AI agents are OpenAI’s bet on the future of software engineering.
During Monday’s “OpenAI Podcast,” cofounder and president Greg Brockman and Codex engineering lead Thibault Sottiaux presented a vision of massive networks of human-supervised autonomous AI agents that can operate continuously in the cloud as full-fledged partners.
“We really believe that the direction this is taking is toward having big populations of agents in the cloud that we, as individuals, groups, and organizations, oversee and manage to provide significant economic value,” Sottiaux stated.
So, this is what it will look like if we’re going in a few years,” Sottiaux continued. “To perform beneficial tasks, millions of agents labor in our data centers and those of other businesses.
GPT-5 Codex was released on Monday by OpenAI. According to OpenAI, GPT-5 Codex can work on complicated software projects, such extensive code refactorings, for hours at a time, unlike previous versions, and it can seamlessly integrate with cloud developers’ workflows.
During tech entrepreneur Azeem Azhar’s “Exponential View” podcast, OpenAI CPO Kevin Weil stated that internal tools such as Codex-based code review processes helped its programmers work more efficiently.
This does not imply that human programmers will become obsolete. Despite successful instances of “vibe coding,” engineers and professors of computer science have told that it is clear when an individual utilizing the AI agent lacks coding skills.
According to Brockman, as AI agents take increasingly challenging jobs, supervision will remain crucial. Since 2017, he added, OpenAI has been planning how humans or even less advanced AIs may keep an eye on more potent AIs to ensure supervision and “be in the driver’s seat.”
Sottiaux stated, this is where it’s headed for him—figuring out this entire system, then making it multi-agent and steerable by individuals, teams, and organizations, and aligning that with the whole intent of organizations. Although it’s a little unclear, it’s also really thrilling.
