Anatomy of ChatGPT Work: An Agent for 1 Billion Users (18-min read)

OpenAI launched ChatGPT Work, an agent product targeting knowledge work, on July 9. This launch was a major update that simultaneously included three new models and 14 configurations, integration of the ChatGPT and Codex desktop apps, and the popularization of cloud agents.

It is reported that users of Work and Codex surpassed 10 million within three weeks of the launch. ChatGPT is also estimated to be approaching 1 billion monthly active users and 1 billion weekly active users, indicating that Work is not a product for a select group of advanced users, but rather shows the direction of ChatGPT, which will be used by a massive user base in the future.

Currently, Chat and Work are provided as separate modes within ChatGPT, but OpenAI co-founder Greg Brockman stated that the two modes would be integrated by the end of the year. Work can be seen as a combination of the following elements.

• ChatGPT's conversational interface
• Codex app and agent execution harness
• Existing Codex cloud agents
• ChatGPT agent
• Related products and features such as Atlas and OpenClaw

The product structure is not yet consistent. Features differ across desktop, web, and mobile versions, and the experience varies depending on whether cloud mode is used. Therefore, Work can be interpreted not as a single new product, but as an integration phase that bundles Memory, Proactivity, Scheduling, Browser Use, Plugins, Skills, and Tools into a single user experience.