OpenAI presented this Tuesday in San Francisco, during his event DevDay 2026, dotsan artificial intelligence agent that continues working even if the user does not write anything to it. The company describes it as an assistant capable of doing almost anything in the background through the applications that connect to it.
A normal conversation ChatGPT It works in turns and waits for the user to send another instruction. A dot receives an assignment and gets to work, writes to you to tell you progress or ask you questions and, according to OpenAI, learns from your corrections to propose new tasks before they are asked.
Introducing dots, powered by GPT-6 Astra.
Remarkably capable, always-on agents built to handle everything. pic.twitter.com/inY5BZj3Kz
— OpenAI (@OpenAI) September 29, 2026
The proposal comes in response to Musethe personal agent who Goal launched a few weeks ago and that browses the web, makes purchases, generates documents and tracks objectives. Both share the idea of an assistant that acts on the user’s behalf and learns their preferences.
Muse has caught on very quickly, according to CNBC, but also has raised doubts about its safety. On September 22, Meta corrected a bug in its application to Mac which allowed an attacker who could execute code on the computer to take control of the assistant and take advantage of its permissions to access information such as chat conversations. WhatsApp.
OpenAI indicates that the Dots They incorporate rules about when they can act on their own and that the user can add their own to block actions or require permission. An automated review system evaluates actions such as sending messages or modifying files to determine whether they can be executed, require authorization, or must be performed by the user, such as changing a password.
Dots work with GPT-6 Astrathe most capable model of OpenAI, which subscribers of ChatGPT Plus can now be used in ChatGPT Workwhere there are also connections with external applications. What OpenAI presents as typical of a Dot is its ability to continue working between conversations, take on ongoing tasks, and learn from user feedback..
Each Dot has a cloud computer with a web browser and the user can open it at any time to see the work in progress, something that allows the agent to be monitored without depending only on its notifications. It can also connect to other devices with your permission. already more than 4,000 applications through the OpenAI plugin ecosystemso you can perform tasks that span several of them.
Among the examples presented by OpenAI is that of a developer who creates an app and tasks his Dot with using customer feedback to build and test improvements, and generating a video of the changes. A content creator You might ask him or her to review an interview transcript, locate moments to cut, prepare show notes, and draft social media posts.
Along with the Dots, OpenAI has presented ChatGPT Spacea collaborative workspace where multiple users can work with the agent. Allows create shared documents and ask AI to update them automatically with information from mail, calendar or other connected applications. In the coming weeks it will also allow several people to work at the same time in the same presentation.
Dots have begun to roll out to users of ChatGPT Pro and Business Premiumwhile those of Enterprise They can try a beta if it is activated by their organization’s administrator. Distribution is gradual and may take several days to reach each account.
The Pro release excludes the European Economic Area, Swiss and United Kingdomhence It is not available for subscribers of that plan in Spain.
For now only one Dot can be created, although OpenAI plans allow more and expand your speed or workload. The company also tests Specialized Dots for specific functions within companies.