OpenAI launches Dots, autonomous agents for Pro subscribers only

At its annual DevDay yesterday, attended by roughly 2,500 developers, OpenAI unveiled Dots — a new family of agents that run continuously in the background, initiate actions and make autonomous decisions while scanning the web. Chief executive Sam Altman walked the audience through his own experience with the agents, framing the release as a pivot away from the chatbot format, which he said is no longer the hot paradigm for developers, toward personal agents vying for control of a new market.

The author set up a Dot named Toolie, styled as a pink frog, and tested it late at night. Asked to scan their ChatGPT history for three immediate tasks, Toolie surfaced a data request that needed a follow-up email and offered to draft the reply; flagged that an upcoming "gay-cation" lacked restaurant plans and suggested options; and proposed reviewing research notes and polishing pitch drafts before sending them to editors — an offer the author declined because that step is their favorite part of the workflow.

Access at launch is restricted to subscribers of OpenAI's Pro plan, which starts at $100 a month. Each user can run a single Dot, with the company promising support for multiple agents later. By contrast, Meta's Muse is free to download or use on the web. Altman argued that the quality of Dots, powered by the GPT-6 Astra model, is the differentiator, and acknowledged the product consumes significant compute resources — though the long-term goal is to reach billions of users.

Both OpenAI and Meta are leaning deliberately into a "cute" aesthetic. Muse adopts a Labubu-inspired look; Dot appears as a two-eyed fur ball. Researchers cited in the piece say the anthropomorphic design may neutralize cultural resistance to generative AI and foster affection for the software. The experience mimics a group chat: agents reply with emojis — thumbs-up, heart — to acknowledge a task before executing it.

Unlike a chatbot that waits for a prompt, the new agents are built to message users on their own initiative, especially for recurring tasks. Altman described his own agent as something that "handles things, alerts if something truly urgent comes up, and responds to what it can" — a description that underscores the shift from a passive conversational interface to a system that manages tasks in the background.