1X's New Tendon-Driven Hand Could Be the Dexterity Breakthrough Humanoids Have Been Waiting For
There's a reason humanoid robots are often photographed from the waist up: the hands have always been the weak link. Building five-fingered appendages that can replicate the sensitivity, strength, and fine motor control of human hands is an engineering challenge that has consumed decades of research funding and humbled some of the brightest teams in the field. Norwegian robotics company 1X Technologies just made a significant move to change that.
The company has unveiled a new dexterous, tendon-driven hand for its Neo humanoid robot platform — and the design philosophy behind it is worth paying attention to.
Why Tendons Matter
Most robotic hands rely on one of two approaches: rigid link-and-joint systems actuated by motors at each knuckle, or pneumatic soft grippers that squeeze and release. Both have real-world uses, but both hit ceilings quickly when you ask them to do something a human hand takes for granted — holding an egg without cracking it, threading a cable through a conduit, or peeling a label from packaging.
Tendons are different. In human hands, tendons transmit force from muscles in the forearm through a network of fine cables running along the fingers. This remote actuation scheme is what allows our hands to be simultaneously strong and slim — the "motors" are placed far from the joints, keeping the fingers light and fast while still generating meaningful grip force.
1X's new hand applies this same principle mechanically. By routing actuation forces through tendon cables rather than embedding motors in each finger segment, the company aims to achieve a more biomechanically faithful range of motion with better force distribution across the entire hand.
The Bigger Picture for Neo
1X has been one of the more philosophically interesting entrants in the humanoid race. Backed by OpenAI, the company has argued consistently that intelligence should lead hardware — that the robot's learning system and behavioral foundations matter more at this stage than raw physical capability. Neo has been designed with that thesis in mind, trained in simulation and deployed in controlled real-world settings.
But dexterous manipulation is where that thesis gets tested hardest. Even the most sophisticated AI policy falls apart if the hand executing it can't reliably grasp, reposition, or assemble objects. The new tendon-driven design suggests 1X is now investing seriously in closing that gap — not just making Neo smarter, but giving it hands that can actually carry out the instructions.
This matters for the use cases 1X is targeting. The company has been focused on domestic and light industrial environments, where the tasks are highly varied and unstructured. That's exactly the context where a hand that can open a jar, fold a towel, and plug in a USB cable is worth far more than one optimized for a single factory operation.
Where This Fits in the Industry Moment
The timing is notable. 2026 has become the year every major humanoid maker is being held to account on manipulation. Figure AI recently highlighted its robots running 17-hour shifts sorting packages — but sorting is still a relatively coarse motor task. Companies like Physical Intelligence (π), Apptronik, and Agility Robotics have all acknowledged that sub-centimeter precision tasks remain the frontier.
A tendon-driven hand with high degrees of freedom moves the conversation forward. If 1X can demonstrate reliable fine manipulation in real-world conditions — not just in polished demo reels — it would represent a genuine step change for what humanoid platforms can be sold to do.
The robotics community has also learned, sometimes painfully, that hand design cascades through every layer of the system. A more capable hand needs new training data, new grasp planning algorithms, new force-feedback loops. It's not a hardware upgrade in isolation — it's a platform shift that requires the entire learning stack to grow with it.
What to Watch
1X hasn't published full technical specifications for the new hand yet, and the critical questions remain unanswered: What is the tendon actuation mechanism — cable-driven servos, shape memory alloys, something else? How many degrees of freedom per finger? What are the load limits and positional repeatability numbers?
Those details will determine whether this is a genuine competitive advance or a well-produced announcement ahead of broader reveal. Either way, the direction is clear: the humanoid hand problem is now the central engineering competition in physical AI, and 1X has put its tendon-driven answer on the table.
For anyone evaluating humanoid platforms — as investors, prospective enterprise buyers, or simply as observers of where robotics is heading — the quality of the hand should be near the top of the scorecard.
---
Source: AI Insider / Google News, July 9, 2026. If you're researching humanoid hardware for enterprise deployment, our robotics buyer's guide covers the leading platforms with specifications updated quarterly.