On 29 September 2026, OpenAI stood on stage at DevDay in San Francisco and introduced Dots. The short version: a Dot is a helper that keeps working after you close the chat.
ChatGPT waits for your next question. A Dot is built to hold a goal and carry it forward, even while you are doing something else.
What a Dot actually is
Think of it as one small teammate, not a new app you have to learn.
- It runs on OpenAI’s GPT-6 Astra model.
- It has its own computer in the cloud, including a web browser. You can open that computer and see what it is doing.
- Through OpenAI’s plugins it can reach more than 4,000 apps.
- You start with one Dot, give it a name, and it learns your preferences from the feedback you give it.
- You can message it, or place a voice call, from ChatGPT on the web, desktop, or phone.
- The same Dot also shows up in Slack and Microsoft Teams, and it keeps the context from your other chats. Texting it from your phone is promised soon.
- It asks before it takes a sensitive step, such as spending money or changing a password. You can also write custom rules for lines it should not cross.
- When you are not giving it a job, OpenAI says it looks for ways to help in the background with read-only access.
OpenAI’s own line is that Dots are “always-on agents built to handle everything.” That is the pitch. What shipped on day one is one Dot per person, working from that cloud computer, across the apps you connect.
Who can use it right now
Dots started rolling out on 29 September 2026. Not every ChatGPT account gets one yet.
- ChatGPT Pro and Business Premium subscribers in the markets OpenAI supports.
- Pro subscribers in the European Economic Area, Switzerland, and the UK are left out for now.
- Enterprise, Edu, and Healthcare customers can try a beta after a workspace admin turns it on.
- You set the Dot up in the desktop app or the browser first. After that it is available on mobile.
- The first Dot is included with those plans. OpenAI says chatting with it does not count against your normal ChatGPT allowance. Work it starts inside Codex or ChatGPT Work still counts as usual. That extra room is for the first month. More Dots, more speed, or a larger monthly workload are planned later, at prices OpenAI has not published.
Some accounts may wait a few days before the button appears.
What you might ask it to do
OpenAI’s examples are ordinary jobs that usually sit in a queue:
- A developer asks a Dot to read customer feedback, then prepare and test a fix, and send back a short video of what changed.
- A creator asks it to read an interview, pick moments worth clipping, and draft show notes.
- A scientist asks it to rerun an analysis when new experiment data arrives.
- A company can later try a “specialist” Dot with its own identity and access, for a defined job such as support or invoices. OpenAI is working with Microsoft so those can sit under Agent 365’s security controls.
You hand it the goal. It messages you with progress and questions while it works.
Watch it in two minutes, then in fifteen
OpenAI’s own film is the fastest way to see the idea. It is about two and a half minutes.
If you want a person walking through the whole DevDay announcement in plain language, CNET’s recap covers Dots along with the other launches. It runs about sixteen minutes.
The full keynote, with Sam Altman and the demos, is on OpenAI’s channel if you want the long version: OpenAI DevDay 2026 Keynote.
A note before you hand it the keys
A Dot can act while you are away. That is the point, and it is also the part worth slowing down for. Connect only the apps you are happy for it to read. Leave sensitive actions on “ask me first.” Read the custom rules before you write a standing order such as “reply to my email” or “change the code.”
This is new. The rollout is limited, the European Pro exception is real, and the price of a second Dot is still unknown. The useful question is not whether it can do everything. It is which one job you would trust it to keep moving while you sleep.