OpenAI Is Reportedly Building an Always-On AI Agent Called 'O'

 

OpenAI may be about to change how we use ChatGPT forever. According to a new leak spotted by TestingCatalog, the company appears to be preparing an always-on AI agent codenamed "O" — and it could be announced as soon as DevDay on September 29.

What the leak reveals

References to "O" were found scattered across ChatGPT's configuration files, including "O" as a display name and an email suffix configured as "-o". That suffix is the interesting part: it suggests the agent would ship with its own email identity from day one, able to send and receive mail on your behalf. On top of that, "O" briefly appeared as a listed benefit on the ChatGPT upgrade page for the $100 Pro plan before disappearing — the classic accidental leak.

A persistent agent

Put together, the clues point to a persistent agent: software designed to keep working outside a normal chat session. Think recurring research, communications, monitoring tasks, or background work that continues while you are away — and greets you with results when you return. The project may be related to "Aeon," a name previously associated with OpenAI's always-on agent work. Aeon is reportedly being used internally around custom agents for ChatGPT Workspace accounts, which hints that "O" could become the consumer-facing version built on the same infrastructure. Two more clues support the branding: OpenAI has been rumored to be exploring a donut-shaped hardware device, making the single-letter name an interesting potential match, and the @o account on X currently appears suspended — possibly reserved for a future launch.

Why the timing makes sense

This fits OpenAI's broader shift from conversational ChatGPT toward software that carries out longer-running work with less supervision. And the timing makes sense: Meta has already pushed this category forward with its Muse agent, putting real pressure on OpenAI to show what its own persistent agent can do. With DevDay only days away — Sam Altman is opening the keynote in San Francisco — "O" is now one of the most notable unreleased ChatGPT features to watch, alongside a recently leaked Pro Max subscription plan. What we still don't know: the supported tasks, permissions, scheduling, memory, pricing, or availability. Leaks like this sometimes describe products months away, and sometimes they confirm what's already shipping.

Why it matters

The jump from chatbot to always-on agent is the biggest shift in how we use AI since ChatGPT itself. A chatbot waits for you; an agent works for you. Whoever nails the always-on experience first — OpenAI, Meta, or someone else — will define what "using AI" means next year.

Our take

Persistent agents are coming whether we're ready or not. The question isn't if your AI gets an email address — it's whether you'll be excited or nervous when it does. We're going with excited. Nervously excited.