OpenAI CEO Sam Altman: Codex Active Users Exceed 4 Million, Quotas Reset to Target 10 Million Developers
OpenAI CEO Sam Altman announced on social media that the active users of the AI coding tool Codex have reached 4 million, with the milestone of 3 million achieved in less than two weeks. To celebrate this new milestone, the official will reset usage rates and quotas for users again. Previously, OpenAI had temporarily relaxed or reset Codex limits multiple times upon reaching 2 million and 3 million users, and has promised in internal communities and short videos to continue doing so for every additional 1 million users until reaching 10 million, using "quota resets" as a marketing strategy to enhance new user experience and retention.
Public reports indicate that since the release of the GPT-5.3 Codex version, downloads of the Codex desktop and VS Code extensions have rapidly increased. In March, OpenAI disclosed that weekly active users of Codex had surpassed 1.6 million, and the user base doubled within three months, continuing to accelerate. Some developer communities have reported that their experience in code generation, refactoring, and automated toolchain invocation is now comparable to or on par with competitors like Claude Code. OpenAI has also established a discussion area on the developer forum specifically for "Codex Quotas and Rate Limits" to address complaints from some users regarding "opaque reset times and random quota changes," attempting to find a balance between increasing activity and controlling costs and abuse.
Source: Public Information
ABAB AI Insight
This update essentially conveys two signals: first, Codex is transitioning from an "ancillary tool" to an independent traffic and revenue engine within the OpenAI ecosystem; second, OpenAI is willing to "burn servers" through frequent quota resets to gain developers' time. For a product centered around developers as its core user group, reaching 4 million active users and adding 1 million in less than two weeks indicates that AI coding has moved from the early curiosity stage to the "everyday workflow tool" stage—many programmers, data engineers, and script developers are relying on it to complete actual tasks rather than just trying it out occasionally. In the competitive environment with OpenAI, Anthropic, Google, xAI, etc., "who occupies the developers' IDE and daily muscle memory" is seen as a precursor to the next round of competition for AI agents and enterprise-level tools.
Frequent "rate limit resets" represent a typical supply-side game: for OpenAI, computing power and inference costs are real expenses, and the more developers use it, the greater the short-term profit and loss pressure. However, if they can periodically relax quotas on a large scale, allowing developers to get accustomed to using Codex as an essential tool, they can later monetize through enterprise packages, team versions, and deeper toolchain integrations, potentially yielding long-term benefits that far exceed initial costs. This is similar to the early cloud storage and developer platform strategy of "crazy free credits": first locking in mindset and workflow, then charging for excessive use, team collaboration, and enterprise-level features. The rhythm of quotas being made public and ritualized (resetting for every additional 1 million users) also reinforces a "community-level growth story," turning user growth into a part of participation and benefits.
On a deeper level, the rapid expansion of Codex users will have long-term impacts on the labor structure and tool landscape of the software industry. As AI coding tools penetrate mainstream IDEs, the truly scarce skills are shifting from "handwriting code speed" to "problem decomposition ability, architectural design, and collaboration with AI capabilities." Mid-level repetitive labor is systematically compressed, while senior engineers leverage AI to manage more code and systems. In this structure, tools like Codex, which have usage data from millions of high-frequency developers, have the opportunity to become the "invisible standard setters" of the next generation of engineering practices: they will solidify certain architectural and library usage habits through recommended templates, default modes, and best practices, thereby shaping the technological paths and lock-in effects of the software industry. Each time Altman announces "user +1 million, quota reset again," it appears to be user operations, but in essence, it accelerates the process of restructuring technology and labor.