< Back to all clusters
[TECHNOLOGY] · United States · 11 sources

started · updated

OpenAI tests persistent mode for Codex AI coding agent

OpenAI is testing a new 'Persistent mode' for its Codex coding agent, which would allow the AI to operate autonomously for extended periods. Unlike current models that stop after a task is completed or after a short period of inactivity, a persistent agent would continue working until it is explicitly 'put to sleep.'

Code discovered in the Codex command-line interface suggests the feature includes a 'proactivity' function. This would enable the agent to independently create and execute follow-up tasks across different sessions, using past user interactions and context to drive progress. The mode appears within a 'reasoning effort' menu, allowing users to control the level of computing power and time allocated to the task.

While internal tests have reportedly shown agents running for approximately 25 hours on complex tasks, OpenAI has stated there are no immediate plans for a broad launch. The development represents a shift from reactive chatbots toward proactive, always-on AI agents capable of managing long-term projects.

Entities

Anthropic · Codex · OpenAI · Thibault Sottiaux · Wired

Claims

What the coverage asserts, and how many sources carry each claim.