The dots demonstrated at OpenAI's DevDay in September are starting to move out of the demonstration stage.
Product lead Thibaut Sotio said on X that a small group of ChatGPT Plus users were invited to test dots, with Business Standard users opening it up more broadly.
The biggest difference between dots and ordinary chat is that it comes with a cloud computer. You explain your goals, provide sufficient background, and connect the required applications, and it will run on its own in the background: arranging agendas, making websites, testing applications, modifying notes, and slowly figuring out your preferences based on your feedback.
Which things it can decide on its own, and which ones must wait for your nod, can be agreed on item by item.
The mobile version is also testing a new design. The bottom tab bar has a separate tab page for dots, and functions such as Codex, Projects, and Images can be fixed according to your own habits. The first batch is open to Pro users who already have dots or are qualified to create dots. Some people will be turned on by default.
If you don't want to use it, there is a "New ChatGPT Design" switch in the settings, and you can switch it back at any time.
Putting a background agent on the same level as the chat, the signal is clear: OpenAI wants ChatGPT to transform from a question and answer box into an agency.
In the same week, Gemini can operate computers by itself on Mac, and Apple is also recruiting people to be its own intelligent agents. Some routes are different. OpenAI chooses the computer in the cloud and does not touch your local things. It reduces concerns about permissions and avoids the new warnings added by macOS.
The price is that it can't do work that requires local files and local software.
There are very few test places for Plus, so it’s normal if you don’t receive an invitation.