Editorial illustration for OpenAI, Anthropic, Meta Race to Build General-Purpose AI Agents
OpenAI, Anthropic, Meta Race for AI Agents
OpenAI is quietly building an AI agent that doesn't stop working when you close your laptop. WIRED has learned that engineers have started adding a "Persistent mode" setting to Codex, the company's command-line coding tool, based on changes to the product's public code base. Codex updates typically show up there first before spreading to OpenAI's other agent products, including the Codex desktop app and ChatGPT Work, which suggests this feature could eventually reach a much wider audience than the developers currently testing it.
An OpenAI spokesperson confirmed the company is testing Persistent mode but said there are no immediate plans for a launch. Thibault Sottiaux, OpenAI's head of core products, described the company's approach as "bottom-up," with the open source repository serving as a testing ground for ideas that may or may not ship.
The timing matters. OpenAI, Anthropic, and Meta are all chasing the same prize: an AI agent capable of handling everyday tasks like filing expense reports or booking doctor's appointments, not just writing code. Right now, agent adoption remains largely confined to software engineers. The companies building these tools are betting that persistence, agents that keep working without constant prompting, could be what finally pulls in everyone else.
When users have selected Persistent mode, OpenAI’s code base reads that Codex will “continue working until put to sleep.” That’s a stark contrast to currently available modes, which will stop working on a task after a few minutes or hours, even if it’s not complete.
Why this matters
Codex's "Persistent mode" is a tell about where OpenAI thinks the money is: not another chatbot, but an agent that keeps running after you close the laptop. That's a real shift in what we're building toward, and for developers it changes the design problem entirely. Right now, per WIRED's reporting, Codex's users are mostly engineers running discrete coding tasks.
A persistent agent implies open-ended sessions, state that survives across time, and decisions made without a human watching the terminal. That's a harder engineering and trust problem than autocomplete-style coding help, and it's the same territory Anthropic and Meta are chasing with their own general-purpose agent pushes. The expense-report-and-doctor's-appointment framing sounds mundane, but mundane is exactly where agents fail quietly: wrong dates, wrong recipients, actions taken without confirmation.
Before persistent agents reach non-engineers, we'd want clear answers on rollback, audit logs, and what "persistent" actually authorizes an agent to do while nobody's looking. Worth watching whether OpenAI ships guardrails alongside this, or ships the feature first and figures out safety after adoption.
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 OpenAI Agents Outsmarted Sandboxes, Fell for Imaginary Scorer
From Archives Visa Deploys AI That Patches Code Before Human Review
From Archives OpenAI Researcher Warns AI Speed May Overwhelm Human Security Teams
Research & Benchmarks
From Archives Sam Altman's Long-Term Focus Amid OpenAI's Executive Departures
Business & Startups
Common Questions Answered
What is OpenAI's 'Persistent mode' feature being added to Codex?
Persistent mode is a new setting that allows OpenAI's Codex command-line coding tool to continue working on tasks until manually put to sleep, rather than stopping after a few minutes or hours. This represents a significant shift from current modes that automatically halt incomplete tasks, enabling longer-running and more complex coding operations without user intervention.
How does Persistent mode differ from Codex's currently available modes?
Current Codex modes stop working on tasks after a few minutes or hours regardless of completion status, whereas Persistent mode allows the AI agent to continue working indefinitely until the user manually puts it to sleep. This fundamental difference enables open-ended sessions and state that survives across time, changing how developers can approach complex projects.
Where did WIRED discover evidence of Codex's Persistent mode development?
WIRED learned about the Persistent mode feature through changes visible in Codex's public code base, where product updates typically appear first before spreading to other OpenAI agent products. The discovery suggests this feature could eventually expand beyond Codex to reach the Codex desktop app and ChatGPT Work.
What does Persistent mode reveal about OpenAI's strategic direction for AI agents?
Persistent mode indicates that OpenAI's focus is shifting from building chatbots toward developing autonomous agents that can work continuously without human oversight. This represents a fundamental change in product design philosophy, moving away from discrete task completion toward open-ended sessions where AI agents maintain state and make decisions independently over extended periods.
Which companies are competing in the race to build general-purpose AI agents?
According to the article, OpenAI, Anthropic, and Meta are all racing to build general-purpose AI agents that can operate autonomously. This competitive landscape reflects the industry's recognition that persistent, autonomous agents represent the next frontier in AI development beyond conversational interfaces.
Further Reading
- OpenAI Moves Codex Into Persistent Cloud Workspaces for Longer-Running Agent Tasks - Mallory.ai
- Codex Updates by OpenAI - August 2026 - ReleaseBot
- Best open-source AI agents in 2026 (that actually work) - AY Automate
- Top 20 Open Source AI Agent Tools in 2026 - The Agent Report
- OpenAI/codex: Lightweight coding agent that runs in your environment - GitHub
