OpenAI Reworks Codex Into a Tool for Autonomous Workflows

OpenAI has rebranded its Codex tool, emphasizing its ability to run workflows independently of constant human involvement. According to the company, the updated Codex can operate autonomously “for hours, if that's what it takes.”

Working for the user and with them

The tool's new positioning is built around the idea of two modes of interaction: Codex can either carry out tasks on its own or work alongside the user in an interactive mode. This approach reflects a broader trend in the industry — the drive to turn AI assistants from mere advisors into doers capable of seeing tasks through to the end without constant oversight.

Long autonomous sessions

The key feature of the update is the tool's ability to sustain long, independent work sessions. This suggests that OpenAI is betting on the model's reliability and resilience in carrying out multi-step tasks, rather than just on how quickly it responds in a dialogue.

Details about the specific capabilities, limitations and use cases of the updated Codex have not yet been disclosed in full. Even so, the very fact of the rebranding and the shift in emphasis toward autonomy indicates that OpenAI sees the future of such tools precisely in the independent execution of work tasks — from writing code to broader office scenarios.

The update fits into the company's overall course toward developing agentic systems capable of acting with minimal human intervention, something that had already been apparent in other OpenAI products.

Source: Ars Technica