With two days left before DevDay, an unreleased OpenAI agent appears to have shown up on the company’s own product surfaces. According to findings shared by TestingCatalog, ChatGPT’s internal configuration includes “O” as a display name, alongside a dedicated “-o” email suffix. That points to an agent that may carry its own email identity and continue working outside the chat window.
The name had also briefly appeared on the upgrade page for the $100-a-month Pro plan, where it was listed as a subscription benefit. Taken together, the clues suggest OpenAI may be preparing an always-on agent called “o.” The company has not confirmed it. DevDay on Sept. 29, which begins early Sept. 30 in Taiwan time, may be when the product is clarified.
Configuration references appeared before any official announcement
TestingCatalog said “Aeon” had previously been tied to OpenAI’s work on an always-on agent. Inside OpenAI, that name has also been used for custom agents in ChatGPT Workspace accounts. Based on that, TestingCatalog speculated that “o” could be a consumer-facing version built on the same foundation.
Two other details have added to the speculation. First, the single-letter naming style has drawn comparisons with a Bloomberg report from August by Mark Gurman about an OpenAI donut-shaped speaker. That device was described as having no screen, moving parts, and a possible price above $300. Second, the @o account on X is currently suspended, which has prompted guesses that the handle may be reserved for a launch.
In a post on X, TestingCatalog wrote that an upcoming always-on assistant from OpenAI will be named “o.” The post said references to it appeared briefly on the ChatGPT upgrade screen for some users, and that internal configuration points to “o” as a display name and “-o” as an email suffix, indicating support for persistent operation.
Meta has already launched Muse
OpenAI would not be entering an empty field. Meta formally introduced its personal AI agent Muse on Sept. 8. Based on the information cited in the report, Muse can keep working after a user closes the app, although actions such as sending emails or placing orders still require user approval.
Muse runs inside a dedicated Muse Secure VM, with a Sentinel agent acting as a guardrail. Actions that have not been approved are not sent onto the internet. At Meta Connect this week, Muse added another capability: its own email address, allowing it to be included in email threads and to control Mac desktop apps.
On currently available evidence, “o” has only been linked to a “-o” email suffix, while Muse already has a full email address. The pricing contrast is also clear. Most Muse features are free, while OpenAI listed “o” as a benefit on its $100-per-month Pro plan. The report also noted OpenAI’s $200 Pro tier and a rumored Pro Max tier priced at $500 per month.
Whether OpenAI’s always-on agent can carve out its own place against that head start remains unanswered for now.