Dots, introduced by OpenAI, is a new type of AI agent designed to keep working on projects even when users are not actively chatting with it. Dots takes the familiar question-and-answer model of ChatGPT toward persistent assistants that can carry context across tasks, use connected apps and proactively look for useful work.
Each dot is powered by GPT-6 Astra and gets its own cloud computer and browser. It can connect to more than 4,000 apps through plugins, allowing them to work with information and tools from services a user has authorized.
Instead of having to create a new conversation for every task, users can give a dot an ongoing project and continue adding work while it handles other assignments. OpenAI gives examples including monitoring customer feedback and preparing software fixes, updating launch materials as plans change, rerunning research analysis when new data arrives and turning interview transcripts into draft content.
Those are examples of how OpenAI expects the system to be used rather than preconfigured abilities available in every situation. What a dot can actually do depends on the apps, permissions and tools available to it.
Dots can work without constant prompting
One of the bigger differences from regular ChatGPT is what OpenAI calls “proactive research.” When a user isn’t actively interacting with a dot, it can examine information from connected apps to identify work that may need attention. OpenAI says this background mode is restricted to read-only tools, meaning the agent cannot independently send messages, modify app data or take control of a browser or computer while operating this way.
Dots can take broader actions when users are working with them, subject to permissions and approval rules. Users can create custom rules specifying which actions a dot may perform automatically, which require approval and which are blocked.
OpenAI says potentially consequential actions are also checked by an automated review system, while certain sensitive tasks, such as changing a password, must remain under direct user control. Users can inspect a dot’s cloud computer and follow its activity. Connecting a personal computer is optional rather than required.
Available first on paid ChatGPT plans
Dots are beginning to roll out in eligible markets to ChatGPT Pro and Business Premium customers. Enterprise users, including Edu and Healthcare customers, can access a beta when enabled by their workspace administrator.
Pro and Business Premium plans initially include one primary dot. OpenAI says users will eventually be able to add more, while it is also testing “specialist dots” for businesses that can be assigned defined organizational roles with their own credentials and system access.
Dots can be contacted through ChatGPT on desktop, web and mobile, as well as through Slack and Microsoft Teams. OpenAI says texting support is coming later. The company also says conversations with a dot don’t count against normal ChatGPT usage limits, although tasks it launches in products such as Codex or ChatGPT Work continue to use the limits associated with those services.
The launch puts OpenAI into the growing market for persistent personal AI agents, alongside products such as Meta’s Muse. The more important change, however, is the model of interaction where instead of waiting for every new prompt, Dots are designed to maintain an ongoing understanding of a user’s work and continue acting on it over time. That also makes permissions more consequential. OpenAI cautions that dots can still make mistakes and recommends reviewing important work, particularly as users give persistent agents access to more of the apps and information they rely on.
Leave a Reply