1X has released a major update for its NEO humanoid robot , not promising much but introducing a significant change to its learning logic. The new module, called 1X World Model , is designed to teach the robot to translate short voice or text commands into real-world actions, even if it has never faced such tasks before.
The upgrade is based on a video model that doesn’t rely on abstract rules, but on an understanding of real-world physics. Until now, humanoid robots have relied primarily on pre-programmed scenarios or large datasets collected through manual control by operators. 1X has taken a different approach: NEO now learns from online videos of people interacting with objects in everyday contexts and translates this understanding directly into physical movements.
The developers believe this approach helps bridge the gap between digital intelligence and the robot’s actual actions . By observing a person opening a door, picking up an object, or performing a household task, NEO learns not a specific movement, but the general logic of the action. This allows it to adapt to new objects and environments without requiring separate training for each scenario.
For the user, interacting with the robot is quite simple. All they need to do is give a short verbal or text command. NEO then uses its cameras to assess the situation, build a visual representation of what should happen , and translate this plan into precise movements using an inverse dynamics model. The company emphasizes that the robot no longer requires engineers to manually add skills, but can now figure out how to act autonomously.
Bernt Börnich, CEO and founder of 1X, says the World Model has been developed over the years, alongside improvements to the robot itself, including its shape and movement mechanics. He explains that the combination of a humanoid body and video-based learning has allowed NEO to apply digital learning to the real world and perform new actions without pre-recorded examples.
In a demonstration video, the company shows NEO performing everyday tasks, such as packing a picnic lunch. Even when the objects are different from those the robot has seen before, it calmly executes the actions and completes the task. Furthermore, in the video, NEO performs completely new tasks: opening a sliding door, raising and lowering the toilet lid, ironing clothes, and brushing a person’s hair.
The update also changes the way NEO evolves over time. Classic humanoid systems rely heavily on data collected by human operators, which requires significant time and resources. As the video models improve overall, NEO’s capabilities also grow, relying directly on this type of AI.
The developers paid particular attention to resilience to real-world conditions. A home environment is rarely predictable: lighting changes, new objects appear, and people are constantly moving. 1X claims that the World Model helps the robot maintain control and continue its task, even when the environment is constantly changing.
NEO is now available to order through the company’s online store. Early access is priced at $20,000 , with priority delivery scheduled for 2026. A subscription plan is also available for $499 per month .
1X acknowledges that the technology is still in its infancy, but believes this represents a step towards robots that learn not from rigidly defined rules, but rather from the same principles as humans.
Follow us on Google News to receive daily updates on cybersecurity. Contact us if you would like to report news, insights or content for publication.
