NEO's hand has 25 force-controlled joints and can feel a glass start to slip

We covered this hand when it washed itself under a tap. The more consequential specification is what it can feel.

The joint count, and why it is not the headline

1X’s hand for the NEO humanoid has 25 degrees of freedom: 22 actuated joints across the fingers and palm, plus three more in the wrist. Joint counts are the easy number to compare between robot hands, and on its own it says little.

The property that matters is that all 25 are natively force-controlled and backdrivable. Push a finger and it yields, then reports exactly how hard it was pushed. The tendons are quasi-direct-drive at low gear ratios, roughly 5:1 to 15:1, which keeps the hand light while leaving enough force for using tools, opening doors and pushing a cart.

Shear is the interesting sense

The tactile stack is moulded into the soft polymer of the hand itself, not bolted on as discrete sensors. It measures three things: normal force, where contact is happening, and shear.

Shear is the one that changes what the hand can do. Normal force tells a robot it is holding something. Shear tells it the object is sliding against its fingers, which is the signal that arrives while a glass is starting to slip and before it is on the floor. A grip can then be tightened in response to the slip rather than pre-clamped hard enough to survive any eventuality, which is how most robot grippers avoid dropping things and why they crush the ones that are fragile.

What it has been shown doing

The demonstration list runs from picking grapes and fitting light bulbs to zipping a jacket, assembling LEGO, pouring tea, plugging in a USB-C cable and signing in sign language. The hand is IP68 waterproof and food safe, so it can be rinsed after cooking.

NEO is on pre-order at $20,000, or $499 a month.

The part of humanoids that is still open

Locomotion has improved fast and photographs well. Manipulation is the capability that decides whether a home robot does the dishes or carries the box they arrived in, and fine manipulation in wet, cluttered, unpredictable spaces is where the field has been stuck. A hand that senses slip in real time is a direct attempt at that, rather than another walking demo.