Mark Zuckerberg has spent three years telling anyone who would listen that the next computer would be worn rather than held. The glasses have been on his face in every product video since 2023, and they are the argument. This week the argument acquired a pair of legs.
According to The Information, which reported the work on Oct. 31, Meta Platforms assembled a humanoid prototype called Max in roughly three months, working with a humanoid robotics startup. The company has put home robots on its list of next-generation hardware and is already preparing an improved version.
Three months is not a normal interval for a robot. It is a normal interval for a partnership, which is what this appears to have been: a startup’s mechanical design and Meta’s models, assembled quickly enough to find out whether the software translates. The interesting question is not whether Max can walk. It is what Zuckerberg believes a body does for an assistant that currently lives in a pair of frames on his nose.
Hands First, Then a Body
Meta’s robotics researchers did not begin with a person. They began with fingers. The group’s published work over the past several years has concentrated on touch: sensors that measure the pressure and slip of a grip, datasets of manipulation recorded at high fidelity, models that let a hand pick up an object it has never seen. It is the least photogenic part of robotics and the part that decides whether a machine is useful in a kitchen.
Going from a hand on a bench to a machine that moves through a house means adding everything the hand did not need. Balance. Power. A safety case for a device that shares a floor with a toddler and a dog. Actuators, the joints that convert electricity into motion, are a supply chain of their own, and the credible suppliers are concentrated in China. The bill of materials for a capable humanoid runs into tens of thousands of dollars before anyone has written the software that makes it worth owning.
That gap is where most humanoid programs have stalled. Tesla has shown Optimus repeatedly and has not put it in a customer’s home. Figure, 1X and Agility have all announced timelines that moved. The demo is easy to stage. The product has to survive a spilled glass of milk and a firmware update that arrives at two in the morning.
The Bet Behind the Timeline
So why would Meta enter a hardware category with that record? The answer is in what Meta already owns and what it does not.
Meta owns models, capital and a distribution channel that reaches three billion people. It does not own a manufacturing culture. Reality Labs, the division that houses the glasses and the headsets, has lost money every year since it was broken out in 2020, and the losses have been measured in the tens of billions. A robot program inside that division inherits both the budget and the skepticism.
The three-month timeline suggests the company is trying to buy information rather than win a race. A prototype built from a partner’s parts and Meta’s models tests one thesis cheaply: that the value in a home robot sits in the perception and planning layer, not in the joints. If that thesis holds, Meta can partner its way to a product, the way it has entered most hardware categories it did not invent. If it fails, the company has spent a quarter and a small team.
There is a commercial logic underneath. The glasses work because they put a camera and a speaker where the assistant can see and be heard. They cannot pick anything up. A house is full of tasks that require a hand, and the companies selling robot vacuums have shown that consumers will pay for a machine that does one chore reliably. A general-purpose machine that folds laundry is a harder sell than a vacuum, but it is the only version that justifies a model as expensive as the ones Meta is training.
The competitive pressure is real. Amazon has been testing humanoid machines in its warehouses, work that started with industrial arms and has moved toward machines that can pick and carry. Apple has been reported to be exploring a robot for the home, drawing on sensor and chip work it already does. Chinese manufacturers can build hardware at a cost Meta cannot match and are already shipping lower-cost humanoids to factories and, in small numbers, to research labs.
Zuckerberg’s own pattern is instructive. He bought Instagram, WhatsApp and Oculus rather than build them, and he has been willing to spend years and billions on a category before it had customers. The glasses took a decade of iteration to become something people wear in public. The robot is at the stage the glasses were in 2016: a demonstration that convinces the person paying for it and almost nobody else.
Meta has not said when a home robot would be sold, at what price, or under what name, and the company has described the prototype only in the terms reported by The Information. What it has said, through the shape of the effort, is that the assistant needs somewhere to go after the phone. The glasses on Zuckerberg’s face are the first answer. Max is the second, and the harder one, because a machine that walks into a living room has to be useful before it can be charming.
