1X World Model: humanoid robot EVE plans in real time via video prediction
In one sentence 1X Technologies presents an end-to-end world model for humanoid robot EVE: it predicts future video frames from current observations and actions, trained purely on robot data. It enables real-time planning without external compute, a key step toward autonomous household robots.
To move intelligently in an environment, a robot needs to be able to "imagine" the consequences of its actions before executing them. This is the idea behind world models: a model that predicts how the world will change in response to the robot's actions.
1X Technologies developed a world model for its humanoid robot EVE that works directly with video: the model watches the current video stream and predicts what the next second will look like based on any action the robot might take. All of this runs in real time, directly on the robot's own hardware.
The important point is that this model was trained only on EVE's own data — no internet video, no simulation. The robot learned to understand its own environment simply by observing its own experiences.
An onboard world model means the robot does not depend on cloud connections or external computers to plan its movements. It can independently understand what will happen if it grabs an object, moves left, or opens a drawer. This is a fundamental step toward robots that think and act independently.
Companies
1X Technologies
Tools
—
Tags
Sources