Give AI the creative power of human hands 🤖
GenRobot just launched Gen DAS Dex, and the real story is not the glove itself. It is the data stack behind it.
Dex captures 23 DoF of hand motion, 1mm fingertip positioning, 0.02° joint precision, and 0.05N tactile sensitivity, then syncs that with wrist vision and GenRobot’s Ego system for millisecond-level head-hand alignment.
That matters because embodied AI does not just need video. It needs ground-truth data for how humans grasp, adjust force, rotate objects, and recover in the middle of real physical interaction.
What GenRobot is building looks less like a device company and more like a human-data infrastructure layer: capture hardware, egocentric sensing, structured datasets, and a pipeline to clean, label, and feed that data into models.
If internet-scale text built language models, this is the same bet for robotics: real human interaction data, collected at scale, with enough precision for robots to actually learn how to use their hands.