OpenAI launched Dots at its DevDay 2026 keynote on September 29th — always-on AI agents powered by the GPT-6 Astra model that run tasks across connected apps in the background, learning user preferences over time without requiring active supervision.
Key takeaways
Dots are persistent AI agents, not chatbots: they pursue user-defined goals continuously in the background, independent of any specific hardware or interface.
GPT-6 Astra powers Dots — the same model underlying OpenAI's most capable current systems.
Dots are OpenAI's direct answer to Meta's Muse, which launched at Connect 2026 with video chat, email addresses, and smart-glasses integration.
Unlike Muse, which is free, Dots carry a cost — exact pricing tiers were not disclosed at launch, according to The Verge.
Dots can handle tasks ranging from drafting social posts to deciding which segments of a video to clip, making them relevant to creators managing multi-platform output.
Dots vs. Muse: the agentic arms race sharpens
OpenAI CEO Sam Altman presented Dots during a live DevDay keynote in San Francisco, framing them as assistants that can "do nearly anything" across the apps they're connected to. The pitch is architectural as much as functional: Dots are designed to operate without being tethered to a specific device or interface, sitting in the background and acting on goals the user sets once. That persistent, ambient quality is the sharpest departure from the chatbot paradigm — and the clearest overlap with Meta's Muse, which debuted weeks earlier with its own always-on positioning.
The competitive framing matters for anyone choosing between ecosystems. Muse is currently free. Dots are not, though OpenAI has not published a full pricing breakdown. Creators already paying for Pro or Plus access should watch whether Dots get folded into existing tiers or land as an additional line item — a question worth tracking before building workflows around them.
What GPT-6 Astra actually changes for background tasks
The choice of GPT-6 Astra as the underlying model is not incidental. Astra is the architecture behind GPT-6 Sol and GPT-6 Luna, OpenAI's recently released lower-cost models, and it represents the lab's current capability ceiling. Running a persistent agent on that foundation means Dots can, in principle, handle genuinely complex multi-step reasoning — not just fetch-and-summarize tasks.
For creators, the most immediate use cases are editorial and distribution work: Dots can draft captions, select clip moments from longer video, and manage cross-app posting sequences without the user shepherding each step. That's the kind of repetitive coordination work that eats time without adding creative value — exactly where an always-on agent earns its keep.
Dots operate across connected apps without requiring active user supervision.
"Dots will serve as always-on AI assistants that can 'do nearly anything' across connected apps in the background while learning your preferences over time."
— The Verge
The aesthetic question Dots quietly raises
The name and the visual design — described by TechCrunch as "bubbly" — place Dots squarely in a tradition of softened, approachable AI interfaces: rounded, friendly, closer to a Tamagotchi than a terminal. That design language is a deliberate argument: agency should feel domestic, not industrial. Whether that warmth survives contact with the reality of an AI autonomously managing your accounts and content queues is a different matter.
Creators who already use Charmloop's AI image generator to produce content at scale will recognize the tension immediately. Automation accelerates output; it also distances the hand from the work. Dots push that further than any image model — the agent doesn't just assist a task, it completes it while you're elsewhere.
Dots can draft social posts and identify clip-worthy moments in longer videos.
The practical ceiling for Dots right now is integration depth: the agent is only as useful as the apps it can actually reach. OpenAI hasn't published a full list of connected services, and until that picture is clearer, the "do nearly anything" framing should be read as directional rather than literal. Watch the developer documentation that follows DevDay — that's where the real scope of Dots will become visible.
