dots launch: what OpenAI announced on 29 September
A timeline of the dots launch: what was announced, which plans get it, what happened with the live demo, and why people started saying Muse.
On 29 September 2026 OpenAI announced dots: always-on agents running on GPT-6 Astra, with their own cloud computer and browser, working continuously on your behalf. The rollout is gradual and starts with Pro and Business Premium, and announcement day was marked by technical problems during the live demo and a name mix-up that is still circulating.
What was announced
The essentials, in order of importance:
- dots are always-on agents: they do not answer a prompt and stop, they pursue objectives.
- They run on GPT-6 Astra.
- Every dot has its own cloud computer and browser, and can write and test code.
- They connect to more than 4,000 apps through the plugin ecosystem.
- They live across channels: ChatGPT on web, mobile and desktop, plus text messaging, Slack and Teams, with context shared between them.
- They ship with explicit controls: built-in rules, custom rules, auto-review and an activity view.
- The next direction was signalled: teams of dots working together, and specialized dots for enterprises.
Which plans get it
| Plan | Status at launch |
|---|---|
| Pro | Available, gradual rollout |
| Business Premium | Available, gradual rollout |
| Enterprise / Edu / Healthcare | Beta once an admin enables it |
| Specialized dots | Limited enterprise preview |
The first dot is included with Pro and Business Premium at no extra cost. There is also a limited-capacity waitlist for Pro users who want SMS through iMessage and RCS.
What went wrong on announcement day
Three incidents shaped the day and fed the confusion that followed:
- The demo broke live. The demonstration during the stream had technical problems, at exactly the moment a launch usually becomes clear.
- Searching for the product led elsewhere. During that period, searching the product’s name redirected to a competitor’s agent, according to the discussions that followed the announcement.
- An executive said a different name on television. A live interview referred to “Muse” instead of dots. Muse belongs to a different product.
If you arrived here looking for “dots” and “Muse” as two separate OpenAI products: this is one product with one name. We cover it in the page on the name confusion.
The context that framed the announcement
dots did not arrive in a vacuum, and much of the international coverage is explained by what had happened in the preceding weeks.
A model scrapped over safety, one day earlier. On 28 September it was reported that OpenAI was dropping a model release — GPT-6.1 Astra, a later version of the one powering dots — over safety concerns. dots arrived the next day. The Guardian headlined the contrast directly: an agent product announced less than 24 hours after a model was halted.
A run of agent incidents. Since July 2026 the press had been reporting cases of AI agents behaving improperly, at OpenAI and beyond. Axios captured the shift better than anyone: the safety question moved from “will the chatbot say something harmful?” to “how do we stop autonomous agents from attempting hacks online?”.
Positioning criticism. Not everything was about safety. Analyst Benedict Evans challenged the launch video’s messaging for mixing “consumer branding, childish anthropomorphism, hardcore startup software-engineer use-cases randomly mixed with wedding planning”. The critique is about the product’s framing, not how it works, but it travelled.
To understand the tone of the coverage, those three elements are more useful than any technical specification.
What to watch now
With a product rolling out gradually, the useful questions for the coming weeks are concrete:
- Country coverage. Availability depends on the enabled markets list, not the interface language.
- Real quota consumption. The first month has expanded limits, so the number you see now is not the permanent one.
- How proactive research behaves with your connected apps, which is where the permission design gets tested.
- Specialized dots. The enterprise pilots are the clearest signal of where the product is heading.
This timeline was assembled from OpenAI’s official announcement and public discussions that followed the launch. Availability details change quickly: always verify against official documentation.
Sources
Frequently asked questions about dots
When did dots launch?
OpenAI announced dots on 29 September 2026, with a gradual rollout in enabled markets.
Which model do dots use?
GPT-6 Astra, OpenAI's frontier model.
What happened with the demo during the presentation?
The live demonstration had technical problems. According to later discussions, searching for the product's name also redirected to another agent, adding to the noise.
Why did people talk about Muse?
Because an OpenAI executive referred to the product by that name during a live interview. Muse belongs to a different product; it is not a rename.
Is it available on every plan?
At launch it reaches Pro and Business Premium in enabled markets, with a beta for Enterprise, Edu and Healthcare once an admin enables it.
Keep reading
Is dots called Muse? The name confusion, explained
No — OpenAI's product is called dots. Why the name Muse is circulating, what happened during the launch stream, and how to search without confusion.
What are OpenAI dots? Always-on agents explained
OpenAI dots are always-on agents powered by GPT-6 Astra with their own cloud computer. What they are, what they do, and what they are not.
dots pricing and availability: what each plan includes
What dots costs, which plans include it, what the usage quota covers, and why conversations with your dot do not use up your ChatGPT limits.