. For users, the coming month sets a daily checkpoint: either a new feature or fresh quotas.

image

What happened

On October 5, 2026, Tibo (Tibo) Sottiaux, who leads Codex and ChatGPT Work at OpenAI, posted on X a promise: for 28 days, the team will either ship a “clear improvement relevant to most Codex/Work users” or announce a full quota reset. The structure is daily: on each individual day, the window requires one of two outcomes, meaning a reset applies only to days without a significant improvement, not guaranteed all 28 times in a row. The day before, Sottiaux listed four team priorities for this period: product simplification, efficiency gains in the form of more useful payload for the same quota, breakthrough features, and new models.

Context

The promise did not come out of nowhere: OpenAI had already applied global quota resets after Codex outages on September 25–26 and October 1–3, 2026, so the public scorecard with a deadline reads as addressing the aftermath of those incidents and competitive pressure following the release of Anthropic Claude Opus 5.5, against which comparisons between Codex and Claude Code intensified. By design, this is a format of public accountability, not a technical release: the sources contain no changes to APIs, prices, or latency, nor benchmarks.

Why this matters for the industry

A public commitment with a deadline from a product leader turns Codex development into a daily public report — a rare accountability format for developer tools. If the focus mode on four priorities holds, the industry may get a new standard for AI tool product team communication, where a day without a significant improvement is compensated with quotas, and tool evaluation shifts from presentation benchmarks at launch to continuous external verification; competitors would then have to either copy the format or explain its absence. The hypotheses testable within the promise window cover only two of the four priorities — new models and increased payload per quota — so it is worth watching measurable artifacts in these two areas.

Why this matters for users

If you work with Codex or ChatGPT Work, the next 28 days provide a daily checkpoint: either a new feature or replenished quotas, the remainder of which can be checked in Settings → Usage. The reset mechanics matter for planning: it updates the remaining limits, not their cap, so hoarding quota in anticipation of compensation is pointless — it makes sense to launch large tasks immediately after replenishment and check the remainder in the morning. Separately, note that on October 30, within the promise window, a reduction of the Pro plan limits from 20x to 10x of Plus for Codex and ChatGPT Work is still scheduled, so when choosing a subscription, focus on available limits rather than the month’s promises.

What is still unknown / limitations

The criterion of a “clear improvement relevant to most Codex/Work users” is not operationalized: it has no metrics, methodology, or way to be falsified, so the promise is by design impossible to fail, and any day can be declared successful — this does not create an evidence base for planning integrations. At present, only the commitment itself is confirmed, not improvements: there are no new models, evaluations, or measurable methodologies in the sources. In addition, it is unknown whether the daily rhythm will hold for all 28 days and whether the format will persist after the window; under constant daily pressure, the bar for a “clear improvement” may blur into cosmetic fixes, and audience trust in such promises may burn out.

Sources

Author

Look at AI, editorial team