Insider Brief
- 1X has launched the 1X World Model Lab, a new research organization focused on developing large-scale AI world models and foundation models to improve the autonomy and adaptability of its Neo humanoid robot.
- The company said the lab will focus on large-scale pretraining using internet data, first-person human video, simulation environments, teleoperated robot data and real-world data collected from Neo robots, reflecting its belief that scaling data and model training is key to more capable humanoid systems.
- 1X also hired former Luma AI researcher Sam Sinha to lead the effort and said it is expanding hiring across data, infrastructure and machine learning as it scales both its AI research and Neo humanoid manufacturing operations.
Humanoid robotics maker 1X has launched a new lab focused on developing AI world models to “to pretrain on the most important data from the very beginning.”
According to the Norwegian-American company, 1X World Model Lab will focus on training large-scale foundation models designed to help robots understand, predict and interact with their environments. The company also announced the hiring of Sam Sinha, a former founding research scientist at Luma AI, as founding AI researcher and head of world models.
“World models will free humanoid robots from the trap of narrow tasks in static environments by unlocking generalization, adaptability, and self-learning. But to reach full autonomy fastest you must own the entire stack — down to pre-training your own video foundation models,” noted founder and CEO Bernt Børnich. “That’s why we built the 1X World Model Lab.”
World Model Lab
The establishment of the lab follows what the company said are recent advances in its internal world model technology that allowed its Neo humanoid robot to perform previously unseen tasks without task-specific training. The company said those results reinforced its belief that scaling data and model training is the most promising path toward more capable humanoid robots.
Rather than focusing solely on fine-tuning models for specific tasks, 1X said the lab will concentrate on large-scale pretraining using a combination of internet data, first-person human video, simulation environments, teleoperated robot data and information collected from Neo robots operating in real-world settings.
Who is Sam Sinha?
Sinha joins the company after helping develop multimodal image and video generation models at Luma AI. According to 1X, his experience building large-scale generative AI systems will be applied to the development of world models designed specifically for robotics and physical AI applications.
“We believe the only durable moat in AI is data — specifically, the ability to collect and learn from the optimal mixture at large scales,” Sinha added. “Embodied data has been treated as a second-class citizen for too long. Embodied AI is too important to be a fine-tuning problem. Scaling along every imaginable axis in robotics is the only way forward. 1X is uniquely positioned to close the full learning loop from data curation and pretraining, through real-world deployment.”
The company is also expanding the team behind the new research organization, recruiting talent across data, infrastructure and machine learning to support development of its next generation of robotics AI models.
1X announced in May it now has the capacity to produce up to 10,000 Neo humanoid robots annually as it expands manufacturing ahead of a broader commercial rollout planned for 2026. The company said demand sold out its first year of planned production within five days of opening preorders and that it is combining scaled manufacturing, teleoperation-assisted learning and world model technology while pursuing deployments of up to 10,000 humanoid robots through a partnership with EQT.
Image credit: 1X
