OpenAI is advancing its vision of a future where millions of AI agents operate in the cloud, working alongside humans to drive productivity and economic value. In a recent episode of the "OpenAI Podcast," cofounder and president Greg Brockman and Codex engineering lead Thibault Sottiaux described a coming era in which autonomous AI agents, supervised by people, will collaborate on complex tasks—particularly in software development. Sottiaux emphasized that within a few years, large populations of AI agents could be running continuously in data centers, executing meaningful work under human guidance. "We have strong conviction that the way this is headed is large populations of agents somewhere in the cloud that we as humanity, as people, teams, organizations supervise and steer in order to produce great economical value," he said. This vision is now being brought to life with the launch of GPT-5 Codex, a new version of OpenAI’s AI coding tool. Unlike earlier models, GPT-5 Codex can operate for extended periods on intricate software projects, such as large-scale code refactoring, and seamlessly integrate into developers’ cloud-based workflows. Kevin Weil, OpenAI’s Chief Product Officer, shared on the podcast "Exponential View" that internal tools powered by Codex have already boosted engineering efficiency at the company. However, he stressed that human expertise remains essential. While AI can assist with coding tasks, the ability to understand context, logic, and intent still requires human judgment. As previous reports from Business Insider noted, attempts at "vibe coding"—relying solely on AI without technical knowledge—often result in flawed or nonsensical code. Brockman highlighted that OpenAI has been working on AI oversight strategies since 2017, focusing on how humans or simpler AI systems can guide more advanced models. Maintaining human control is critical, he said, to ensure that AI systems remain aligned with human values and organizational goals. Sottiaux described the emerging system as both complex and exciting—a multi-agent framework that can be steered by individuals, teams, or entire organizations. "It's a bit nebulous, but it's also very exciting," he said, underscoring the transformative potential of AI agents working in concert with people to solve real-world problems.