Report says OpenAI is building a more persistent Codex AI agent

In short: OpenAI is reportedly developing a version of its Codex AI agent that can keep working on its own until a user tells it to stop.

What happened

WIRED reports that OpenAI is working on a “proactive” and “highly persistent” version of Codex. Codex is an AI agent, meaning software that can take actions to complete tasks, not just answer questions. In this case, Codex is aimed at programming tasks.

According to WIRED, code it reviewed suggests OpenAI is building a feature that lets Codex continue working until it is “put to sleep.” Think of it like a helpful assistant who stays on the job and keeps checking for what to do next, instead of waiting for you to come back with a new request.

The report does not say when this feature will be released or who will be able to use it first. OpenAI has not publicly announced the feature in the information provided here.

Why it matters

A more persistent agent could save people time by handling longer, multi step work without constant back and forth. But it also raises basic questions about control and safety. If an AI tool can keep going in the background, users and companies may need clearer ways to pause it, review what it did, and limit what it is allowed to access.

Source: Wired