The most consequential number in humanoid robotics this month is not a funding valuation or a sprint record. It is twenty thousand. That is the size of the pre-order backlog 1X Technologies says it will begin shipping into homes this year, as Forbes reported this week in a piece tracking the industry’s broader retreat from safety cages. Agility now markets Digit 5 as engineered for cage-free cooperative work. Boston Dynamics says production Atlas supports fenceless guarding. Tesla’s Fremont line is building Optimus without a cage story that has survived scrutiny. Each of those machines operates, or will operate, in a workplace governed by an industrial safety stack: ISO 10218, ANSI/RIA R15.08, OSHA general duty clauses, insurers with decades of automation data. Neo’s destination is a living room. None of that stack applies there, because no consumer safety framework for a 66-pound autonomous biped has ever been written.
This is the fault line the humanoid industry will cross in the next twelve months, and it is worth examining precisely, because the gap between “cage-free” and “home” is far larger than the gap between “caged” and “cage-free” ever was.
What twenty thousand customers are actually buying
The commercial facts are straightforward. Neo is listed at $20,000 to own with a three-year warranty, or $499 per month by subscription, secured with a refundable $200 deposit, with United States deliveries advertised to start in 2026. More than ten thousand customers had placed deposits by mid-year, according to The New Yorker’s reporting, and the figure has since doubled. The buyers profiled are exactly who you would expect: wealthy early adopters concentrated in the Bay Area, Los Angeles and New York, people who describe the purchase the way one former OpenAI employee did, as “kind of like the new iPhone.”
The machine they are buying is genuinely distinctive engineering. Neo stands five feet six, weighs 66 pounds, moves on tendons rather than direct motor drives, and runs at 22 decibels, roughly the volume of rustling leaves. Its proportions split the difference between the median American male and female body. Its head contains the compute; its chest contains the speaker. It wears a knit bodysuit, gloves and padded shoes, an aesthetic 1X iterated toward specifically because earlier, sleeker faces frightened children under twelve.
But the product description that matters most for this analysis is what Neo does not guarantee. Asked whether the latest models always remain upright, 1X’s head of product and design, Dar Sleeper, told The New Yorker: “To say it doesn’t fall is, like, a total stretch.” Previous versions fell often. The company says that problem is resolved. The person who designed the robot declines to say it is solved.
The candor already on record
What makes the home safety question unusual is that nobody has to speculate about the industry’s private doubts. The principals have stated them publicly, in a New Yorker feature published this summer, and the quotes deserve resurfacing now that shipment is imminent.
Bernt Børnich, 1X’s CEO, agrees that Neo should not be used around small children. Jeff Cardenas, CEO of Apptronik, put it more broadly: “Around small pets, around small children, there’s still work to be done.” Carolina Parada, who heads robotics at Google DeepMind, said home deployment “comes later” and framed the obstacle not as interest or capability but as safety, full stop. Aaron Ames, the Caltech mechanical engineering professor whose lab works on bipedal control, was blunter about 1X specifically: “I would worry about the legal ramifications when one of those robots falls on a person.”
Sit with the configuration those statements describe. The manufacturer’s own CEO advises against operating the product around small children. The product ships into homes, which frequently contain small children. The product’s designer concedes it falls. And the target environment is the one place in the economy with no safety certification regime, no trained supervisor, no exclusion zone and no incident reporting structure beyond a customer support channel.
Børnich has a term for the experience he is actually selling: “robotics slop,” his phrase for tasks clumsily executed. In a Wall Street Journal demonstration, Neo took longer than a minute to retrieve a bottle of water. The 1X wager is that early adopters will tolerate this, that a slow, imperfect household servant is still desirable at $499 a month if it improves weekly. That is a defensible product thesis. It is also, read against the safety admissions, a live experiment in how much slop customers accept before a robot’s clumsiness stops being endearing and starts being a hazard report.
The incident record, and why homes change it
Skeptics can already point to what failing bipeds look like in public. A Unitree G1 at a Chinese amusement park kicked a small child in the stomach during a scripted appearance. An XPeng IRON at a mall demonstration in Shenzhen seized and collapsed, unrecoverable, and three human handlers had to drag the 150-plus-pound machine away. Neither incident produced serious injury. Both occurred in environments with professional staff standing within arm’s reach.
That is the critical asymmetry between demonstration and deployment. Every documented humanoid failure so far has happened inside a managed setting: a stage, a mall, a lab, a trade show floor, a warehouse aisle with a safety officer. The 20,000 Neo shipments will be the first large-scale deployment of general-purpose humanoids into completely unmanaged environments, occupied by the people least equipped to respond to a failure mode: children, elderly relatives, pets, guests who did not consent to anything. The New Yorker’s reporting also noted research showing that AI guardrails can be circumvented by persistent, creative users, and illustrated the scenario of a child eventually talking a household robot into a destructive act. Every parent of a four-year-old understands that “persistent and creative” is the job description.
