# Nori A3 Is $1,688. Who Teaches It to Fold the Laundry? **Summary:** Nori A3 puts a 19-DOF, two-armed mobile robot at a $1,688 price, but its product model asks the buyer to train and operate it. I see a compelling developer platform, not a finished household appliance. - Canonical: https://markhuang.ai/news/nori-a3-1688-who-teaches-it - Language: en - Author: [Mark Huang](https://markhuang.ai/about) - Published: 2026-09-01 - Section: News - Tags: Nori A3, Home Robotics, Humanoid Robots, Robot Learning, Privacy - Source: [Nori Robotics](https://www.norirobotics.com/) - License: https://creativecommons.org/licenses/by-nc/4.0/ --- ![A compact wheeled robot holds a folded towel beside a basket of laundry in a home utility room](https://cdn.markhuang.ai/news/nori-a3-1688-who-teaches-it/hero.webp) *A robot folding one towel looks like an appliance. The pile beside it is a better test of how much work the owner still has to do.* [Nori Robotics is selling the Nori A3 for $1,688](https://www.norirobotics.com/), with orders advertised to ship in fall 2026. The company shows a two-armed robot on wheels and lists kitchen help, tidying, fetching food, loading dishes, folding clothes, and pouring ingredients as everyday tasks. That price got my attention. My read, though, is that the A3 makes more sense as an affordable robotics platform than as a finished household appliance. Nori expects the buyer to teach, operate, and manage the machine. A developer may hear an invitation in those verbs. Someone who wants the laundry folded may hear a second job. ## The hardware is unusually accessible The A3 is a bimanual mobile manipulator, which is a less romantic and more useful description than "humanoid." It rolls instead of walking. According to [Nori founder Antonio Li's technical preprint](https://arxiv.org/abs/2605.16537), the robot has 19 degrees of freedom, weighs 20.4 kg, and fits on a 45 by 45 cm base. A telescoping column moves its head and arms from 69 cm to 145 cm, letting it reach from the floor to a counter. Nori's product page rates each arm for a 1.5 kg payload and gives the robot four 720p cameras, lidar, a microphone, and a speaker. The advertised battery life is 6 to 8 hours. Those are manufacturer figures, but they describe a serious amount of hardware at this price. The evidence is still close to the company. Li is the sole author of the five-page preprint, and the product page is a sales page. I would treat both as documentation of the design and Nori's claims, not as an independent durability or household-performance test. ## Who teaches it? The owner does Nori Lab is the center of the pitch. The product page says the laptop app lets owners "train, operate, and manage" the robot. Skills can then be shared through a marketplace. [New Atlas reports](https://newatlas.com/robotics/household-robot-chores-nori-a3/) that owners can write code or train the robot through a simulated version on a desktop. Its caution is fair: the A3 is not like a robot vacuum with a narrow job and a familiar expectation of what happens after unboxing. The [public Nori documentation](https://docs.norirobotics.com/) is marked beta and says full A3 documentation is coming soon. It leads owners toward a desktop app and a first teleoperation session. The site also offers Python and TypeScript SDKs and says developers can stream video and telemetry over WebRTC. I do not see that as a gotcha. A low-cost, programmable robot with remote control and an SDK could be useful to makers, classrooms, and small robotics teams. It does mean the $1,688 price answers only the hardware question. It does not tell me how long a skill takes to teach or how often it fails when a chair moves. The [Y Combinator company profile](https://www.ycombinator.com/companies/noril1) is direct about the larger plan. Nori wants affordable robots in many homes and businesses so their use can produce diverse training data for future generalist robot policies. The profile says, "Every user of Nori makes Nori smarter." If that works, early buyers help build the product's future capabilities. They are customers, but they are also part of the data strategy. ## A chore demo is not a chore guarantee The website shows the A3 doing recognizable work. That is better than selling a robot through specification sheets alone. What I still need is boring evidence: repeated task success rates, recovery after a failed grasp, setup time in a new room, breakage and spill rates, noise, charging behavior, and the hours an owner spends teaching one dependable routine. Those numbers do not appear on the submitted product page. The page says the A3 can help with household tasks, but it does not give a benchmark or a service-level promise for them. New Atlas reaches a similar boundary, warning that buyers may not get much ready-to-use functionality or a guarantee that tasks will run without mistakes. In a [Reddit discussion about the launch](https://www.reddit.com/r/technology/comments/1vsz8g3/american_humanoid_robot_launches_for_just_1688/), people asked about camera privacy, internet dependence, remote operators, stairs, clutter, and the difference between the 1.5 kg arm payload and the much larger vertical-lift figure repeated in coverage. The thread proves nothing about reliability. It does show what buyers want answered before a four-camera machine moves around private rooms. > **Info:** > > Before ordering, I would ask Nori for an unedited repeated-task demo, the expected teaching time, a clear local-versus-cloud breakdown, data-retention terms, and the recovery path when the robot drops or misidentifies an object. ## The cameras change the buying decision A trainable home robot needs to see the home. That does not settle who receives the video, which recordings become training data, or what happens when a shared skill was learned in someone else's kitchen. Nori may have good answers. I would want them in writing before connecting the A3 to a home network. A home contains people who did not buy the robot. Visitors, children, caregivers, and cleaners can all enter its cameras and microphones. I argued in [my piece on smart-home AI and domestic workers](/news/smart-home-ai-needs-worker-mode) that owner-only controls are a poor fit for devices that observe bystanders. A mobile robot carries that problem from room to room. The skills marketplace adds another layer. A shared skill could save owners hours of training, which is the best argument for Nori's model. It also needs a clear account of what travels with that skill. The policy, demonstrations, video, room geometry, and logs do not all deserve the same treatment. I would want the marketplace to separate reusable behavior from household data by default. ## I would buy the platform, not the promise For a robotics developer, the rough edges may be the point. The A3 puts two arms, mobile sensing, teleoperation, and an SDK into a price range where experimenting on real hardware becomes easier to justify. A school or small lab could learn more from a limited machine it can modify than from a polished robot it cannot afford. For a household buyer, I would wait for evidence from repeated use outside Nori's own demos. The decision changes when independent owners publish how long setup takes, how the robot behaves after small changes to a room, and whether downloaded skills transfer cleanly. Privacy terms and support matter just as much as another folding video. So who teaches the $1,688 robot to fold the laundry? At first, you do, or someone in the Nori community does. That bargain is why the price can open robotics to more developers. It is also why I would not yet mistake the A3 for an appliance that gives household time back on day one.