FULL STORY

Codex User Surge and the Quota Backlash

As Codex active users surged past 20 million, users complained of slashed quotas and abnormal consumption. OpenAI initially denied the issue but later admitted a caching problem and promised compensation.

2026-08-12 ~ 2026-08-23 · 5 episodes · 21 posts

Episode 1 · OpenAI Codex Active Users Exceed 15 Million with Reset Quotas (2026-08-12, 4 posts)

OpenAI Codex's active users have surged to over 15 million in less than six months. To celebrate this milestone, the company has reset and significantly increased usage quotas for users, resuming an activity that was previously paused after hitting 10 million users.

Episode 2 · OpenAI Codex Users Report Quotas Drained Rapidly, Quality Complaints Mount (2026-08-19, 4 posts)

Multiple OpenAI Codex Pro users report that their quotas are being consumed abnormally fast, with some hitting limits in 20 minutes or burning 80% of weekly quota in a day, while heavy users found the $200 Pro plan lasts only about 12 hours. Complaints about declining code quality are also mounting.

Episode 3 · OpenAI Codex Hits 20 Million Active Users, Resets Usage Credits (2026-08-21, 4 posts)

OpenAI announced Codex surpassed 20 million weekly active users, a 20x increase in six months, and reset banked credits for all Codex and ChatGPT Work users while apologizing and investigating abnormal usage consumption.

Episode 4 · OpenAI Acknowledges Codex Cache Issue and Promises Compensation After Quota Complaints (2026-08-22, 7 posts)

On August 22, OpenAI Codex users erupted with complaints about abnormal credit consumption. Product lead Tibo Inoue's denial was rebutted by community evidence, and the incident was temporarily settled when OpenAI admitted cache hit rates had dropped and promised compensation.

Confirmed

  • Multiple users independently reported abnormal credit consumption: MaziyarPanahi said that with token usage exactly the same as usual, they burned through 60% of weekly credits in a single day, with earlier feedback to developers going nowhere; RileyRalmuto, a long-time developer building the same app on Codex, measured at least 20% higher usage for the same workload, suggesting either the model burns tokens more heavily or OpenAI quietly cut the credits.
  • Official follow-up admission: After community pushback, Tibo Inoue partially acknowledged the issue in a tweet, saying some users' cache hit rates this week were below previously stable levels, causing faster consumption, that the team is investigating with an update expected the next day, and that users would be compensated.
  • Tibo Inoue initially denied weakening Codex credits and accused users of fraud; his claims were refuted by community members with evidence and publicly flagged by X's Community Notes.

Not Yet Confirmed

  • The root cause of the abnormal consumption (model behavior changes or a caching mechanism issue) awaits the official investigation.
  • RileyRalmuto said consumption felt slightly better afterward but couldn't confirm it was fixed.
  • The specific scope and form of compensation (the post only mentioned compensating "B…", truncated) remains unclear.

Why It Matters

  • The incident shows how X's Community Notes can instantly correct vendor claims; executive denials can't hold up against user-provided evidence.
  • For developers relying on Codex, credit consumption directly affects cost and availability, and fluctuations in infrastructure parameters like cache hit rates materially impact user experience—transparent communication limits damage far better than denial.

Episode 5 · OpenAI Users Complain of Frequent Usage Limits Without Reset (2026-08-22, 2 posts)

Multiple OpenAI users report hitting usage limits prematurely, with no reset even after upgrading to the $100 plan, raising questions about OpenAI's quota management and service stability.