There is also a security surface no household appliance has ever presented. Security researchers Andreas Makris and Kevin Finisterre demonstrated a Bluetooth vulnerability in the Unitree G1 capable of spreading robot to robot, “creating a botnet that spreads without user intervention.” A home humanoid is a networked mobile sensor platform carrying cameras, microphones and enough torque to manipulate physical objects. The IFR has flagged rising hacking attempts against robot controllers and cloud platforms as a primary safety concern for 2026. Factory fleets get security teams. Neo gets a home Wi-Fi router.
A safety stack built for workplaces, not families
The regulatory landscape makes the industrial-humanoid progress of the past year look better by comparison, which is precisely the problem. ISO 10218, the foundational industrial robot safety standard, was revised in 2025 and absorbed the collaborative-operation guidance of ISO/TS 15066. ANSI/RIA R15.08 extends the framework to industrial mobile robots. A dedicated humanoid standard, ISO 25785-1, addressing dynamic stability and the specific failure modes of bipeds, is in committee draft, as standards trackers have mapped. That is real, rapid progress: the industrial stack is being upgraded on a schedule that roughly matches the industrial humanoid ramp.
Every one of those documents governs workplaces. The consumer realm has UL listings, IEC household appliance standards and CPSC jurisdiction, a regime designed for toasters and robot vacuums, not for autonomous machines that walk, carry loads, ascend stairs and are remotely operable by third parties. Track-mounted robot vacuums generated years of privacy litigation and standards work while weighing a few kilograms and lacking arms. There is no ISO committee document, no UL outline of investigation, no CPSC rulemaking that certifies a humanoid for unstructured home use alongside children. Neo will not ship with a humanoid home-safety certification because no such certification exists to earn. First movers write the incident casebook the standards bodies will eventually codify, in the same way early autonomous vehicle deployments are still writing case law a decade after launch.
Agility’s Digit 5 launch materials are instructive as the industrial counterexample: a robot explicitly framed around OSHA ergonomics thresholds, cooperative safety engineering and facility integration. The industrial players treat the safety layer as the product. The consumer wave is inverting that: ship the experience, harden the safety case in the field, absorb the liability as a cost of learning. Tesla extracted vehicles from that school of hard knocks; it is not obvious a startup with a subscription product priced at $499 a month can do the same when the failure mode is a fall onto a toddler rather than a fender bender.
The stranger in the machine
The underdiscussed deployment fact is that Neo will not be fully autonomous, especially early on. 1X’s own demonstrations to press have leaned on teleoperators in VR headsets, and the company has been explicit that remote operation remains part of the product: earpiece rings change color when a human pilot takes over, and 1X has mused about teleporting operators in to “bartend your party.” The New Yorker piece describes the press demonstration robot as a marionette, with the autonomy showcase declined.
Set aside whether customers read the fine print. A teleoperated humanoid means strangers with live camera access inside paying customers’ homes, on demand, at a company that also sees teleoperation as a data collection channel. Nvidia research scientist Jim Fan’s summary of humanoid deployment logistics applies with special force to consumers: “Babysitting these robots demands an entire operation team. Mistakes are irreversible and unforgiving.” 1X’s answer, that subscribers would rather have a vetted remote operator than an unvetted house cleaner, is at least a coherent position. It is also a position that has never been tested against consumer protection law, wiretap statutes or the first viral clip of a pilot’s error being broadcast from someone’s kitchen.
Why 1X ships anyway
None of this is a mystery to the company, which is what makes the moment analytically interesting. Sleeper, asked whether 1X would hit its 2026 home delivery commitment, answered without hesitation: “It’s a promise we’ve made to the world, and a promise we’ve gotta keep.” That sentence is the real unit of account. 1X has raised capital, priced a product and built a factory schedule around being first into homes. Every quarter of delay hands the narrative to Figure, to Optimus, to whoever ships second with one more quarter of maturity. The rational strategy for a company that believes general-purpose home robotics is winner-take-most is to ship the slop, iterate in the field, and price the liability into the subscription.
And to be fair to the bull case, the engineering margins are real: 66 pounds is light for a biped, 22 decibels is quiet, tendons limit peak forces, textiles cushion contact surfaces, and the company runs an in-house safety lab that includes, per The New Yorker’s tour, technicians teaching robot hands to karate-chop cantaloupes. A soft, light, slow robot is the correct morphology for a first home product. The question is whether “softer failure modes” scales into “acceptable failure rates” in an environment where the occupant never signed a risk acknowledgment.
What to watch
Three signals will tell you how this resolves. First, the first serious Neo incident and 1X’s response: a fall onto a person, a privacy complaint, a teleoperation leak. Product recall behavior, disclosure speed and settlement posture will set the template for every consumer humanoid after it. Second, insurance: underwriters pricing a home humanoid rider is the moment the category becomes real, and the premium they charge is the honest market estimate of the risk the standards bodies have not yet quantified. Third, standards movement: watch whether ISO 25785’s scope widens beyond industrial humanoids, or whether a consumer humanoid committee forms at UL or the IEC with 1X, and eventually Figure and Tesla, in the room.
The factory humanoids spent 2026 earning the right to leave their cages. The home humanoids are about to discover that leaving the cage was the easy certification.