OpenAI Develops 'Persistent' AI Agent for Codex
Overview
OpenAI is developing a new "persistent" AI agent capability for its Codex coding assistant. Code reviewed by WIRED reveals the feature is designed to allow the AI to continue working proactively without constant user input, functioning until it is explicitly "put to sleep."
How It Works
Unlike current AI assistants that typically respond only when prompted, the persistent agent can monitor tasks, make decisions, and take actions autonomously over extended periods. Users would maintain control through the ability to deactivate the agent when needed.
Implications
This development reflects a broader industry trend toward AI agents that can handle multi-step workflows with less human oversight. Such systems could potentially increase productivity for developers and other users by automating repetitive tasks and monitoring processes.
Context
The persistent agent builds on OpenAI's existing Codex technology, which was designed to understand and generate code. Making the assistant persistent could expand its utility for ongoing development projects and continuous integration workflows.