OpenAI just launched Dots. Here’s why they matter for developers.
The New Stack Amanda Caswell ● Covered by 17 sources
OpenAI launched Dots, agents that keep working on their own cloud computers instead of waiting for the next prompt. They can watch bugs, build fixes, and hand developers a pull request while you’ve moved on.
Based on reporting by The New Stack, Amanda Caswell — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
OpenAI used DevDay on Tuesday to show off Dots, a new take on the company’s agents. They’re built on GPT-6 Astra, live on their own cloud computers, use their own browsers, and can connect to more than 4,000 apps through OpenAI’s plugin ecosystem. The pitch is simple: a Dot doesn’t stop when you stop typing. It keeps going, carries context across ChatGPT, Slack, and Microsoft Teams, and can juggle multiple projects at once.
That changes the unit of work. In OpenAI’s developer example, a Dot sits between a problem and the fix: it watches feedback for repeat complaints, scopes smaller bugs and improvements, builds and tests the changes, and returns a pull request with a video showing what changed. The company also showed Dots starting from a Slack bug report, turning a new design into a working app, and taking on more tasks without being restarted in a fresh thread.
And the use cases are not just code-shaped. OpenAI showed Dots rerunning scientific analyses when new data arrives, updating sales proposals as customer requirements change, and turning interview transcripts into clips, show notes, and social posts. The point is less “AI teammate” as a slogan and more a system that can keep a workflow moving while a human is somewhere else doing the higher-value part, which in practice usually means being interrupted by something else.
But OpenAI is also drawing a hard line between finding work and doing it. When a user isn’t actively working with a Dot, it can do “proactive research” across connected apps, but that mode is read-only. It can’t send messages, change app content, or control a browser or computer. Actions that could touch accounts or share information go through an auto-review step, and sensitive tasks like changing a password stay with the user.
There are more guardrails: monitoring can pause or stop a Dot if safety problems show up, users can see background activity in an Activity View, and the Dot’s environment stays separate from the user’s own computer unless they connect it. OpenAI says Business, Enterprise, and Education workspace content is not used to train models by default, while personal users can choose whether Dot conversations and work are used for training. The company also said Dots are rolling out now to ChatGPT Pro and Business Premium users in eligible markets, with one included at no extra cost, while enterprise customers can use the beta once an admin turns it on.
My take — AI-written commentary, not fact-checked reporting
The clever part here isn’t the agent demo; it’s the constraints. OpenAI is finally acting like an agent product can be useful only if it can’t go feral the moment it reads a Slack channel. That’s a healthier instinct than the usual Silicon Valley tradition of shipping a robot with a permission slip and calling it safety.
Read more about this at: The New Stack
