Laundry Is Becoming the Home-Robot Proving Ground

Three Home-Robot Strategies, One Laundry Basket

The comparison is useful precisely because the robots are not the same. Memo and Isaac 1 use wheeled bases. NEO walks on two legs. Memo is moving toward a household beta with retail pricing still undisclosed. Isaac 1 has a stated preorder price and fall-2026 shipment target. NEO has a consumer preorder path and a higher-priced humanoid body intended to reach far beyond laundry.

Why Laundry Is So Much Harder Than It Looks

Rigid objects are kind to robots. A mug keeps being a mug when you grasp it. A box has edges that stay where the vision system expects them. Clothing changes shape every time it moves. A T-shirt can be flat, crumpled, twisted, inside-out, partly covered by another garment, hanging over a basket, or lying on the floor with the sleeve hidden underneath the torso.

That makes laundry a compact version of several unsolved home-robot problems at once. The robot needs to infer garment state from partial observations, decide where to grasp, coordinate two hands, preserve a fold while regrasping, recover after the cloth slips, and know when the result is good enough to stack. The task is also long horizon: a useful load is not one grasp. It is dozens or hundreds of decisions chained together without a human resetting the scene after every mistake.

Research is moving in the same direction. Physical Intelligence’s π0.7 generalist robotics paper reports zero-shot cross-embodiment behavior that includes folding laundry on a robot that had not seen that task on that embodiment. That is not a consumer-product result, but it reinforces the strategic point: laundry has become a useful test of whether manipulation skills can transfer rather than being rebuilt from scratch for every robot and room.

Sunday’s ACT-2 Result Deserves Attention — and Context

Sunday Robotics has published the most detailed laundry reliability claim of the three companies. In its July ACT-2 preview, Sunday reports 785 autonomous fold attempts across nine major garment types, with an overall success rate of 99.1% ±0.3% standard error. Sunday defines success as a garment being folded and stacked autonomously.

The distribution is more interesting than the headline percentage. Sunday reports 514 attempts starting from a pile on a bed, 73 from a basket on a bed, and 198 from a basket on the ground. It varied robot position and sheet color and included shorts, pants, leggings, T-shirts, polos, blouses, sleeveless tops and thick and thin long-sleeved garments. Blouses were the weakest category at a company-reported 94.7% success rate, while several categories reached 100% in the reported sample.

Sunday also reports fold quality rather than success alone. Across 778 completed folds, the company says ACT-2 averaged 4.72 out of 5, with 98.3% meeting its four- or five-star quality threshold. The median successful fold took 2 minutes 13 seconds. Those speed numbers matter: a robot can be extremely reliable and still fail the household value test if a normal load takes most of the afternoon.

Sunday deserves credit for publishing enough detail to interrogate the claim. Its own “Solve” framework argues that a robotics result should state three things: performance, scope and adaptation cost. That is a much better standard than a viral demo. It also gives outsiders a clearer checklist for what still needs to be tested.

The most consequential question is whether this recipe transfers from a controlled evaluation program into the company’s planned household beta. Real families do not preserve benchmark conditions. They interrupt the robot. They mix towels with children’s clothes and fitted sheets. They move furniture. They expect the robot to do dishes after laundry, not be reconfigured by a research team.

Sunday says Memo will go to families through its Founding Family beta in fall/late 2026. Until that happens, Biped treats Memo as a high-interest adjacent home-robot benchmark, not a verified consumer deployment. Its wheeled body also means it does not increase Biped’s core humanoid/bipedal counts.

Weave Is Trying to Sell the Whole Laundry Flow

Weave Robotics is making a different bet. Isaac 1’s product page does not stop at folding. Its “Laundry Flow” is supposed to find and pick up dirty clothes, handle loaded hampers, fold clothes and put them away. Depending on the home, Weave says the robot may also assist with washer and dryer loading and unloading. That moves the benchmark closer to what customers actually mean when they say, “do the laundry.”

Isaac 1 is also priced like a product rather than a research platform. Weave currently lists $7,999 upfront or $449 per month, with a $250 refundable deposit, first shipments beginning in fall 2026 and California deliveries first. Weave says Isaac is autonomous by default for Laundry Flow and Daily Reset, with teleoperation assistance when needed to guarantee task completion.

That last sentence may matter as much as the price. Early household robotics is likely to be a service stack, not just an autonomy model: robot hardware, onboard AI, cloud software, remote assistance, support, repairs and continuing skill updates. The economic question is therefore not only “what percentage of actions are autonomous?” It is “how much hidden human labor and support does it take to deliver one hour of useful household work?”

Weave’s wheeled base is a pragmatic choice for that experiment. Wheels reduce the balance problem, lower fall risk and avoid spending battery and control complexity on walking when most laundry happens on flat floors. The tradeoff is obvious too: stairs and some human spaces remain harder or inaccessible. The home market may ultimately segment by architecture rather than converge on one humanoid shape.

