OpenAI is assigning a new position to its coding assistant, Codex. The platform was initially used as a passive code generator, but now the tech giant is more focused on transforming it into an autonomous collaborator. Competition is intensifying among AI coding tools. While Codex already has 3 million+ weekly users, the shift in strategy will help boost its utility in developer workflows.
The most notable upgrade is ‘background computer use.’ Codex can now view the screen, click, and type using its own cursor. This shifts the focus from simple API connections to practical, real-world application workflows and software traditionally excluded from automation. Multiple AI agents can also run simultaneously, without disrupting the user’s ongoing tasks.
The desktop app now includes a built-in browser, allowing users to comment directly on webpages to guide the AI’s actions. Codex has also gained image generation and iteration via the gpt-image-1.5 model, signaling a broader push into multimodal capabilities. OpenAI has also introduced more than 90 plugins.
Codex can now remember user preferences and past work, making future interactions more tailored. It can also schedule tasks and resume long-running processes over days or weeks, an attempt to reduce the need for constant human prompting.
The rollout is currently limited to desktop app users logged into ChatGPT, with macOS getting first access to computer-use features. Broader availability, including enterprise and regional expansion, is planned. Meanwhile, rivals like Anthropic are pushing competing tools, raising the stakes in this rapidly evolving space.
OpenAI’s update is about redefining the scope in place of incremental improvement. Codex is trying to operate within the developer's environment. Whether this move translates into real productivity gains or added complexity will depend on how reliably it performs.
Also Read: Google Brings Gemini to Mac, Built Using Its Own AI Coding Tool Antigravity