The first question a reporter asked about the machine was whether to call it he or she.
This was a warehouse south of Seattle, in the fall of 2023, a working fulfillment center that doubled as a place to test things. Two robots were moving totes on the floor, close enough to human that the question came out before anyone could stop it. The man who ran the robot company answered plainly. We use it for Digit, he said. His head of communications sharpened the point: it is not a person, it does not have a gender, it is a machine.
I keep coming back to that room, and I still cannot settle what happened in it.
Look at the body first. About five foot nine. LED eyes set in the head that blink to show which way it is about to turn. Cameras, sensors, and lidar to scan the room. Two arms. Legs that bend the wrong way, backward at the knee like a large bird, an ostrich or a crane. It lifts about thirty five pounds. It walks at roughly one and a half meters a second, which the company will tell you, unprompted, is close to the speed a person prefers, because it was built for rooms built for people. A reporter asked the Amazon engineer running the test whether workers would find the thing creepy. She said it was a really interesting form factor, and that was why they test, to get the feedback.
Now the job it was doing. Amazon called it tote recycling. When the inventory has been picked out of a plastic bin, the empty bin still has to be moved somewhere. That is the task. Not saving a life, not solving a puzzle, not choosing anything. Picking up a container that is empty because a person, or a machine standing in for a person, already took everything out of it, and setting it down again.
That is the image I cannot put down. The tote was already empty when the robot reached it. Whatever looked like the point had happened before. What was left was moving the emptiness from here to there, over and over, for a shift.
Set the scale beside the scene. In June of 2026 the company that makes the robot agreed to go public at a pre-money value of two and a half billion dollars, with more than six hundred and twenty million in expected proceeds, the largest raise in the short history of humanoid robots. It said its machines had logged more than sixty five thousand hours across nine customer sites. It said it had booked more than three hundred million dollars in milestone-dependent orders for the next model, a thousand robots on a three year rental contract to a single customer it would not name. A factory in Oregon is built to make ten thousand of them a year. The company burned about a hundred million dollars in 2025 and had not disclosed revenue in its initial filings.
All of that capital, all of that engineering, aimed at a body that can pick up an empty box.
Here is what I think the money is actually buying, and it is not the thing the headlines say. The headline is that the machines are coming for the work. The truer and stranger thing is quieter. The machine can do the work only because the work was already made machine-shaped. Before a robot could move the empty tote, someone had to carve out a task whose whole content was moving the empty tote. The person-shaped robot is not the start of that story. It is the receipt.
You can hear this in the builders' own words if you listen for it. The company says it went looking for its first customers in the places where the work had become, in its phrase, increasingly structured, repetitive, and robotic. The woman who runs it now describes a modern warehouse as mostly automated, with pockets of humans left inside it who serve as the glue between the machines. The glue between the machines. The robot is not replacing a craftsman. It is being slotted into a gap that the machines on either side had already defined, a gap where a person used to stand.
At a logistics warehouse in Georgia you can see the exact shape of the gap. A small wheeled robot rolls up carrying totes from the picking area. Digit lifts the totes off the cart and sets them on a conveyor, which carries them away to be packed. That is the whole loop. When the conveyor backs up, or when the human workers go on break and the line stalls, the robot stacks the totes to one side and waits, then feeds them back once things clear. Even the machine's day runs on the rhythm of a human shift. It is a link in a chain, standing where a link used to stand.
I want to be fair to the case for this, because it is strong and it is not cynical. The work is genuinely bad for bodies. Warehousing runs on lifting and bending and reaching and pushing, shift after shift, and the government safety office recommends keeping the boxes people pick under thirty five pounds precisely because backs and shoulders and wrists give out. In a single year the sector logged seventy seven thousand injuries and illnesses serious enough to record, tens of thousands of them bad enough to cost days of work, across nearly two million jobs. The man who ran the robot company put it cleanly: there is an awful lot of dull, dirty, and dangerous work that people probably should not be doing, work that is repetitive or hard or both, and just causes wear and tear. If a machine takes the tote so a spine does not, that is a real good.
But hold the two facts together and refuse to drop either one. Relief from moving empty totes all day is humane. And the job of moving empty totes all day was something we made. Nobody is born to be the glue between two conveyors. That role was engineered, step by step, over years, by taking a person's work and cutting it into pieces small enough and plain enough that a machine could one day copy them. The company that sells the robot likes to say the deployment lets humans be more human. The task it is removing is picking up empty plastic. The whole argument is folded into that sentence, and it is not comfortable.
To be more human, first you spend years being asked to be less.
None of this began with legs. The warehouse was reshaped long before anything walked. Fourteen years ago the same retailer paid about three quarters of a billion dollars for a startup whose squat robots carried shelves to the workers, so the workers would stop walking to the shelves; it now runs more than half a million of them. Then the job itself was taken apart. A robot to move the cart. A camera to read the label so no one had to hold the scanner. A system that lifts the bin to you and holds it at the exact height between your thigh and your chest, so you never reach up or squat down. Each step was sold, honestly, as easier on the body. Each step also lifted one more decision out of the person, until what remained was a motion so simple you could write it in four words. Bring the tote here. The humanoid did not arrive to do a human's job. It arrived after the job had been emptied out, the way the tote was.
It is early, and I do not want to oversell how far the machines have come. They still work behind barriers, or off in a fenced corner of the floor with no people near them, because a walking machine that weighs as much as a person and might fall is not yet safe to stand beside. In April 2025 one large logistics firm had two of these robots against more than a hundred and fifty thousand human employees. An industry analyst guessed that fewer than one in twenty robots entering workplaces that year could even walk. The co-founder is blunt about why: the reason they have deployed so few is safety, and that is the blocker for everyone. So no, the warehouse is not full of them. But that cuts the wrong way for the comfortable story. Even now, slow, alone, fenced off, the robot can already do the task. Not because it got so good. Because the task got so small.
There is one more thing in the record from that first year. When videos of the robot moving totes went around, some Amazon fulfillment center employees watched online and said, a little flatly, that the thing was slow, that they could do it faster. They are probably right. And under the small pride of it I keep hearing something the comments do not quite say out loud: that the work is work a machine can be built to imitate, and that everyone can now watch the imitation being made. You can be faster than the robot and still be the one it was built to stand in for.
The people who made the machine were careful about the pronoun, and they were right to be. It is not a person. It does not have a gender. It is a machine, and calling it one is honest. What the scene will not answer is a smaller and worse question. Long before anyone built a thing we would be so careful to call it, we built a job we were not careful about at all. We shaped the work until the truest description of the person doing it was the same one you would write for the machine now taking it away. The robot gets to be it. We should sit with how close we came to meaning the same word for the person beside it.