1X founds the World Model Lab to accelerate the autonomy of its humanoids.

1X launches the World Model Lab, led by Sam Sinha, to pretrain video foundation models and accelerate the autonomy of its NEO humanoid robots.

« Robotics is not a fine-tuning problem. »

Based on this thesis, 1X is establishing the 1X World Model Lab, a frontier research organization dedicated to large-scale pretraining of embodied world models, with fully autonomous humanoids as its ultimate goal. The company behind NEO believes that embodied intelligence requires models exposed to the physical world from the outset, not in post-training: space, movement, objects, causality, and physics must be learned natively. The initiative follows advances in the in-house World Model that have enabled NEO to execute zero-shot tasks it had never encountered.

The highlighted scaling recipe combines web-scale media, human egocentric videos, simulation, teleoperated robot data, and on-policy data from NEOs, with real-world deployment then fueling data collection and RL. With the NEO factory operational, 1X considers that the bottleneck is no longer data collection but the pace of innovation on the world model itself. Bernt Børnich, its CEO, advocates the idea that achieving full autonomy as quickly as possible requires owning the entire stack, down to pretraining its own video foundation models.

The lab will be led by Sam Sinha, a former founding researcher at Luma AI and a specialist in scaling multimodal generative video models, named Founding AI Researcher and Head of World Models. Recruitment is open for the three pillars that 1X identifies for AI: data, infrastructure, and learning.