The reset button has become a public signal

Codex Resets is an independent tracker that organizes public X announcements from OpenAI's Thibault Sottiaux into a chronological history of usage-limit resets for Codex and ChatGPT Work. The 35 resets and 8.9-day average reported when the link first circulated are already outdated. As of September 2, the site shows 49 resets, a 7.2-day average interval, a 67.7-day longest gap, and a latest reset 2 days ago. Part of the site's appeal is that its own statistics move quickly as resets become more frequent.

From incident compensation to growth events

The history shows that resets are more than compensation for outages. They have followed service incidents, unusual latency, and usage-accounting errors, but they have also accompanied new plans, model launches, plugin experiments, and weekend events. During the stretch when Codex and ChatGPT Work successively reached 7 million, 8 million, and 9 million active users, ordinary and banked resets appeared almost like growth celebrations.

This pattern shows OpenAI reducing friction so people can keep attempting larger projects while infrastructure catches up with demand. Usage limits have become an operating tool that connects capacity management, trust recovery, and product discovery—not merely a hard stop.

What makes a banked reset different

An ordinary reset immediately refreshes eligible limits for the covered accounts. A banked reset is a one-time benefit stored on an account until the user chooses to apply it. According to official guidance, it is not purchased credit and does not permanently increase a plan's allowance. Eligibility, expiration, and scope can vary by announcement and account, so users should verify their actual balance in their own Usage view.

The best way to read the tracker

Codex Resets is an observation log of public announcements, not a calendar that guarantees the next reset. Urgent development work should therefore be planned around current account limits and priorities rather than an assumption that another reset is imminent. Even so, the tracker offers a rare timeline where service growth, incident response, and product adoption meet in one dataset. Frequent Codex users may find it worth bookmarking.

What to watch next

The more interesting question is not simply how many resets occur, but how the operating model changes: which events receive banked resets, whether announcement intervals stabilize after major growth, and whether the Usage experience becomes more predictable.

Sources

• Codex Resets
• Public announcements from Thibault Sottiaux
• OpenAI: How banked Codex resets work
• GeekNews introduction

How could this be used in biotech?

Teams using Codex for bioinformatics pipelines, analysis-script review, literature-data organization, or repetitive experiment automation can treat reset history as a secondary planning signal. Free resets are not guaranteed resources, however, so critical analysis should still rely on reproducible code, checkpoints, and local execution alternatives. A reset is an opportunity to experiment more boldly—not a reason to skip data-integrity or validation procedures.