OpenAI Personal Assistant Agent Leaks Again: This Time It Might Be Called Dots

Beating AI News Flash: OpenAI's next-generation personal Agent may be called Dots. Leaked unpublished copy from M1Astra and ChrisGPT shows that it can execute tasks on its own. Users can reach it via messages, phone calls, Slack, and email, and after approval, it can also directly make purchases on the user's behalf.

Dots can also determine on its own when to act directly and when to first seek user approval, and users can set rules in advance. Each Dot has an independent VM, with an interface called "Your dot's computer." It can also gain access to the user's computer, which users can revoke at any time and then reauthorize through the desktop app.

The same batch of copy also mentions the ability to securely save passwords in the browser, allowing ChatGPT to automatically log in to websites. Each Dot will also have an independent 3D avatar, initially available only to Pro users.

A few days ago, the ChatGPT Pro upgrade page briefly showed "o, your always-on assistant," and the client configuration also showed the `o` and `-o` email suffixes; in previously leaked code, Dots was also linked to internal names such as Orbit and Aeon. OpenAI has not yet officially announced this product, so whether `o`, Dots, and Aeon are different products, internal codenames, or names for the same product at different stages cannot currently be confirmed.