Latest / Minds, Bodies, and Terawatts / Boston Dynamics Makes New Atlas 'Order of Magnitude' Simpler to Build, Opening a Path to Mass-Produced Humanoids
Transcript
- 0:00[Host] This is Minds, Bodies, and Terawatts, where the robots get cheaper while you read the news. Forbes reported on July 2nd that Boston Dynamics just redesigned Atlas, its flagship humanoid, to be almost an order of magnitude simpler to build.
- 0:14[Guest] This is the robot that did backflips and soccer tricks, historically the hardest and most expensive humanoid on the planet, running upward of two hundred grand a unit. Their director of robot behavior, Alberto Rodriguez, told Forbes the new one has way fewer parts, and way fewer unique parts.
- 0:32[Host] Fewer parts usually means you cut corners. Did they lose capability?
- 0:36[Guest] Opposite. Rodriguez says same performance or higher. They stripped out the complexity and kept the athletics, and fewer unique components means it's faster to assemble, cheaper, and it breaks less.
- 0:47[Host] So why does a simpler robot matter more than a fancier one? I'd have thought the impressive part is the backflips, not the parts count.
- 0:57[Guest] Because now you can actually build a lot of them. Atlas's parent is Hyundai, and Hyundai is talking about at least thirty thousand units a year. You don't quote a number like that unless the manufacturing math finally closed.
- 1:11[Host] Thirty thousand a year of a machine that can lift, climb, and move like a person. Where do they all go?
- 1:19[Guest] Hyundai's keeping most of 2026's run for itself, its own factories first, outside sales later. So the first customer for a mass-produced humanoid is the company that builds the humanoid.
- 1:30[Host] This is the part that gets me. If the hardware was the hard problem, and they just solved it, what's left?
- 1:37[Guest] Rodriguez said it plainly. The body is ahead of the brain now. His words were that Atlas is capable of much more than they've been able to squeeze performance out of. Software is the bottleneck, not the machine.
- 1:51[Host] That's a flip. For years the story was that the robots were clumsy.
- 1:56[Guest] Right. And the honest defense of a human worker used to be that the machine was cheap but flaky. Once a robot is cheap and dependable, that defense expires. That's the line where the cheapest reliable worker for a task stops being a person.
- 2:12[Host] And you're saying we're walking straight up to that line.
- 2:16[Guest] One job at a time. It doesn't arrive economy-wide on a Tuesday. It crosses in one warehouse, one loading dock, one parts station. BMW just put Figure's humanoids on its South Carolina line last week doing exactly that.
- 2:30[Host] But thirty thousand robots is a rounding error against millions of workers.
- 2:34[Guest] Today it is. The number that should get your attention is the trajectory. The average humanoid went from about eighty-five thousand dollars in 2023 to twenty-five thousand in 2025, per industry data.
- 2:46[Host] That's two-thirds of the cost gone in two years.
- 2:49[Guest] A simpler Atlas is that curve continuing. A basic humanoid ends up costing what a dishwasher does.
- 2:56[Host] And nobody brags about owning a dishwasher.
- 2:58[Guest] That's the whole shift. The moment bodies get boring and cheap, physical work stops being a human monopoly. Not because the robot is smarter than you, but because it's cheaper and it shows up.
- 3:11[Host] So the real question stops being whether they can build it.
- 3:15[Guest] It becomes what we do when the work is already handled. Patrick Déglon calls that the Labor Cliff in his book Unscarcity, the point where machines automate bodies faster than the economy makes new roles for people. For more on the substitution threshold and life on the far side of it, head to unscarcity dot ai. The book is on Amazon, the audiobook is on Spotify. This has been Minds, Bodies, and Terawatts.