OpenAI Debuts Dots, Background AI Agents That Run Without a Screen

OpenAI has introduced Dots, a new class of software agent built to keep working toward goals a user assigns, running quietly in the background rather than waiting for prompts. The company describes the avatars as hardware-agnostic, meaning they are not tied to a phone, a desktop app, or any single interface. That framing matters because most assistants today live inside one product and stop the moment a user closes a tab.

The core idea is persistence. A Dot is meant to accept a goal, then keep pursuing it over time with minimal oversight, checking in only when it needs a decision or hits a wall. OpenAI has not published detailed technical documentation on how the agents schedule work, how they handle conflicting instructions, or where their computation actually runs. Those gaps will shape how developers judge whether Dots are a real platform or a demo.

Hardware independence is the sharper break from the current crop of assistants. Apple's Siri, Google's Gemini, and Amazon's Alexa are each anchored to devices and ecosystems that their owners control. An agent that follows a user across a laptop, a phone, or a wearable would sit above those silos, which is precisely the position OpenAI has been maneuvering toward since it began shipping consumer hardware partnerships and its own device ambitions.

The launch also lands in a crowded field of autonomous agents. Anthropic, Google, and a string of startups have all pushed tools that take multi-step actions on a user's behalf, from booking travel to filing expense reports. Most of those run inside a browser or a chat window and require the user to stay nearby. Dots' pitch — continuous background work with the human mostly out of the loop — raises the same unresolved questions the whole sector faces about reliability, cost, and what happens when an agent acts wrongly.

OpenAI has not said when Dots will be broadly available, what they will cost, or which models power them. It also has not explained how users will audit what a Dot did while they were away, a detail that could decide adoption among businesses wary of unsupervised automation. The next signals to watch are pricing, enterprise controls, and whether third-party developers get a way to build on top of the system.

For now, Dots read as OpenAI's bet that the assistant of the near future is not a voice you summon but a process that runs whether or not you are watching. The company has the distribution to test that thesis at scale. Whether users tolerate an agent acting on their behalf without a screen in front of them is the question the launch leaves open.