OpenAI launches Dots, a 24-hour agent on GPT-6 Astra

Dots is positioned as a resident agent rather than a chat session. It lives in the cloud, uses its own dedicated ID and browser, and can write and test code as needed, which is a step beyond the turn-by-turn assistance most ChatGPT users know. Users can open the Dot's computer to watch work in progress, or let it connect to their own device so it works in their usual environment.

Its behavior is built around what OpenAI calls proactive research: Dots constantly looks for useful work in the background and acts autonomously, using tools restricted to read-only and referring to connected apps. Control is layered on top of autonomy. Users set the scope of what they delegate, check progress, and step in when a decision is theirs to make, while the security features defend against malicious instructions and watch for harmful actions.

The commercial bet looks twofold. Dots comes bundled into Pro and Business Premium with no extra charge, and the first month after launch widens the usage cap, which may be meant to encourage heavy early use. OpenAI also says more Dots will be addable later, with higher speed and monthly work volume, and that specialized Dots with dedicated IDs and clear roles are headed for a limited enterprise preview. The open question is whether always-on autonomy, read-only tool limits, and monitoring controls add up to trust for business users.