What are OpenAI's new 'dots' agents — and should you use one?

OpenAI unveiled dots — always-on AI agents with their own cloud computer, a browser, and connections to 4,000+ apps. ChatGPT Pro and Business Premium subscribers get them first, not Plus.

OpenAI's developer conference, DevDay 2026, hit Fort Mason in San Francisco on September 29 with more than twenty announcements — but one headline matters most for regular people: ‘dots,’ always-on AI agents. Each dot gets its own cloud computer with a browser, runs on GPT-6 Astra, connects to more than 4,000 apps, keeps working across ChatGPT, Slack, and Teams toward goals you set, learns your preferences over time, and stays in read-only mode when you are not watching. First dibs go to ChatGPT Pro and Business Premium subscribers — not Plus.

The same event launched GPT-6.1 Sol, a model with near-Astra ability at one-fifth the token price ($2 for input and $10 for output per million tokens, with cached input at just $0.10); a $500-a-month ‘Pro 500’ plan with 25 times the usage of Plus; shared ‘ChatGPT Space’ team hubs; Codex coding fully in the cloud; and an Agents API for developers. Stay realistic about permissions: Ars Technica reports OpenAI paused another model release it judged too insecure, a reminder that any always-on agent poking through your apps deserves healthy skepticism. The practical take: if you already pay for Pro, dots are worth trying for research briefings and document prep; for everyone else, waiting for the wider rollout is the sensible move.