Tesla Sued Him for Secrets. Now He Ships Robot Hands.

Jay Li ran Tesla's Optimus humanoid robot program. Last year, Tesla sued him for allegedly stealing trade secrets when he left. Earlier this month, the suit was settled and dismissed. On Sunday, Li's new company, Proception, announced an $11 million seed round led by First Round Capital, with Y Combinator and BoxGroup participating — and said the first batch of its ProHand robotic hands is already shipping to robotics researchers and companies.

ProHand has 22 degrees of freedom — roughly the same joint count as a human hand — which makes it capable of a far wider range of manipulations than the gripper-style hands that dominate industrial robotics today. That level of dexterity has been the missing piece in humanoid robots: a machine that can walk but can't handle objects with any subtlety is limited to a narrow set of tasks. Bill Trenchard, the First Round partner who led the deal, described dexterous manipulation as 'a very, very, very important part of the whole humanoid story going forward.'

The data flywheel: gloves that become hands

Proception's approach to training is what separates it from other robotics hardware startups. The company equips human testers with sensor-packed gloves to capture detailed manipulation data — how fingers curl, how much force is applied, how a grasp adjusts when an object slips. That same sensor array is then built into the ProHand's skin, creating a direct link between the training data and what the robot actually feels. The result is a training loop that doesn't require a physical robot to be in the room: data collection scales with the number of human testers, not with the cost of hardware.

Dexterous robotic hands have been called unsolved for decades. Northwestern University robotics professor Kevin Lynch told the Wall Street Journal last year that fully functional robotic hands remained 'a decade' away. Li's argument is that the bottleneck was never the hardware alone — it was the shortage of high-quality, detailed manipulation data to train the robots on. The glove approach is Proception's bet that solving the data problem unlocks the hardware problem faster than building a better gripper in isolation.

Proception is entering a crowded but still wide-open field. Figure, Physical Intelligence, Apptronik, and several others are working on humanoid dexterity from different angles. What sets the Proception story apart is the specific combination of executive pedigree — Li built hands for the highest-profile humanoid project in existence — and a differentiated data strategy that treats glove sensors and robot skin as two ends of the same pipeline.