Editorial illustration for OpenAI's Codex Gains a "Persistent Mode" in New Code
OpenAI's Codex Gets Persistent Mode for Continuous Coding
OpenAI's Codex Gains a "Persistent Mode" in New Code
Codex has always worked in bursts, spinning up, finishing a task, then shutting down. Code spotted by WIRED suggests OpenAI wants to change that. Buried in publicly available code for the AI coding agent is a "Persistent Mode," built to keep the agent running until someone actively puts it to sleep. That's a departure from the current setup, where sessions end after minutes or hours.
The code points to something bigger than a longer timer. A "proactivity" feature would let Codex generate its own follow-up tasks, carry work across sessions, and message users unprompted, no request needed. OpenAI has capped what it can touch outside a user's own system, requiring approval for anything external. The company confirmed the tests to WIRED but said there's no launch date set.
TIME reported earlier on the same broader concept: "persistent agents" acting as virtual coworkers handling tasks over long stretches without hand-holding. Sam Altman has talked repeatedly about turning ChatGPT into something closer to a full personal assistant, and an always-on Codex would fit that ambition. Whether users, or OpenAI's own security team, are ready for an agent that acts without being asked is a separate question entirely.
OpenAI is building a "Persistent Mode" for its AI agent Codex, according to publicly available code WIRED found. Unlike earlier modes that shut down after minutes or hours, this agent is designed to "continue working proactively until it is 'put to sleep.'" TIME had previously reported on "persistent agents," virtual coworkers that handle tasks on their own over long stretches of time.
Why this matters
An agent that runs until someone tells it to stop is a different product than one that answers a prompt and quits. For developers, that shift changes what "using Codex" even means: less prompt-and-wait, more monitoring a process that might file pull requests, open tickets, or touch production code while you're at lunch. The "proactivity" feature WIRED flagged suggests OpenAI wants Codex initiating work, not just executing it.
That's useful for founders trying to cut engineering overhead, but it also means someone has to answer for what an unsupervised agent does at 3 a.m. Persistent, self-starting systems need sleep switches for a reason, and "put to sleep" language in the code hints OpenAI already knows this needs a kill mechanism. We'd want to know what guardrails ship alongside this, not after.
TIME's earlier reporting on "persistent agents" as virtual coworkers points to where this is headed industry-wide. Codex may just be the first version of it we can actually inspect in code before OpenAI announces it officially.
Related Reading
Browse all OpenAI news →
Open Source D&B rebuilds 642 million‑business database after AI agents hit limits
Open Source OpenAI's GPT Transcribe Cuts Error Rate to 3.31%, Improving on GPT-4o
Open Source Study: 6.7% of Deepfake Requests on Hugging Face Targeted Children
Related Topic OpenAI researcher quits, citing distrust over ad‑driven engagement metrics
LLMs & Generative AI
Related Topic OpenAI launches GPT-Image 1.5 with precise editing for enterprise visuals
LLMs & Generative AI
From Archives Anthropic, OpenAI to speak at TechCrunch Disrupt 2026
From Archives Hugging Face Launches USD 399 Microduck Robot Amid Reported Nvidia Acquisition Talks
From Archives Barret Zoph Leaves OpenAI for Google After Five Months
Business & Startups
From Archives OpenAI, Anthropic, Meta Race to Build General-Purpose AI Agents
Common Questions Answered
What is the main difference between Codex's new Persistent Mode and its current operating model?
Currently, Codex operates in bursts, spinning up to complete a task and then shutting down after minutes or hours. The new Persistent Mode is designed to keep the agent running continuously until someone actively puts it to sleep, fundamentally changing how the AI coding agent operates.
How does the 'proactivity' feature in Codex's Persistent Mode change developer workflows?
The proactivity feature enables Codex to generate and initiate work on its own rather than just executing tasks in response to prompts. This means developers would shift from a prompt-and-wait model to monitoring a process that could autonomously file pull requests, open tickets, or modify production code while they're away.
What evidence did WIRED find about OpenAI's plans for Codex's Persistent Mode?
WIRED discovered references to Persistent Mode buried in publicly available code for OpenAI's AI coding agent. The code indicates that OpenAI is building functionality to keep the agent running proactively until it receives a command to stop, representing a significant departure from the current session-based approach.
Why is a persistent, self-starting AI agent significant for software development teams?
A persistent agent that works autonomously over long stretches of time represents a different product category than traditional prompt-response tools. For founders and engineering teams, this shift could reduce manual coding work by enabling the AI agent to handle tasks independently, similar to having a virtual coworker available continuously.
Further Reading
- OpenAI Is Developing a ‘Persistent' AI Agent - WIRED
- OpenAI tests persistent Codex mode that can keep working across sessions - Crypto Briefing
- Nevertheless, OpenAI Persists With New Always-On Agent - Gizmodo
- OpenAI introduces Codex, its first full-fledged AI agent for coding - Ars Technica
- Codex is now generally available - OpenAI
