From the paper of Wednesday, September 30, 2026
OpenAI launches Dots, always-on AI agents that work on their own
OpenAI announced Dots on Tuesday at its DevDay event in San Francisco: personal AI agents that run continuously, take on multistep projects, and connect to apps a user already has. The company calls them "remarkably capable, always-on agents built to handle everything," powered by a model it names GPT-6 Astra.
Each Dot gets its own cloud computer and browser and can reach more than 4,000 apps through OpenAI's plugin system, the company said. A user names a first Dot and talks to it through ChatGPT, Slack, or Teams. Wired reports that Pro users can join a waitlist to text a Dot over iMessage or RCS, and that texting is coming more broadly later.
The pitch is accumulation. A Dot learns a user's preferences and standards over time, OpenAI said, and can work on several projects at once without being directed at each step. In demos the company shared, one Dot saw that its user's calendar showed they would be working through dinner and messaged two GrubHub options with prices; in another, a user worked with the agent to launch a website. Inside OpenAI, the company said, a bug appearing in Slack sends Dots investigating, and an early tester's Dot noticed he had forgotten to invoice a publication, prepared the invoice, and sent it once he approved.
OpenAI says users stay in control. Dots get built-in rules about when to act alone and when to ask, a Custom Rules tool lets users allow, gate, or block specific actions, and sensitive tasks like changing a password stay with the person. When a user is not actively working with it, a Dot looks for ways to help in the background using read-only access to connected apps — a feature OpenAI calls "proactive research" — and cannot send messages or change app content while doing it. Company monitoring can pause or stop a Dot. "Dots can still make mistakes, so always review consequential work," OpenAI wrote.
Wired reports that automating parts of your life means trusting the agents not to share information you didn't intend to share or be tricked by an online adversary into acting against your interest, and notes that an account setting allowing model training on your data carries over to the agent's interactions.
Dots begin rolling out Tuesday to ChatGPT Pro and Business Premium subscribers in eligible markets, with the first Dot included at no extra cost. Wired puts the Pro tier at $100 a month. Enterprise, Edu, and Healthcare users get a beta when a workspace administrator enables it. More Dots per user, and the option to scale each one's output, come later, OpenAI said.
For companies, OpenAI previewed "specialist Dots" with their own identities and credentials, assigned to defined jobs. It says it is testing them in procurement, invoice processing, email marketing, customer support, and commercial contracting, and is working with Microsoft to manage them through that company's Agent 365 controls. TechCrunch notes much of this was already possible through Codex and similar tools; Dots packages it in a new product.