OpenAI's always-on assistant launched as "dots," not "o"

What we checked

• Will OpenAI publicly confirm "o" at DevDay and specify its permissions, availability, and persistent-work capabilities? partly confirmed. OpenAI did announce an always-on agent at DevDay on September 29, 2026, and it described the permissions, availability and persistent capabilities. But the product is called "dots" (one agent is a "dot"), not "o". None of the launch coverage I opened mentions the "o" name. [3][5][4][6]
• Did the leaked email identity (email_suffix "-o") ship as a capability at launch? unresolved. The "-o" email suffix came from a config string reported by BleepingComputer and TestingCatalog. None of the launch reports I opened say consumer dots have their own email address. For the enterprise preview, OpenAI says it gives "specialist dots" their own identity and credentials, which partly matches the leak's idea of a separate agent identity. [1][3][5]

What we found

• The pre-DevDay evidence for "o" was entirely leaks. BleepingComputer (September 27) based its report on a briefly visible Pro-plan benefit line and a config with display_name "o" and email_suffix "-o", and said OpenAI had neither confirmed nor denied the feature. TestingCatalog linked it, speculatively, to the internal "Aeon" codename and to a suspended @o account on X. [1][2]
• On September 29, OpenAI launched "dots," always-on agents inside ChatGPT that run on GPT-6 Astra. Each dot has its own cloud computer, which the owner can open, and a browser. Dots can run several projects at once, learn the owner's preferences from feedback, and reach more than 4,000 apps through OpenAI's plugins. None of the launch coverage I opened uses the "o" name. [3][4][5]
• Permissions are tiered. When a dot works unprompted (what OpenAI calls "proactive research"), its tools are read-only: it can't send messages, edit anything, or control the user's computer. Planned Custom Rules will let users allow specific actions, require approval for them, or block them. Sensitive actions such as changing a password stay with the user. SiliconANGLE adds that actions touching accounts or sharing information go through an "auto-review" check, and that a monitoring system can pause or stop an agent. With the owner's permission, a dot can also work on the owner's laptop. [5][3]
• Availability is restricted. Dots are rolling out to ChatGPT Pro and Business Premium subscribers, and some accounts may wait several days. Pro subscribers in the EEA, Switzerland and the UK are excluded at launch. Enterprise, Edu and Healthcare workspaces get a beta that an admin must enable. One dot is included at no extra cost, and per OpenAI's help docs as reported by SiliconANGLE, dot usage won't count against plan allowances for the first month. Sources disagree on which Pro tier qualifies: BGR says the $100 Pro tier is excluded, while the leak placed the feature on the $100 plan. [3][5][6]
• Dots can be reached in ChatGPT on web, desktop and mobile, and in Slack and Microsoft Teams, with context carried across channels. Texting is "coming soon," and teams of dots are a stated future plan. "Specialist dots," which get their own identity and credentials for a defined organizational job, are in enterprise preview, and a Microsoft Agent 365 integration is planned. OpenAI's developer-forum recap mentions no API access to dots themselves. Developers instead get a separate Agents API in public beta, with hosted execution, memory and computer use. [6][3]
• Dots are already being updated. An October 9 post on OpenAI's developer forum says users can now create a dot in the iOS and Android apps, and that a dot can start work in Codex, follow up on existing Codex threads, and review or edit Scheduled Tasks in ChatGPT Work. [7]

What it means

Stop calling the product "o" and drop features that only the leak supported, especially email. The always-on behavior and the DevDay timing in the leak were right. The name was wrong, and nothing I found confirms that consumer dots have their own mailbox. If you build on or compete with ChatGPT, the main thing to study is OpenAI's permission model. Unprompted background work is read-only by default. Write actions go through rules, approvals and an auto-review check. Some sensitive actions can never be delegated. Your own persistent agents should be held to at least that standard, because users will now expect it.

For platform planning, dots are a consumer and workspace product, not an API. If you want comparable persistent agents in your own product, OpenAI's recap points to the Agents API (public beta) and Codex Cloud. The enterprise "specialist dots" preview, with its own identities, credentials and planned Agent 365 management, shows OpenAI moving into managed agent identities. Vendors selling agent identity, governance or scheduling should treat that as direct competition. Check eligibility before you promise users access: there are regional exclusions in the EEA, Switzerland and the UK, and reports disagree on which Pro tiers qualify.

Still open

• Why does the shipped product differ from the leaked "o" naming and "-o" email suffix? Was "o" a placeholder, an abandoned name, or a separate product still to come?
• Which Pro tiers include a dot? BGR and the leak disagree about the $100 tier, and OpenAI's own pricing page couldn't be fetched to settle it.
• What will dot usage cost after the first month, and what will additional dots or extra capacity cost? Prices were not disclosed.
• When will Custom Rules, texting, teams of dots and EEA/UK/Swiss availability arrive?