1X Is Asking Laundry to Prove the Humanoid Thesis

NEO carries the largest category burden. 1X is not pitching a laundry machine with arms; it is pitching a home humanoid. The company currently lists a $20,000 Early Access ownership option and a $499-per-month Standard subscription, with a $200 refundable deposit and U.S. deliveries beginning in 2026.

1X says NEO arrives with basic autonomy and uses its Redwood generalist AI to learn and repeat tasks. For complex chores it does not know, “Scheduled Expert Mode” allows a 1X expert to remotely supervise the robot at scheduled times. Laundry appears explicitly in that Expert Mode section. That is refreshingly concrete because it acknowledges that the early product will mix autonomy with human help.

Laundry also gives NEO’s new hands something meaningful to prove. In July, 1X disclosed a 25-degree-of-freedom tendon-driven hand architecture with force control, backdrivability and tactile sensing. Those are manufacturer specifications and claims, not independent proof of home-task reliability. But deformable fabric is exactly the kind of object where contact feedback, compliant fingers and fine regrasping should matter.

If NEO can perform laundry at comparable reliability while also handling doors, dishes, tidying and mobility through human spaces, legs and higher hardware complexity may pay for themselves through breadth. If a cheaper wheeled robot can deliver most high-value household chores sooner, consumers may care much less about whether their helper looks human.

Biped’s Proposed Basket-to-Drawer Benchmark

The industry still lacks a shared consumer benchmark for laundry. “Folded a shirt” is too easy to game; a single success rate can hide preparation, resets, garment selection, human intervention and time. For home buyers, Biped would rather see companies report a basket-to-drawer workload.

Start with a realistic mixed basket and end only when the useful household outcome is complete. The robot should have to retrieve, sort when necessary, fold or hang, stack, transport and put away the load. Then report the following metrics in a standardized way.

The Real Product Is Time Returned

Laundry is compelling because the value proposition does not need to be invented. People already outsource portions of it to cleaners, wash-and-fold services and household labor. A robot therefore has a natural comparison set: minutes saved, service cost avoided and how much supervision the user must provide.

But consumers will not calculate ROI like a warehouse manager. Home adoption will also depend on annoyance. Does the robot block a hallway? Wake someone up? Misplace a favorite shirt? Need the basket arranged just so? Require an expert to look through a bedroom camera? Turn a 15-minute chore into 40 minutes of troubleshooting? Reliability that feels invisible may matter more than peak dexterity.

This is where the three strategies diverge. Sunday is trying to push reliability and generalization high enough that autonomy survives novel homes. Weave is wrapping a narrower set of useful workflows in a concrete product and support model. 1X is spending more hardware complexity to pursue a broader humanoid that can move from laundry to many other chores. The winner may be the architecture that creates the most unattended useful minutes per dollar, not the robot that looks most like us.

Privacy and Remote Help Are Part of the Benchmark

Home robots operate where industrial robots rarely do: around bedrooms, medicine, children, visitors, personal conversations and daily routines. Laundry can be especially sensitive because the robot may work in bedrooms, closets and private storage areas. A system that needs remote assistance can still be useful, but the assistance model must be part of the product specification.

Weave says Isaac 1 uses teleoperation assistance when needed. 1X explicitly describes Scheduled Expert Mode and remote-control capability. Sunday’s public ACT-2 material emphasizes autonomous evaluation, but broad beta-scale exception handling and intervention-rate metrics are still to come. Those differences should not be reduced to a binary “autonomous/not autonomous” label.

The consumer questions are more practical: Who can see the camera? Is the owner notified before a human joins? Can rooms or times be excluded? Is video retained? Can it be used for training? How many minutes of remote intervention does a typical week require? A home robot that depends on human rescue is partly a robotics product and partly a remote-labor service. Buyers deserve to know both halves.

What Would Change Biped’s View

So, Which Home Robot Is Ahead?

On published laundry reliability: Sunday currently provides the richest company-reported evaluation evidence. Its ACT-2 report is unusually detailed for a startup robotics claim.

On a concrete consumer laundry product: Weave is the clearest. Isaac 1 has published pricing, a delivery sequence and an explicit Laundry Flow that tries to cover more of the chore than folding alone.

On the general-purpose home-humanoid thesis: 1X is making the biggest bet. NEO has a consumer order path and a body designed to operate across the human environment, but that ambition raises the standard it must meet.

None of those statements makes one robot the overall winner. The evidence is still too early and too heterogeneous. That is why Biped’s Home Humanoid Robots in 2026 guide tracks availability, task evidence, supervision, privacy and readiness separately instead of collapsing everything into a single hype score.

The Laundry Test Is Really a Trust Test

Laundry is not glamorous, which is exactly why it matters. It takes robots away from the stage and into repetitive work where failures accumulate, customers notice, and the economics cannot hide behind novelty.

If Memo can preserve its ACT-2 reliability as Sunday moves into real family homes, if Isaac 1 can turn Laundry Flow into a dependable paid service, and if NEO can make laundry one ordinary skill inside a larger humanoid repertoire, home robotics will have crossed a more important threshold than another viral demo.

The first great home robot may not be the one that can do everything. It may be the one that quietly gives people back the most time—and makes the laundry basket feel boring again.