Péter Fankhauser, Bernt Børnich, Amanda McMaster: The $1/Hour Worker: Four Robotics CEOs on Humanoids at Home, China's Threat, and the End of Dangerous Jobs

This part of the episode was recorded in Paris at a conference called Machina, described as "AI in the real world." After a sponsor read, it opens a series of interviews with robotics companies. The first is with the co-founder and CEO of a quadruped-robot company (the transcript spells it "Ennebotics"). The second is with the founder and CEO of OneX, maker of the Neo household humanoid. The transcript has no speaker labels, so this article describes the exchanges without assigning individual lines to named people.

The company is described as about ten years old, with its founder working in the field for close to twenty. The first five years were essentially a research lab, and the last five have been commercial. The product is framed less as a robot than as an inspection solution for collecting and understanding data in critical infrastructure. The four-legged "dog-obot" is just the form factor that delivers it.

Asked why the dog shape became the first to reach scale, the answer draws on nature and on the job sites. Four legs give mobility: the robot can climb stairs and "go anywhere a person can go." They also give balance and stability on a wide footprint with many footholds, which matters because the robots work in "nasty environments" with slippery floors, rain, snow and growing grass.

The interviewer half-jokingly proposes centaur-style humanoids, noting that two-legged robots such as Optimus, Neo and Boston Dynamics' Atlas keep falling over in demos and could injure someone. The reply is that the right form depends on the use case. A humanoid may suit narrow spaces like a coffee shop, or tasks done at eye level. The industrial facilities this company serves have enough room for a four-legged robot to move around.

Hundreds of these robots have been deployed over the last five years. The interviewer puts the purchase price in the "low hundreds of thousands of dollars," which the other side accepts. The interviewer also assumes service contracts of tens of thousands a year. That second figure is the interviewer's assumption and is not confirmed in the excerpt. The robots are industrial and heavily equipped with sensors, not built for the home.

For us it's not about labor replacement, right? It's what can we do better? What can we do superhuman?

Inspection is the example given. Human eyes and ears miss signals such as micro gas leaks or overheating equipment. The robot carries thermal cameras, acoustic microphones, gas-concentration sensors and AI. The financial case is avoided downtime: when these assets stop, they reportedly lose revenue in the hundreds of thousands per hour, so each minute or hour saved "essentially pays for the robot." That margin is what justifies expensive sensors and GPUs on the machine.

They don't care about the robot. Actually, they don't even want the robot. They want the data. They want the insights.

A mission lasts about two hours. The robot then returns to a docking station to charge and goes out again. After some back-and-forth over whether the number was 14 or 40, the figure is clarified: some customers run missions 40 times a day. One example is an electric arc furnace, where the customer wants to know what happens at the specific minute the furnace ramps up. It is too dangerous to send a person, and the transcript notes thermal cameras have burned. Batteries are charged at the dock rather than hot-swapped, because the goal is hands-free autonomy that nobody has to think about.

Computing is split between the robot and the cloud. Obstacle avoidance and data-quality checks must run in real time on the robot, both because connectivity cannot be guaranteed and because "If you upload a blurry image with the cloud, it's too late." The cloud handles contextual analysis and historical downtime analysis. Customers are not asking for 24-hour operation. The maximum is described as around eight hours, which leaves time for charging, with diminishing returns beyond that.

The comparison offered is with manual inspection. People might inspect once or twice a day, while the robot can do it 8, 15 or 20 times. That raises frequency without putting people in harm's way, and the data is described as higher quality.

Offshore sites are named as the most exciting deployments. Crews fly out by helicopter at tens of thousands per flight, almost nobody is on site, and the equipment has to keep running. These sites include oil rigs and offshore wind. For wind, the robot does not work in the water. It works at the transformer station where hundreds of turbines connect, a facility that is typically manned and contains large converters. The transcript says this station converts "AC to DC" before transmission.

The robots have been deployed in Norway at minus 20 degrees and in deserts at plus 40, 50 or 60 degrees, with dust and humidity as further stresses. The point stressed most is a special robot built for explosive atmospheres in oil, gas and chemicals, where methane may be in the air and any spark is forbidden. That robot is described as guaranteed not to create a spark. The interviewer raises gas leaks in the Permian Basin as a setting where people seriously get hurt.

Asked about putting robots on the ocean floor, the answer is that the company does not do that now. It agrees with the principle that robots should work where people shouldn't be: dangerous, remote, and boring repetitive tasks. The interviewer adds that other companies are building surface and shallow-subsurface robots for monitoring, including for the military.

The interviewer calls fixing a leak once it is detected the "holy grail." The answer is direct: "Not today." It may work in a demo, but reliability is the obstacle.

But in reality, getting it to 99.9% reliability in explosive atmosphere.