OpenAI has announced a major upgrade to its AI coding assistant, Codex, transforming it from a mere code generator into a more versatile and autonomous partner for developers. This update arrives amid increasing competition in the AI coding tool market.
Autonomous Computer Interaction
A standout new feature allows Codex to operate a computer in the background. The AI can now view the screen, click elements, and type, enabling it to work across various applications autonomously. This capability is particularly useful for tasks such as testing applications, making frontend design adjustments, or utilizing tools that lack direct APIs.
OpenAI highlighted that multiple AI agents can run concurrently on a user's Mac without interfering with other ongoing work. This parallel processing capability is a boon for developers needing to iterate on frontend changes or work within applications without exposed APIs.
Enhanced Functionality and Integration
The updated desktop application now includes a built-in browser, allowing users to guide the AI by commenting directly on webpages. Furthermore, Codex can now generate and edit images, leveraging the advanced capabilities of the GPT-image-1.5 model.
To further expand its utility, OpenAI has integrated over 90 new plugins. These plugins combine various skills, application integrations, and MCP servers, providing Codex with richer context-gathering abilities and a wider range of actions across different tools.
Memory and Automation Capabilities
Beyond active interaction, Codex now boasts improved memory and automation features. It can remember user preferences and past work information, significantly enhancing its performance on future tasks. Users can also schedule tasks for later execution, and Codex can automatically resume long-running work over extended periods, boosting productivity.
These new features are progressively rolling out to users of the Codex desktop app who are signed in with ChatGPT. Personalization features, including context-aware suggestions and memory, will soon become available to Enterprise, Edu, and users in the EU and UK. Initially, the computer control feature is exclusive to macOS, with plans for a rapid expansion to EU and UK users.