Thibault “Tibo” Sottiaux, who heads products and platforms at OpenAI and is the public face of its Codex coding agent, has turned the team’s next month into a public accountability drill. In a post on X on Monday, he pledged that every day for the next 28 days, Codex and ChatGPT Work will either ship a daily update delivering “clear improvements for most users” or undergo a full usage reset.

Four areas, one clock

The pledge follows a post the day before in which Sottiaux said the team had “locked in its focus.” User feedback, he said, is clear: people want the product to be simpler. Going forward the team will concentrate on exactly four areas: simplifying the product, increasing efficiency to give users more usable capacity, breakthrough features, and new models.

By attaching a daily outcome to the sprint, OpenAI is converting the most complained-about part of Codex — its usage limits — into the incentive. The promise is blunt: a visible improvement every single day, or everyone gets their usage reset.

Close-up of a developer’s hands typing code on a laptop keyboard.
Sottiaux’s four focus areas: a simpler product, more usable capacity, breakthrough features, and new models. Photo: Pexels.

The Claude Code pressure

The sprint lands amid intensifying competition. Anthropic recently released Opus 5.5, and comparisons between Claude Code and Codex have become routine — centering on model capabilities, speed, and subscription allowances. The reception of Sottiaux’s pledge was a telling mix, per explainx.ai: hope, gambling jokes, and cancelled subscriptions.

OpenAI has been here before. When Codex and ChatGPT Work reached 8 million active users, the company reset usage limits for everyone and dropped the five-hour rate limit for a time — an early taste of resets as a goodwill lever. Plus subscribers have also lived with a usage allowance that resets every five hours, alongside a weekly quota.

A developer working late at a desk with multiple screens, evoking a daily shipping sprint.
28 daily checkpoints: a ship date every single day, or a full reset. Photo: Pexels.

What to watch

Two things will decide whether this is a stunt or a turning point. First, whether the daily improvements are actually visible to ordinary users — “clear improvements for most users” is a high bar for a 28-day streak. Second, whether the reset side of the deal is real: a full usage reset on a day the team falls short would cost OpenAI actual inference budget, which is exactly why the promise is credible. Keep an eye on the Codex changelog and the replies under Sottiaux’s thread — developers will notice a missed day instantly.