Tracking when OpenAI's Codex usage limits get reset, announced (often out of the blue) by Thomas Sottiaux on X. Total resets 35, average interval 8.9d, longest gap 67.7d. Reset activity over the last 26 weeks shows near-weekly cadence.

Recent logs show resets coinciding with user milestones: 9M active users across Codex and ChatGPT Work, plus 8M, 7M, and 3M markers. Announcements cite scaling traffic, reliability incidents, and new features like plugins and GPT-5.6 Sol. One reset celebrated a 'RESET week' internally.

Banked resets let users replenish weekly limits on demand; a web/mobile rollout fixed a bug affecting <10% of users. Resets also followed rate-limit drains traced to cache optimizations, cyber-risk blocks, and a 2X promotional increase missed for ~9% of Plus/Pro users. Engineer Rajan Agarwal pressed the button for one reset.

Holiday periods brought 2X limits until Jan 1, while a $100 plan launch and system rewrite triggered additional resets. The tracker captures a pattern of frequent, often unannounced limit refreshes as OpenAI iterates at speed.