1X's hand teaser puts humanoid dexterity back at the center of the home robot race
The post did not confirm a new launch, but it targeted the exact bottleneck 1X must solve for NEO.
By Ryan Merket ยท Published
Why it matters
Humanoid startups are moving from walking demos to manipulation proof. For 1X, NEO's home pitch depends on whether its hands can handle real objects reliably, safely and with less human supervision.
An X post teasing an advanced robotic hand from 1X pulled the home-humanoid debate back to where it will be judged: fingers, tendons, control, and endurance.
https://x.com/TheQuietFeed09/status/2074978211885183171
The post does not establish a formal launch, a new product name, a dated company announcement, or a spec sheet for the hand. That distinction matters. The verified story around 1X is still NEO, the home humanoid 1X opened for preorder on October 28, 2025, with company-claimed 22 degrees of freedom in its hands, a $20,000 Early Access price, and first consumer-home deliveries planned for 2026. A hand teaser can help focus attention. It cannot answer whether 1X has solved the hard part of domestic labor.
NEO targets domestic work, which is far less structured than a warehouse. A warehouse robot can be trained around bins, aisles, and repeatable workflows. A home robot has to grab mugs, laundry, cabinet handles, charging cables, toys, food packaging, and everything else people leave in the way.
The hand is where the home robot promise gets expensive
1X describes NEO around tendon-based actuation, a soft body, and 22-DoF hands. The NEO page also says owners can schedule a 1X Expert to guide chores the robot does not yet know, so NEO can learn while the job gets done.
That is the commercial bridge: ship a robot into homes before full autonomy is solved, use supervised work to complete tasks, and feed the resulting robot data back into the autonomy stack. It is also why hand dexterity has become the fault line. Remote operation can move a robot through a chore. It cannot make brittle hardware acceptable if tendons stretch, fingers wear, contact sensing is weak, or grasp recovery fails after the fifth awkward object.
RuntimeWire reported in June that 1X had launched a World Model Lab and hired Luma AI veteran Sam Sinha to lead it, while 1X had not disclosed the lab's budget, hiring plan, or NEO shipment volume. 1X said in a June 4 announcement that large-scale embodied world-model pretraining is its path toward fully autonomous humanoids and that it aims to own the stack down to pretraining its own video foundation models.
A better hand would fit that roadmap, but the public record does not yet say what the teased hand is. It could be new hardware for NEO, an internal prototype, a control-system demonstration, or a marketing cut of existing work. The source post's discussion around tendons, durability, control, and real home tasks is the right discussion precisely because the available evidence stops short of a product claim.
Figure made the hand race explicit
Figure launched Figure 03 on October 9, 2025, and made the hand a central part of the robot's pitch. Figure said each hand integrated an embedded palm camera for close-range visual feedback during grasps, and that its internally developed fingertip tactile sensors could detect forces as small as three grams. Figure tied those hand changes directly to Helix, its vision-language-action system.
That is a more sensor-explicit dexterity story than 1X has publicly told for the teased hand. 1X has disclosed NEO's 22-DoF hand claim and tendon-driven hardware. Figure has disclosed palm cameras, tactile sensing thresholds, and the control rationale behind them. The difference matters because humanoid demos increasingly depend on what happens after the first successful grasp: whether the robot knows a cup is slipping, whether a soft package is deforming, whether a drawer pull needs a different angle, whether the robot can adjust before it breaks something.
The capital around the category has also moved from research budgets to deployment war chests. Figure announced more than $1 billion in Series C funding at a $39 billion post-money valuation on September 16, 2025. Apptronik said on February 11, 2026, that it had closed a $520 million Series A-X extension, bringing its Series A total to more than $935 million. Those companies are largely commercial and industrial in their early deployment posture, but the technical race is the same: hands, data, control, and reliability.
Sanctuary AI offers the clearest dexterity-first contrast. In a recent demonstration, Sanctuary said its proprietary hydraulic hand autonomously reoriented a cube ten times without dropping it, using a policy trained in simulation and transferred to the real world. Sanctuary is focused on Physical AI for industry rather than selling a home companion, but its framing is useful: in-hand manipulation is a product category on its own, because useful labor depends on the hand doing work without the arm resetting the whole motion every time.
1X has a distribution path if the hand works
1X's advantage is that it has already chosen a brutally specific market. NEO is a home product first, priced for early adopters and designed to collect the kind of messy interaction data robotics labs struggle to simulate. 1X also has a commercial path through EQT. On December 11, 2025, 1X and EQT announced a strategic partnership framework intended to make up to 10,000 humanoid robots available across EQT's portfolio companies between 2026 and 2030.
That EQT arrangement gives 1X a second proving ground. Homes are varied and privacy-sensitive. Portfolio-company environments in logistics, facility operations, warehousing, manufacturing, and healthcare can be constrained enough for pilots while still producing real operational data. If NEO's hands can survive repeated contact, varied objects, and human proximity in those settings, 1X gets a cleaner story than a living-room demo can provide.
The privacy and autonomy questions remain attached to the home strategy. Tom's Guide raised concerns in October 2025 about Expert Mode, remote operation, and camera access, while noting that human-in-the-loop control is also how 1X can gather the data needed to improve NEO. 1X's own NEO page says experts can guide the robot through unknown chores. For buyers, that makes the hand problem inseparable from the trust problem: a teleoperated robot still needs safe hardware, and an autonomous robot needs enough sensing and control to avoid turning routine chores into household incidents.
The hand teaser is useful because it points at the right metric. Humanoid robotics spent years proving that robots could walk, wave, and carry boxes in polished videos. The next phase will be judged by quieter motions: pinching a cable without crushing it, folding fabric without losing track of an edge, opening a cabinet without scraping the finish, recovering when an object shifts mid-grip. If 1X wants NEO to be judged as a product rather than a supervised research platform, 1X's team has to make the hand boringly reliable.