Ask a robotics company what stands between it and a useful humanoid and you will increasingly get the same answer. Not legs, not batteries, not even the model driving the thing. Hands. Elon Musk named it the "hands problem," and last week 1X, the Norwegian company behind the $20,000 home humanoid NEO, claimed to have solved it.
The specification is genuinely impressive. 1X's new hand has 25 degrees of freedom, against 27 in a human hand. It can pinch, twist, grasp and use tools, and its fingers move faster and bend further than yours. It is waterproof, so the robot can wash its own hands, and strong enough to carry grocery bags and suitcases. In the launch video the hand is struck repeatedly with a hammer, survives, then opens a bag of Funyuns and offers one to the person who hit it. Asked by Business Insider whether the problem was now solved, 1X's VP of design and product Dar Sleeper said: "The short answer is yes. This is by far the closest hand to human-level dexterity."
One detail in the coverage deserves its weight. A 1X spokesperson confirmed the robot in that video was running on a mix of autonomous operation and remote control, in order to "show the upper limit of the hardware's capabilities." That is an honest disclosure, and it also relocates the problem. What 1X has demonstrated is a mechanism capable of human-like dexterity when a human is helping to drive it. Whether a policy can control 25 degrees of freedom autonomously, in a stranger's kitchen, is the part that remains open. The hands problem has two halves, and the hardware half is the one that just moved.
Why do hands matter so much? Because they are where the promise of a humanoid actually gets cashed. A robot that can walk into a building designed for people is only useful if it can then manipulate the objects in it: the zip, the dish, the charger, the folded shirt. Human hands combine strength, agility and flexibility in a package that has resisted decades of mechanical imitation, which is why so much humanoid capital has quietly concentrated on this one joint of the problem.
Meanwhile, everyone else has stopped waiting for general dexterity and started selling around it. Robot.com launched R-noid this week, a humanoid pitched at "the repetitive, multi-shift, and hard-to-staff jobs" and rented out under a robot-as-a-service model with eight to twelve weeks from first site visit to autonomous operation. Its five launch roles are strikingly narrow: restaurant assistant, packer, picker, folder, host. The economics it cites are not about intelligence at all. Quick-service restaurants see staff turnover above 130%. The average warehouse picker stays 1.2 years. More than 67% of hotel operators report critical gaps in housekeeping and laundry. Chief executive Felipe Chavez Cortes offers the softest possible framing, that "the future of work isn't fewer people, it's people freed from the parts of the job that grind them down," but the harder line in the announcement is the one that actually sells: "R-noid never resigns."
In Bengaluru, Axiom Robotics has unveiled the Axiom-1 for Indian manufacturing, priced to compete with conventional robotic arms rather than with imported humanoids, hardened with sensors for dust and heat, and aimed at small and medium enterprises via the government's Production Linked Incentive scheme. And in the research lane, the ergoCub project is going the other way entirely, building a humanoid designed around anticipating human movement and working safely alongside people rather than replacing them, integrating vision, proprioception and touch into a shared model of self and others.
Put these together and the shape of the industry in mid-2026 is clear. The general-purpose humanoid, the one that does anything a person can do, is still a research programme, and its bottleneck is a hand that can be trusted without a teleoperator. The commercial humanoid is a much humbler animal: a machine rented by the month to fold, pick and greet in jobs that humans keep quitting. The gap between those two is where all the money is, and it is currently bridged by a labour shortage rather than by a breakthrough. If 1X's hand really does work autonomously at scale, that gap closes fast, and the pitch changes from "R-noid never resigns" to something considerably harder to put in a press release.