The robot that learns by watching videos.

Teaching theft was always expensive, slow and artificial, 1X NEO decided to break that logic. The humanoid Neo now learns only by watching videos, exactly how humans learn, the company introduced the 1X world model, a video-based already system that allows the robot to generalize behaviors by watching people interact with the world. They are not robotic examples, they are common Internet-scale videos, the robot analyzes the scene, visually predicts future actions and converts this prediction into real movements using an inverse dynamics model, imagines what is going to happen and then does it, in demonstrations, Neo prepares lunch boxes, opens sliding doors, lifts toilet lids, irons clothes and even combs a person's hair, often with objects it has never seen before.

It was the time he spent watching the videos that taught him how to do it. Afterwards he only needed to remember what he saw, rather than executing tasks, he learns continuously, each interaction generates new data, creating a self-reinforcing cycle, the more he acts, the more he learns, this solves one of the biggest bottlenecks of humanoid robotics, the transfer of digital intelligence to reliable physical actions. When robots learn like humans learn, the question changes, it is no longer a yes, they will evolve, it is at what speed, but learning alone is not enough, it is necessary to learn quickly.