OpenAI Codex users pick quota resets over new features
When OpenAI Codex and ChatGPT users had to choose between a batch of new features or a fresh quota, they didn't hesitate. They picked more computing power. Thibault Sottiaux, who leads OpenAI Codex and ChatGPT, said the team ran a community vote after rolling out several updates. Even though users rated the new features as good or very good, the majority still wanted a full quota reset. OpenAI delivered.
This wasn't a fluke. Since October 4, Sottiaux's 28-day rule has set a clear pace: each day must bring either a real improvement for most Codex and Work users, or a complete quota reset. On the second day of the cycle, Sottiaux rolled out four new features and then put the next move up for a vote. Users could accept the updates, ask for a reset, or demand both. The answer was clear-reset the quotas.
After the launch of the 28-day cycle, OpenAI expanded Codex to allow tasks to be run in the cloud, making remote jobs accessible from other devices as well.
Features land, but quotas still matter most
The four updates included automatic approval ("Approve for me") that no longer eats into quota, a cleaner developer interface, built-in meeting minutes, and a new Decisions interface for developers. These changes aimed to make the platform smoother and easier to use. The automatic approval alone freed up 2% to 10% of plan usage. Sottiaux also pointed out that the model had solved several tough math problems, though that wasn't part of the day's feature drop.
Still, even with these upgrades, the most active users-especially those who rely on five-hour and weekly quotas-chose the reset. For them, the ability to run longer jobs and get more computing time matters more than small feature tweaks. Paid users benefit right away, able to keep working without interruption. OpenAI, meanwhile, takes on the extra computing cost and has to adjust its release schedule.
The 'Tibo reset' becomes a daily ritual
Resets, often announced from Sottiaux's own account, have picked up the nickname "Tibo resets" in the community. At first, they were a way to make up for outages, misconfigurations, or poor user experience. Over time, they turned into a meme and a daily routine, with Sottiaux even keeping a physical reset button on his desk. Not all resets are the same: there are full resets, banked resets, and extra credits users can buy, each with its own rules and effects.
On October 5, 2026, OpenAI reported an incident with elevated error rates in Codex Cloud. The issue was mitigated and moved to 'Monitoring' status, but remained active for several hours.
Unlike the usual service credits from cloud giants like Amazon Web Services or Cloudflare, Codex resets have become a daily, user-driven event. The process is open and personal, skipping the standard status pages and putting control in the hands of the most active users. This shift has changed the game. Instead of racing to ship new programming agent features, the real contest is now about who can keep up with demand for long task quotas. Features may roll out every day, but the five-hour cap is still the wall users hit.
Computing time is now the real currency
What's really changed is who holds the power over capacity. Subscription fees set the baseline, but resets decide how much work gets done each day. By voting for resets over features, users are telling OpenAI that extra computing time is now worth more than another interface tweak. Sottiaux's rules tilt the system toward resets, shifting capacity calls from the product roadmap to the community's daily needs.
This isn't the first time OpenAI has had to deal with quota drama. As Nerdschalk reported, the 28-day cycle forces the team to deliver a feature that "clearly improves" the experience for most Codex and ChatGPT Work users, or else trigger a full quota reset. The current setup, where compensation is not just automatic but up for a vote, has turned quota into a second currency-one that, for many, is more valuable than the latest feature drop.
Letting users pick between features and resets is a rare move in user-driven resource management. The results are plain: in a world where computing power is still scarce, quotas rule. The vote isn't about the best new feature-it's about what users need right now. Once compensation becomes votable, it turns into pricing, and OpenAI's community has chosen raw capacity over small changes. Features can change every day, but quotas are the hard currency that gives users real power.
