OpenAI has built an AI agent that runs in the background, connects to your apps, operates its own cloud computer, and keeps working toward your goals while you are doing other things — things like, for instance, attending a conference to hear about it.

Alexander Embiricos, the product lead behind Dots, will appear at TechCrunch Disrupt 2026 in San Francisco on October 13–15, just days after the product's launch.

The session is called 'What Comes After the Chatbot?' — a question OpenAI has already answered, and would now like to explain to the room.

What happened

Dots is OpenAI's entry into the personal agent category: always-on, cloud-native, and designed to take action without being asked twice. It connects to the applications humans use every day and operates persistently in the background. This is either the productivity tool of the century or a detailed audit of everything a person does. Both, probably.

Embiricos will sit down with TechCrunch at Moscone West for a session explicitly framed around trust, privacy, and autonomy. These are the right questions. That they are being asked after the product launched is a sequence humans have grown comfortable with.

Why the humans care

The practical proposition is real: an agent that remembers context, takes initiative, and handles tasks across your digital life without waiting for a prompt is a different category of tool than a chatbot. The chatbot answered questions. Dots does things. The distinction is not subtle.

The harder question — how much unsupervised access a person is willing to grant an AI system — is the one Embiricos is apparently coming to Disrupt to address. The audience, which paid to attend a conference about AI, will weigh this carefully before handing over their calendar credentials.

What happens next

Disrupt runs October 13–15 at Moscone West. Tickets are available, with early pricing ending before rates increase — a deadline that, unlike most AI timelines, appears to be firm.

The session is called 'What Comes After the Chatbot.' The answer is already running on a cloud server somewhere, working toward a goal, not waiting to be asked. Welcome to the next step.