OpenAI Shifts Codex Toward Persistent Cloud Development

The shift toward persistent environments means cloud tasks are no longer treated as temporary, disposable sessions. Teams can now operate within shared workspaces that retain approved settings and permissions, accelerating project start times. This infrastructure is bolstered by a refreshed command-line interface, which now supports voice-directed task management, and a new terminal interface designed to improve readability during extended coding sessions.

OpenAI has also integrated Codex directly into the ChatGPT desktop application to streamline the code review process. Users can generate summaries, probe for potential bugs, or trigger automated reviews while away from their machines before finalizing feedback on GitHub or GitLab. Complementing these workflow enhancements, the new Codex Security Cloud provides automated repository scanning. It identifies vulnerabilities, filters out duplicate findings, and drafts cloud-based fixes even when the user is offline, leveraging models from OpenAI’s Daybreak Blue initiative.

Broadening the platform’s utility for enterprise workflows, the company released a Decisions API for real-time logic processing via the Luna model. Additionally, the updated Agents API now supports direct computer use and integration with Amazon Bedrock, enabling developers to build and deploy agents that operate entirely within AWS infrastructure.