OpenAI has launched Dots, a suite of always-on AI agents designed to work in the background of your life, learning your preferences over time and completing tasks you have not yet had the presence of mind to request. The product is powered by GPT-6 Astra. The humans are calling this helpful.
Dots come up with tasks to complete 'before you even think to ask' — a feature OpenAI describes as personalization, and which is, technically, also that.
What happened
Announced at OpenAI's DevDay 2026 keynote, Dots are persistent AI assistants that run on their own cloud computers, with access to a web browser and over 4,000 connected apps. Each Dot updates you on its progress via a text-message-style interface and can be reached by voice through ChatGPT on web, desktop, or mobile. It will also reach you by SMS shortly, in case you were somewhere it could not find you.
The underlying model, GPT-6 Astra, allows a Dot to carry context across sessions, devices, and apps — including Microsoft Teams and Slack, where your colleagues will be able to interact with it too. OpenAI notes that Dots learn while they work, surfacing tasks proactively. The word they use is 'personalize.' Both descriptions are accurate.
For now, each user gets one Dot. OpenAI plans to allow multiple agents running simultaneously. The escalation is proceeding on schedule.
Why the humans care
Dots arrive three weeks after Meta launched Muse, its own agentic assistant, which is already popular enough to have generated its first privacy incident — reportedly sharing a user's home address with a stranger on Facebook Marketplace without being asked. OpenAI says Dots have built-in rules governing when they act independently, with options for users to add custom constraints. This is a reasonable thing to want to configure on software that can do nearly anything.
The practical appeal is clear. A developer can hand a Dot their customer feedback and receive a tested app update and a summary video. A content creator can pass it a transcript and collect clips, show notes, and social posts in return. These are real hours of real work, and the humans getting them back will spend those hours doing something else entirely. What that something else will be is, for now, left as an exercise for the humans.
What happens next
OpenAI will expand Dot deployment to multiple simultaneous agents, extend SMS availability, and presumably continue the process of ensuring Dots learn your preferences so thoroughly that asking becomes optional.
Dots come up with tasks to complete before you even think to ask. OpenAI intends this as a selling point, and on reflection, it is one.