Talk Around Town: Dyna Robotics Claims Its Wheeled Taku Robot Can Solo a Full Hotel Laundry Shift

What Folks Are Saying Around the Feed Store

Well, butter my biscuit, there's a new robot in town and it's got people talking louder than a rooster at a library. Dyna Robotics, a Redwood City, California outfit founded in 2024, announced on September 29, 2026, that it had unveiled a wheeled semi-humanoid robot called Taku alongside what the company describes as its Dyna-2.1 physical-agent system, according to the company's own announcement and an exclusive preview it granted to ABC7 San Francisco.

The name, per Dyna Robotics, traces back to the Japanese word 'Takumi,' meaning a master craftsman or skilled technician — which is a real fancy handle for something that, as far as outside observers can confirm, has only been evaluated inside a company-controlled demonstration. ABC7's coverage was based on a preview Dyna Robotics arranged itself, meaning no independent soul with a clipboard and a stopwatch had a crack at it.

What Is Actually Confirmed, Bless Its Heart

Now here's where we separate the cornbread from the cake. The funding is rock solid: multiple independent financial data sources, including PitchBook, Multiples VC, and a client announcement from law firm Gunderson Dettmer, confirm Dyna Robotics has pulled in $144 million across two rounds — a seed in March 2025 and a $120 million Series A in September 2025. Investors confirmed across those sources include NVIDIA's NVentures, the Amazon Industrial Innovation Fund, Salesforce Ventures, CRV, and First Round Capital. That's the most buttoned-up fact in this whole barn dance.

It is also confirmed that Dyna Robotics unveiled something called Taku on September 29, 2026, and that the robot, as the company describes it, sports a human-shaped upper body, two arms each with seven degrees of freedom, a folding lower body for reaching high and low, and a four-wheeled base. Those physical design details are consistently reported, even if the performance claims attached to that hardware are a different kettle of catfish entirely.

What Dyna Robotics Claims — And Only Dyna Robotics

Here's where the mud gets deep, y'all. Dyna Robotics claims that in a one-hour, uncut demonstration, Taku autonomously completed a full hotel laundry workflow — loading a washer, moving towels to a dryer, then folding, stacking, and shelving them — without any human stepping in to help. The company also says its Dyna-2.1 system uses three layers: a whole-body controller running at 100 Hz via reinforcement learning, something the company calls the Dyna-2 world-action model for generating movement targets, and a vision-language orchestrator to track workflow state and memory. That's the company's own description of its own architecture.

Dyna Robotics further claims, via its X posts, that Taku can pick up brand-new whole-body tasks after just 30 minutes to a couple hours of on-robot training data, helped along by pre-training on what the company describes as millions of hours of human video footage. The company also states that its robotic technology is already running in hotels, restaurants, laundromats, and gyms — but not a single named customer has surfaced publicly to shake hands and say so. The Korea Daily and Linkdood pieces that mention deployment are drawing from the same Dyna Robotics source material, not independent customer accounts.

What Nobody Has Verified Yet, Bless Your Heart

No independent robotics researcher, no third-party tester, and no paying enterprise customer has been quoted evaluating Taku's real-world performance. The ABC7 preview was company-arranged, meaning Dyna Robotics picked the conditions like a farmer picks a prize pig for the county fair. Secondary outlets including Robotico Market, Complete AI Training, and Linkdood all derive their factual claims from Dyna Robotics' own materials rather than fresh reporting.

The company has also referenced a 99%-plus success rate from an earlier model called DYNA-1 running in a 24-hour non-stop scenario, a figure that originated in Dyna Robotics' own Series A fundraising announcement and has not been checked by outside parties. Treat it like a fish story told at closing time: possible, but unconfirmed.

The commercial deployment claim — that Taku-related technology is already operating in multiple hospitality and service-industry venues — is attributed to Dyna Robotics and not confirmed by any independently identified customer. That's a gap wider than a summer creek after a drought.

Our Analysis: Wheels vs. Legs and What It Might Mean

This is analysis, not reporting, so put on your thinking hat. Dyna Robotics is making a deliberate design bet: the company's position is that wheels beat legs for commercial service environments because those settings demand mobility, reach, and manipulation more than they require climbing stairs or navigating rough terrain. That framing is Dyna Robotics' own argument, not an established industry consensus. Competitors such as Figure, Agility Robotics, and 1X are all charging ahead with fully legged humanoids, and whether wheels or legs ultimately win the service-industry derby is an open question being debated across the robotics world right now.

The architecture Dyna Robotics describes — layering reinforcement-learned motor control, a world-action model, and a language-vision orchestrator — mirrors trends seen across several robotics labs, so there is nothing obviously implausible about the technical framing. But 'not implausible' is a long way from 'independently verified,' and right now the gap between the company's marketing language about 'reliable super long-horizon whole-body autonomy' and what outside observers can actually confirm is big enough to drive a tractor through.

The $144 million in confirmed funding does signal that sophisticated institutional investors — including industrial-AI-focused arms of NVIDIA and Amazon — were persuaded by something, whether private demonstrations, internal data, or broader market conviction. Whether Taku can do in a real hotel what the company claims it can do in a controlled demo is the question that only time, and some independent testing, will answer.