1X's Neo Robot Ditches Human Trainers for AI 'World Model'

The era of humans training humanoid robots may be coming to an end. 1X, a startup backed by OpenAI and a direct competitor to Tesla, has announced a fundamental shift in how its robots learn. The company is moving away from relying on armies of human operators and toward a new artificial intelligence model that allows the robot to teach itself.

This technological leap centers on the company's new world model, an AI system designed to interpret video footage captured by the robot's own cameras. By analyzing this visual data, the Neo humanoid can learn to navigate the physical world and perform household tasks without the need for a human pilot in a virtual reality headset. The change promises to scale the robot's intelligence faster and more efficiently than ever before.

For years, training humanoid robots has been a labor-intensive process. Companies like 1X and Tesla have employed operators to perform simple, repetitive tasks—such as squatting, washing dishes, or folding laundry—while wearing motion-capture suits and VR headsets. This data collection method, known as teleoperation, has been the industry standard for teaching robots how to interact with the physical world.

However, 1X CEO Bernt Børnich told Business Insider that this model is no longer sustainable. The startup's new world model fundamentally changes the equation by allowing Neo to learn from its own experiences.

"Essentially, the world model does the same thing as the operator would do,"

said Børnich. He explained that this advancement allows intelligence to scale with the number of deployed robots rather than the number of human operators gathering data. A spokesperson for the company confirmed that the new model "significantly reduces" their reliance on teleoperation, with future data collection largely being performed by the robots themselves.

The Neo humanoid robot first captured public attention in October with a viral video demonstration. The footage showed the robot performing a variety of domestic chores, including vacuuming, folding laundry, and unloading a dishwasher. Designed to tackle boring and mundane tasks around the house, Neo is now preparing for its commercial debut.

Customers can purchase the robot for a one-time payment of $20,000 or subscribe for $500 per month. 1X expects to ship units this year and has already produced over 10,000 robots. Following its viral demo, the company reported that it "sold out in the first few days."

Early adopters should be aware of a temporary catch. To perform tasks the robot cannot yet handle autonomously and to gather further training data, Neo will occasionally be piloted by a remote human operator. Børnich emphasized that this teleoperation will be transparent to the user, comparing the experience to "inviting a human inside your home to help out." The company is taking steps to protect user privacy during these sessions.

While teleoperation is part of the initial rollout, 1X has a clear roadmap toward complete independence. The company's advances in its world model have accelerated the timeline for autonomy. According to Børnich, Neo should be able to perform most tasks without human assistance by the end of this year.

The ultimate goal is a fully autonomous product. Børnich stated his expectation that by 2026, 1X will be able to deliver a robot that operates "out of the box" without any human intervention except for the user's own commands.

A company spokesperson noted a significant shift in the robot's capabilities. Previously, customers had a limited set of "hero" tasks that Neo could complete autonomously. Now, the robot will attempt all assigned tasks without support. The spokesperson added a realistic caveat:

"Task execution may not always be perfect, and it may struggle at hard tasks, but it will learn over time."

1X is not alone in its pursuit of more efficient training methods. The entire robotics industry is grappling with the challenge of acquiring high-quality data to train humanoids. World models—AI capable of simulating realistic environments and real-world physics—are becoming a key focus for labs seeking to advance their technology.

Even industry giant Tesla is pivoting its strategy. Reports indicate that Tesla has revamped its Optimus training program to rely more heavily on video, moving away from data collected from humans in motion-capture suits. This industry-wide trend highlights the growing importance of AI-driven learning over manual human input.

The shift also has implications for the human workforce. Robot training has become a source of employment, with positions advertised at $25 per hour for physically demanding and often tedious work. Some operators on Tesla's Optimus project have reported sustaining injuries. As AI models like 1X's world model mature, the demand for these human training roles may diminish, marking a significant transformation in the robotics labor market.

The introduction of 1X's world model represents a pivotal moment in the evolution of humanoid robots. By enabling Neo to learn from its own visual perception, the company is moving toward a future where robots can adapt to new environments and tasks with unprecedented speed and efficiency. This leap from human-guided training to self-supervised learning is the key to unlocking the true potential of domestic automation.

As 1X ramps up production and prepares for its first wave of customer deployments, the focus will be on how effectively the robot learns and adapts in real-world homes. The journey from teleoperated helper to fully autonomous assistant is well underway, setting the stage for a new era where intelligent robots become a common presence in our daily lives.