Most robots do party tricks. They fold one shirt, on camera, with an engineer standing just out of frame. Dyna Robotics, a Redwood City startup, just unveiled a machine built for the opposite: Taku, a service robot designed to complete entire workplace chores start to finish, with nobody babysitting it.

The announcement landed around September 29–30 and was widely reported October 1. Taku — named for takumi (匠), the Japanese word for “master of a craft” — is a semi-humanoid “full-stack physical agent” that loads washing machines, folds towels, chops vegetables, and stocks shelves. The company released an uncut demo showing an hour-long hotel laundry cycle. No cuts, no engineer hovering. Just a robot doing laundry for an hour.

That uncut hour is the whole pitch.

Boring on purpose

Look at Taku and the first thing you notice is what it doesn’t do: it doesn’t walk. The robot has a human-like upper body with two 7-degree-of-freedom arms, a folding lower body that reaches high and low shelves, and a four-wheeled base. The arms move at near-human speed. The legs — there are none, deliberately.

This is the most interesting design decision in service robotics right now. Bipedal walking is a magnificent engineering achievement and, for most jobs, a terrible idea. Wheels are faster, more stable, more energy-efficient, and far less likely to fall over in a commercial laundry. Dyna chose the practical morphology. Boston Dynamics’ Atlas just got a remarkable new dexterous hand, and Atlas walks beautifully — but Atlas is a research platform working toward factory deployment. Taku is a product for hotels that need towels folded tonight.

CEO Lindon Gao put the philosophy bluntly: “If a human still has to feed the robot, clear it, reset it and fix everything that goes wrong, then the robot hasn’t completed a whole job.”

That sentence should be printed on the wall of every robotics lab on Earth. The industry’s dirty secret is how much human labor props up every “autonomous” demo. Taku’s bet is that finishing the job — the whole job, including the boring parts — matters more than any single capability benchmark.

Already working

Dyna says the technology is already in commercial use at hotels, restaurants, and laundromats. That’s a meaningful claim: it means Taku isn’t a prototype looking for a market, it’s a product with customers. One background report cites a 24-hour test run with 99%+ folding success — the kind of reliability number that matters more to a hotel manager than any dexterity showcase.

The money has noticed. Dyna has reportedly raised $143M, with investments from NVIDIA, Amazon, and Salesforce. When the three companies that define cloud infrastructure, e-commerce logistics, and enterprise software all back your laundry robot, the thesis isn’t really about laundry. It’s about physical AI — machines that do useful work in the real world — becoming an investable category.

That tracks with the broader data. Global humanoid robot shipments surged 432% in the first half of 2026. An $18,000 humanoid just debuted in the U.S. The hardware is getting cheaper and the software is getting better at the same time. Taku sits at the intersection: affordable-ish morphology, serious software, and a job customers will pay to automate.

What it means

For hospitality and food service: labor is your biggest cost and your hardest problem. A machine that genuinely completes laundry, prep, and stocking workflows — not as a demo, as a shift — changes the staffing math. The early adopters will be hotels and restaurant groups already struggling to hire; the question is how fast the economics work at scale.

For workers: the honest version is that these jobs are exactly the ones humans don’t want to do at 2 a.m. for minimum wage. Automation of towel-folding isn’t the story; what happens to the workforce around it is. The companies that deploy well will retrain; the ones that don’t will just cut.

For investors: the “full-stack physical agent” framing is the signal. The winners in service robotics won’t be the best arm or the best vision model — they’ll be the companies that own the whole workflow, from perceiving the mess to finishing the job. Dyna’s $143M says the market agrees.

Taku won’t do your laundry at home. Not yet. But somewhere in a hotel laundry room, a wheeled robot with two arms is folding towels right now, unsupervised, for an hour at a time.

The home question

Some of the coverage framed Taku as a household-chores robot, which is worth correcting. Taku is a commercial product for hotels, restaurants, and laundromats — controlled environments with predictable workflows and a clear ROI calculation. Your kitchen, with its idiosyncratic drawer handles and the mug collection, is a much harder problem.

That’s the right order, though. Commercial first is how every transformative machine arrived: the dishwasher started in hotels, the microwave in restaurants. Controlled environments let companies like Dyna rack up operating hours, harden the software, and drive costs down before tackling the chaos of a real home. The hour-long uncut laundry demo is a commercial credential, not a consumer promise.

The path from hotel laundry to home laundry runs through reliability statistics and price curves. Dyna’s 99%+ folding figure, if it holds across deployments, is the kind of number that eventually makes the home version thinkable. Until then, Taku’s job is to make the commercial case undeniable — one folded towel at a time.

The future of robotics was supposed to walk in on two legs. It might roll in on four wheels instead.