OpenAI’s Dots puts personal AI agents in the spotlight

OpenAI introduced Dots at its annual DevDay, pitching always-on agents that can proactively handle tasks and make decisions while browsing the web. WIRED tested Dots against Meta’s Muse as the companies compete to bring personal AI agents to a broader audience.

In the report’s testing, Dots reviewed the author’s ChatGPT history and suggested drafting a follow-up email, finding dinner spots and refining pitch drafts. OpenAI said users would initially have one Dot, with multiple agents planned later; it gave no firm timing.

At launch, Dots was limited to OpenAI Pro subscribers, whose plan starts at US$100 a month. Meta’s Muse was free through its app or website. OpenAI chief executive Sam Altman said Dots uses substantial computing resources and described his own agent handling morning tasks.

WIRED’s author found agents more capable at computer tasks than a year earlier, while stressing that the tools are not perfect. The report’s account reflects the author’s experience and does not establish how reliably the products perform across users or tasks.

The test also raised a practical question for users: how much access to grant an agent. OpenAI suggested connecting Gmail and Google Drive during Dots setup. WIRED’s author remained cautious about letting the agent access personal accounts; the report did not describe a security incident involving Dots.