Thinking in Frames: How 1X Technologies Uses Visual Imagination to Advance Robotic Autonomy

The landscape of robotics is moving toward a future where machines no longer rely on rigid, pre-programmed scripts, but instead utilize an internal "imagination" to navigate the physical world. Norwegian-American startup 1X Technologies has officially moved closer to this vision with the deployment of its 1X World Model (1XWM) for the Neo humanoid robot. A World Model is essentially an AI-driven internal simulator that allows a robot to predict the physical consequences of its actions before executing them, much like a human might visualize catching a ball before reaching out. This development marks a shift in how robots are trained, moving away from a total reliance on human teleoperators—specialists who remotely control a robot to teach it specific movements—and toward a system that learns autonomously from vast amounts of visual data.

At the heart of this innovation is a text-conditioned diffusion model, a type of AI architecture similar to those used in high-end video generators that can create complex visuals from simple written descriptions. 1X Technologies has trained this model on hundreds of hours of "egocentric" human video—footage filmed from a first-person perspective—to help Neo understand how humans naturally manipulate objects. By fine-tuning this system with specific robot data tailored to Neo’s unique visual appearance and kinematics, the robot can now translate a simple voice or text prompt into a sequence of physical movements. Instead of searching for a pre-recorded instruction, the AI first "imagines" a short video of itself completing the task and then uses an Inverse Dynamics Model (IDM)—a secondary AI system that calculates the exact motor forces needed to match those imagined images—to perform the action in real life.

This approach is designed to address the "data bottleneck," a significant challenge in robotics where progress is slowed by the thousands of hours of manual training required for even the simplest tasks. Because Neo is built with a human-like shape and soft, organic-inspired movements, 1X Technologies claims the robot can more effectively apply knowledge harvested from video of people interacting with their environment. Recent demonstrations have shown Neo performing tasks such as handling kitchenware, cleaning, and general object manipulation. The company suggests this technology could eventually generalize to more experimental tasks, such as garment care or delicate household assistance, without specific prior training. This ability to "generalize"—or apply learned logic to entirely new situations—is viewed by developers as a critical step toward creating a "flywheel effect," where a robot's intelligence grows autonomously as it experiences more of the world.

The physical hardware supporting this intelligence is equally sophisticated. Neo stands approximately 167–168 cm (5'6") tall, weighs 30 kg, and is powered by the 1X Neo Cortex. This system utilizes high-performance chipsets to manage the heavy computational lifting required for real-time AI processing. Its battery allows for several hours of continuous operation, and its soft design—often seen wearing a form-fitting textile suit—is intended to make it a non-threatening presence in domestic and professional spaces. To address the privacy concerns inherent in placing a camera-equipped robot in a home, the company is implementing a suite of privacy protocols and security tools designed to ensure that the robot's integration into domestic settings respects user confidentiality.

As 1X Technologies begins its initial deliveries to the U.S. market in 2026, the focus is shifting toward large-scale industrial and domestic utility. Through a strategic partnership with the investment firm EQT, the company aims to deploy thousands of Neo units across sectors like logistics, manufacturing, and healthcare by 2030. While the current units may still use remote oversight for highly complex or unexpected scenarios, the stated objective is to steadily increase the level of autonomous capability throughout 2026. By turning the global archive of human motion into a classroom for AI, 1X Technologies is working to prove that the future of robotics lies in a machine’s ability to learn, adapt, and eventually navigate the world with a high degree of independence.