OpenAI's New 'Dots' Are AI Helpers That Never Sleep

OpenAI unveiled a new kind of AI assistant on Tuesday at its developer conference in San Francisco. They're called Dots, and the idea is simple: instead of a chatbot that only answers when you type, a Dot keeps working in the background. Assign it a project — plan a trip, build a website, hunt down a cheaper phone plan — and it keeps checking the web and nudging you until the job is finished. Visually, Dots are cute blobs you can name and personalize, so they feel less like software and more like a small helper.

The most talked-about demo showed the practical upside. A Dot noticed its user's calendar said they'd be working straight through dinner, so it messaged two GrubHub options with prices and asked which one to order, and when. You can reach your Dot through ChatGPT, Slack or Microsoft Teams, and your information carries over between them. Pro users can even join a waitlist to text their Dot by iMessage or Android messaging.

Dots launch first for ChatGPT Pro subscribers at $100 a month, with wider access expected over time. You start by controlling one Dot, and OpenAI says multiple agents at once are coming. The company also announced "specialist Dots" for businesses — versions tuned for accounting, email marketing and legal research — a clear sign that AI "coworkers" are moving into real jobs, not just hobbies.

The catch is trust. To be useful, a Dot needs access to your calendar, messages and connected apps. OpenAI says a Dot asks permission before sensitive actions like installing software or changing passwords, and you can set custom rules about what it should never do. Still, anything that runs nonstop and reads your data can leak information or get tricked by an online scammer. And Meta's rival agent, Muse, already topped the app download charts.