Huh? Is this a robot's hand?

Wait, is this even a robotic hand? Where on earth did this come from?

1X Technologies has just dropped a game-changing breakthrough.

A 25-degree-of-freedom, tendon-driven robotic hand mounted on their in-house humanoid robot NEO can effortlessly perform tasks like screwing in light bulbs, picking up tiny screws, zipping zippers, pouring tea, plugging in USB-C cables, and even signing.

It features force transparency, backdrivability, full fingertip tactile sensing, a positioning accuracy of ±0.2mm, a wrist joint that has passed 2 million cycle tests, and IP68 water resistance that lets it wash itself.

Most critically: Production capacity is set at 10,000 units this year, and the assembly lines are already up and running.

1X Founder and CEO Bernt Børnich stated: With these hands, robots can now perform the manual tasks that humans do every single day.

Joints are sensors in their own right

The vast majority of robotic hands are "write-only" devices.

Send a position command, and the fingers move to the target spot, but the system has no idea what the fingers touched, how hard the object is, or if it's slipping.

The root cause lies in the transmission ratio:

The industry-standard 100:1, even 200:1 gear reduction ratio swallows contact forces in friction, so the signals vanish before they can ever travel back to the motors.

The result is a numb hand, forcing engineers to attach external cameras and rely on vision to guess what's happening at the fingertips.

1X took a completely different approach.

The NEO hand was developed entirely from scratch, using quasi-direct-drive tendon transmission, with transmission ratios reduced to a range between 5:1 and 15:1.

Of its 25 degrees of freedom, 22 are fully actuated in the fingers and palm, and 3 in the wrist — all with native force control and full backdrivability.

Push a finger, and it will yield compliantly while accurately reporting exactly how much force you applied. Force flows out from the fingertips, and sensory information travels back along the exact same physical path.

1X calls this property "force transparency" — a feature that turns every single push, press, and pinch into a precise measurement.

On top of force feedback comes full proprioception: every joint operates in closed-loop control, so the hand always knows its exact configuration without needing visual input, just like how humans can touch their index fingers together with their eyes closed.

What it can do: from picking coins to signing

What do 25 force-transparent degrees of freedom make possible?

They enable a rich variety of grasping postures and manipulation motions. The DOF distribution was modeled after human hand anatomy, with special emphasis on enhancing the thumb's opposition capability — the core of human dexterous fine manipulation.

According to 1X, this configuration strikes the optimal balance between manipulation capability, manufacturability, controllability, and serviceability.

NEO can assemble Lego bricks, pick out individual coins and screws from a wallet, rotate and install light bulbs, use a screwdriver, reorient objects within its palm, zip up a jacket, sort grapes by color, pour tea from a kettle, catch a soft ball, insert a USB-C charging cable, pick up a wine glass, wipe down surfaces with spray and paper towels, and even perform sign language.

Its strength is equally impressive. The thumb's carpometacarpal joint delivers a peak torque of 3.5 Nm, the finger metacarpophalangeal joints hit 2.6 Nm, fingertip flexion force reaches 45N, and wrist torque peaks at 17.75 Nm. Opening doors, pushing loaded carts, lifting heavy objects, and operating tools are all well within its capabilities.

The ±0.2mm positioning accuracy perfectly covers the most common small-object manipulation scenarios in human daily work.

At the tactile level, high-resolution tactile sensors cover the fingertips and contact surfaces, detecting normal force, contact location, and shear force.

When an object starts to slip, the hand detects this in real time and re-grasps it securely.

1X's visualization demos show surface contact normal vectors, pressure heatmaps during handshakes, and the process of gently picking up fragile origami without causing any damage.

This synthetic skin is not a standalone functional material — it is co-designed with internal sensors and the underlying tendon system, because vision alone is simply insufficient for small, transparent, deformable, or occluded objects.

Production lines are already operational

The hand's motors are housed in the forearm, matching the location of the human hand's grip strength muscles, and pull the fingers via proprietary tendons that run through the wrist.

This keeps the hand itself lightweight, while allowing sustained high-force output during continuous operations without overheating.

The entire hand is deeply vertically integrated: in-house developed motors, custom electronics, embedded sensors, proprietary tendon systems, compact transmission mechanisms, and hand-specific firmware are all designed and manufactured in-house. From tendon materials to the outermost soft polymer and tactile sensing layers, every step is completed end-to-end internally.

For durability, individual components and full finger assemblies have undergone millions of cycle tests. Actuation units have been validated under extreme temperature conditions, and the wrist joint has passed over 2 million high-load cycles. It carries an IP67 ingress protection rating, uses food-safe materials, can operate near sinks, and rinses itself clean when dirty.

Safety is guaranteed mechanically: the extremely low transmission ratio, paired with tendon actuation and low distal inertia, ensures that external impacts can safely backdrive the fingers.

In 1X's slow-motion demos, the hand is slapped, struck with a hammer, pinched in a drawer, and slammed into foam — all the while compliantly yielding without suffering any damage.

Hundreds of these hands have already been mass-produced on scalable assembly lines, with an annual production target of 10,000 units, and every single unit undergoes full end-of-line testing.

A robotic hand that cannot be mass-produced cannot generate large-scale datasets, and large-scale datasets are the foundation for advancing embodied intelligence.

Reference link: [1]https://www.1x.tech/discover/neos-hands

This article is sourced from the WeChat Official Account "QbitAI", written by Mengchen, and published with authorization from 36Kr.