OpenAI Testing "Persistent Mode" for Codex AI Agents

OpenAI is developing a "Persistent" AI agent. [3] The company is testing a "Persistent mode" in Codex. [1][2] This mode is designed to let AI agents "continue working until put to sleep." [1][2] Furthermore, the system is designed to let agents proactively generate follow-up tasks. [1][2]

The introduction of a "Persistent mode" in Codex suggests a shift from standard prompt-and-response paradigms to continuous background execution. By allowing agents to operate until explicitly put to sleep and proactively generate follow-up tasks, OpenAI is exploring more autonomous developer tools. This functionality points toward systems that manage long-running workflows without requiring constant human intervention, moving beyond isolated interactions. What the sources don't address: How OpenAI will manage the compute costs, loop-prevention, and safety guardrails for agents that run continuously and generate their own tasks.