Your ChatGPT might soon have a new “person” added!
Elon Musk’s bot has been operational for over a month, Zuckerberg’s Muse has topped the App Store charts, and a small invite-only company has reached a $10 billion valuation.
In just a few weeks, OpenAI has become a follower in the "AI doing work for people" space.
According to The Information, OpenAI is developing features more directly comparable to Grok Bot and is internally discussing how to respond to Meta’s personal assistant, Muse.
This article labels OpenAI as a "follower." A year and a half ago, these three words would never have been applied to it.
In January 2025, when it launched Operator, it was among the first companies to turn "AI operating browsers for users" into a consumer-grade product.
ChatGPT brought generative AI into the hands of ordinary people, reaching over 1 billion weekly active users this summer, according to The Information.
After the report was published, long-time ChatGPT code sleuth and whistleblower blogger Tibor Blaho shared journalist Stephanie Palazzolo’s post, dropping one word: Aeon.
Aeon means "eternity" or "endless time" in Greek.
If this is truly the name OpenAI gave to the agent, it’s hard not to imagine that it’s also going to create an AI colleague for you that never clocks out and is always online.
Tibor also attached a screenshot to the post, showing code he pulled from the ChatGPT client.
In the code for the web version of ChatGPT, there is a line:
If a message is sent by a user, it is recorded under the user’s account name; if not sent by a user, it is recorded under Aeon.
In other words, in ChatGPT's view, Aeon and the user are two parallel identities.
Even as it was labeled a "follower," leaked code suggests it may already have reserved a desk for an AI colleague that can do your work inside ChatGPT.
Elon Musk's bot has been working for over a month.
Why is OpenAI restless?
Because Musk's bot has been working for over a month.
On August 11, SpaceXAI launched Grok Bot. Its official positioning is minimalist: your always-on agent team.
Note that you are dealing with a group of agents, not just one. Their biggest difference from ordinary AI assistants is that each has its own computer.
According to the official description, all bots under the same username share a persistent cloud-based computer, where files, browsers, and login sessions are stored. Even when you close your laptop and go to sleep, it continues running.
These bots will log in to your Gmail, Notion, financial software, and even legacy websites without APIs on their own.
They manipulate the mouse and fill out forms like humans, completing the entire task on their own—only coming to ask you for approval when they need your nod.
Even more aggressive is the coordination of multiple bots.
You can run multiple bots simultaneously—one managing your inbox, another handling sales outreach—with a "Chief of Staff" bot overseeing everything. They can autonomously assign tasks among group chats.
What Grok Bot is truly selling is simply this: handing off the work.
An extra "member" appeared in the code.
In the code screenshot forwarded by Tibor, aeonId and accountUserId appear side by side, indicating that within ChatGPT's system, Aeon may be a distinct, identifiable entity rather than a simple feature toggle.
It’s not clear from the screenshot whether it has the same identity or permissions as the user.
In the two most recent consecutive versions on Android, aeonId appears repeatedly, alongside a field named memberAeonIds.
"Member" means member; this naming is hard not to associate with the scenario of multiple bots coexisting in a single chat group, like Grok Bot.
Above Tibor's screenshot, there is another test instruction:
This test scenario is specifically designed to protect "Aeon-style behavior," where the capability state must be recaptured at each sampling step, rather than waiting until the next user input.
Ordinary ChatGPT operates one turn at a time, while "Aeon-style behavior" requires the system to reconfirm which capabilities are available at each step of a multi-step process.
This aligns with the logic required by an agent that needs to perform a sequence of actions.
OpenAI has not yet made an official statement; what’s currently available is only a media leak, a suspected codename, and a few lines of code fields—but hints are already beginning to emerge.
The ChatGPT menu only lists "Chat".
Most intriguingly, the media reports that OpenAI will not build a new product from scratch, but rather repurpose the existing agent technologies from Codex and ChatGPT.
This statement reveals an awkward truth: OpenAI already had the very capabilities that its competitors are now selling so successfully.
On July 9 of this year, OpenAI released ChatGPT Work, which, according to its official description, includes features such as executing tasks across applications, continuously handling complex projects, and scheduling tasks, along with integration of Codex technology.
The official documentation also states that when you select a cloud runtime, your tasks will continue running even after you close the app or shut your computer.
Talent has also been strategically positioned: In February, Peter Steinberger, the author of OpenClaw, joined OpenAI, and Altman said he will be in charge of building the next generation of personal agents.
The issue is that many OpenAI users may not be aware of these features.
It's like the kitchen can prepare any dish, but the menu only lists one: "chatting."
If the rumors about Aeon are true, the real question to ask is: beyond ChatGPT Work’s existing capabilities, will it gain a new “digital member” that can directly call out names and assign tasks?
This is likely its response to the Grok Bot.
Why can't OpenAI afford to lose?
According to The Information, ChatGPT's growth has significantly slowed over the past year, with OpenAI originally targeting one billion weekly active users last year, but only achieving it this summer.
This round, AI giants aren't just competing for rankings on model leaderboards—they're racing to claim intelligent agents that can complete tasks for you, now seen as the next growth engine.
Elon Musk's Grok Bot wants a cloud computer that stores all your login sessions.
Anthropic launched Claude Tag earlier, on June 23, allowing Claude to join Slack as a team member to take on tasks, targeting entire team channels.
Meta’s Muse reached #2 on the U.S. App Store Free Chart within two days of its launch and climbed to #1 by September 22, pushing ChatGPT down. It aims to be the chat app you open every day.
There's also Instinct, an invite-only startup whose valuation quadrupled in a month. It helps you book restaurants and make purchases via text message and phone calls, requiring your credit card details and password.
Four competitors, four strategies—all betting on your workflow.
Imagine next Monday morning: just after 8 a.m., you pick up your first cup of coffee for the day and glance at your phone.
In the past, you had to reply to a dozen emails, silence each annoying group message one by one, and then cram a pile of meetings into your calendar.
Later, after you finish this cup of coffee, all these tasks may already be done. All that will remain in front of you is a simple checklist waiting for your approval: whether to send this email, pay this bill, or attend this meeting.
Whoever holds this list controls your daily entry point to work.
Once a gateway becomes a habit, it’s hard to switch.
If a bot takes over your inbox, learns your email tone, and remembers which clients to prioritize, replacing it is much harder than switching chat apps.
But the hardest hurdle to overcome may lie beyond technology: trust.
How many ordinary people are willing to let an AI access their personal information, social media accounts, and even authorize it to make payments on their behalf?
This hesitation is the final hurdle all "AI colleagues" must overcome.
In the era of chat boxes, users are asking: Is AI smart?
In the age of agents, users are asking: Can I trust it with my tasks?
The only question left for OpenAI is: how to make one billion users feel comfortable handing over their keys.
Reference materials:
https://x.com/btibor91/status/2102110053721370778/photo/1
https://www.theinformation.com/newsletters/ai-agenda/openai-develops-features-counter-grok-bot-mulls-response-metas-muse
This article is from the WeChat public account "New Intelligence Yuan" (ID: AI_era), author: ASI Revelation; editor: Yuan Yu