The home is the hard case
The state of humanoid robotics surveyed the whole field and found real, paid deployment concentrated in warehouses and factories, not living rooms. This page goes deep on the one exception that matters: 1X Technologies’ NEO, currently the only humanoid robot a consumer can actually order for home delivery. The question here is narrow and specific — what can NEO do on its own, right now, versus what a human is quietly doing for it — not the separate privacy and governance questions that teleoperation inside a home raises, which the companion page household robots: privacy and control covers in full.
This page is a capability and reliability assessment based on public reporting, company statements, and one independent journalistic test. It is not a product recommendation, a safety certification, or a substitute for reading a purchase agreement before putting a networked robot in a home.
What NEO costs and claims to do
1X sells NEO two ways: a $20,000 outright purchase with a $200 deposit, or a $499-a-month subscription with a six-month minimum — both plans include “Expert Mode,” the company’s name for human teleoperation of the robot. First U.S. deliveries are expected in the third or fourth quarter of 2026. By 1X’s own published specifications, NEO weighs 66 pounds, can lift more than 150 pounds, and has 22 degrees of freedom across its hands (1X). Those figures are the manufacturer’s own claims; nothing in public reporting independently verifies them against a household robot actually operating unsupervised.
What “Expert Mode” turns out to mean
When the Wall Street Journal’s Joanna Stern tested NEO for a hands-on review, she found that every household task the robot performed was in fact being driven by a remote human operator through a VR headset, not run autonomously. Retrieving a water bottle from the refrigerator took one minute and seven seconds, most of it spent on the mechanics of opening the door (WSJ, via secondary reporting). That finding matters because it is the one independent, hands-on test of the product publicly available, as opposed to a company demo or a press release. It does not prove NEO can never operate autonomously for simpler tasks; it does establish that, for the tasks demonstrated to a journalist, a human was doing the work through the robot rather than the robot doing the work itself.
This is not unique to 1X. The pattern recurs across the chapter’s opening survey: Tesla’s own executives admit Optimus remains in an “R&D phase,” and a widely circulated December 2025 Optimus stumble at a live event fueled speculation about a hidden operator. NEO is simply the case where the human-in-the-loop reality is now confirmed rather than merely suspected, because 1X built the entire commercial product around it and discloses it in its own marketing rather than only under outside scrutiny.
The mitigations 1X built around a known privacy problem
Because Expert Mode gives a remote human operator camera access inside a customer’s home, 1X has built specific mitigations: teleoperation sessions require the household’s explicit approval before they start, people visible in a room can be blurred in the operator’s feed, users can define “no-go zones” the robot will not enter even under remote control, and operators are vetted, with one manager overseeing roughly eight teleoperators at a time. 1X’s chief executive, Bernt Børnich, has defended the arrangement directly, describing it as a kind of “social contract” with early adopters and stating plainly, “If we don’t have your data, we can’t make the product better” (Humanoids Daily). He has separately argued the model can be more secure than hiring a human cleaning service, on the reasoning that vetted, monitored, camera-logged remote operators are a more auditable arrangement than an unsupervised person alone in the house. Whether that argument holds up — and what it means for data retention, third-party access, and a household’s actual legal leverage over a remote operator on the other side of the world — is a distinct governance question the household robots: privacy and control page addresses in depth; it is flagged here only as the direct consequence of the capability gap this page documents.
What this means for “home robot” as a category
It is worth being precise about what NEO’s current reality does and does not show. It does not show that autonomous household robotics is a hoax — 1X, like its competitors, is explicit that it is using teleoperated sessions partly to collect training data intended to reduce the need for a human operator over time. It does show that, as of today, the category’s first commercially shipping product is closer to a monitored, camera-equipped remote-labor service wrapped around a robot body than to an autonomous appliance, regardless of how it is marketed. A buyer paying $20,000 or $499 a month is paying, in significant part, for human labor delivered through a machine.
Morgan Stanley’s own analysts, covering the humanoid market broadly rather than 1X specifically, are explicit that this gap is expected to persist for some time: “We are not going to see a robot in every home overnight,” they wrote, adding that general-purpose home capability needs “about another decade” of development (CNBC). That is a mainstream financial-analyst view, not a critic’s view, and it lines up with what NEO’s own launch reality shows: the hardware can exist, the business model can exist, but the autonomy that would make it something other than teleoperation has not yet arrived on any independently verified timeline.
The bottom line
Judged strictly on capability rather than marketing, 1X’s NEO is the most honest data point the home-robot category currently offers — not because it performs best, but because its teleoperation dependency has been independently tested and publicly confirmed rather than merely inferred from a stumble on stage. Treat any home-robot claim of autonomy through 2026 with that baseline in mind: ask what specific task was performed, under what supervision, and whether the source is the company itself, a demo reel, or an independent hands-on test. On the evidence available as of September 13, 2026, no home humanoid robot has publicly demonstrated open-ended autonomous household competence; what exists is a monitored human-robot hybrid, priced and marketed as a product.