OpenAI Debuts Dots, an Always-On Agentic Assistant Powered by GPT-6 Astra
OpenAI used its DevDay keynote to unveil Dots, a persistent AI helper that works across connected apps in the background, answering Meta's Muse.
OpenAI announced Dots on Tuesday at its DevDay keynote, a new class of always-on assistant that runs across connected applications in the background rather than waiting for a prompt. The move is a direct answer to Meta's Muse, the buzzy agentic helper that has dominated AI chatter in recent weeks. Dots, OpenAI says, can handle nearly any task its users hand it, learning their preferences as it goes.
The timing matters. Muse gave Meta an early lead in the race to ship agents that live inside a user's daily software rather than sitting in a chat window, and OpenAI has spent the past year under pressure to show its own agent strategy is more than a demo. By putting Dots on the DevDay stage, the company is signaling that persistent, multi-app automation is now a core product line, not a research curiosity.
The technical centerpiece is GPT-6 Astra, the model OpenAI says powers Dots. The company did not disclose benchmark numbers or pricing at the keynote, but the framing suggests Astra was built for long-running, tool-using tasks rather than single-turn conversation. That distinction is the whole pitch: an assistant that remembers context, holds state across sessions, and acts on your behalf without being re-briefed each time.
Dots also raises the stakes for developers in OpenAI's ecosystem. If the assistant can reach into connected apps on its own, third-party services face a choice between integrating and being bypassed. OpenAI has not yet detailed which apps will be supported at launch, how permissions will be scoped, or what happens when an agent takes a wrong action on a user's account.
Those unanswered questions are where the real fight will play out. Meta's Muse already has distribution through the company's family of apps, and Google is expected to push its own agentic offerings deeper into Android and Workspace. OpenAI's advantage is developer mindshare and the Astra model; its risk is that always-on agents invite scrutiny over privacy, reliability, and liability once they start doing things rather than just saying them.
Watch for OpenAI to publish integration details, safety guardrails, and a rollout timeline in the coming weeks. Until those specifics land, Dots is a promise about ambient automation — and the industry's next argument will be over who can make that promise safe enough to actually ship.