From a functional perspective, I don’t get it: why would anyone want humanoid robots?

If the job is “move objects from A to B,” wheels beat legs on cost, reliability, energy use, and engineering simplicity. If the job is “clean floors,” we already have robots for that, and they don’t need knees, a spine, or a face. If the job is “open the door,” then the door can be smart. If the job is “lift heavy things,” forklifts have been around for decades.

We keep finding new ways to package the human body as a commodity and call it futuristic innovation, even when the result is stupid, fragile, power-hungry, slow, and expensive.

Humanoid robot evangelists have one serious technical argument: we built the world for humans. Stairs. Door handles. Narrow corridors. Tool grips. The pitch is for a general-purpose platform that can use the infrastructure we already have, rather than forcing every building to adapt to a new machine species (or become a machine itself).

But the practical story immediately collapses under the weight of its own constraints.

The body is the bottleneck

Humanoid robots won’t fail because engineers forgot to add more “AI”. They will fail because the human-shaped body is a brutal constraint.

For example, a recent paper notes that payloads rarely exceed ~20–25 kg and emphasises that energy efficiency is a critical bottleneck. Dynamic locomotion and constant balance correction chew through power fast. In other words, the very thing that makes a humanoid humanoid is an expensive way to move a box across a room.

And that’s before we get to the home butler fantasy. Domestic tasks aren’t just about strength; they’re about dexterity, judgement, and error tolerance in messy, high-variance environments. A warehouse can be engineered to be robot-friendly, sure. Our kitchens cannot. My kitchen, certainly, cannot.

So when I see a humanoid robot marketed as a household assistant, I see a business model searching for a plot, not innovation.

Take 1X’s NEO. A large part of its current deployment is dependent on teleoperation. That is, a human operator remotely controls the robot so the system can learn the chores and tasks it is expected to perform. But industry coverage is explicit: teleop is not temporary, it is their main path forward.

A teleoperated robot butler is not really a robot butler at all. It’s a device that lets someone else do your chores through a machine.

Which raises two concrete issues.

First: privacy. A teleoperated robot in the home is effectively a spy with its cameras, microphones, mapping capabilities, and continuous connectivity. Multiple outlets have made the obvious point. A device of this nature in your home creates privacy and security risks that go way beyond the usual smart speaker creepiness.

Second: labour. If “autonomy” is really remote humans doing the hard parts, then the techno-future is just old labour relations wearing a shiny mask. The hidden human work behind supposedly automated systems is well documented in the broader AI economy.

So when someone says, “Look, our robot can do the dishes,” I want to ask: whose hands are actually doing the dishes, and on what terms?

Why not just automate the environment?

A lot of humanoid use cases are actually arguments for ambient automation, i.e., making the environment smarter rather than building a person-shaped machine to stumble through it.

Door opening? Smart locks. Lights? Sensors. Temperature? Thermostats. Cleaning? Dedicated robots. Security? Cameras (locally controlled, please). Lifting? Assistive devices, hoists, and ergonomic tools.

A humanoid body is a universal adapter, yes, but it is the most expensive adapter imaginable. If a problem is frequent enough to justify automation, it usually justifies purpose-built automation.

Which is why I keep circling back to the question: what’s left, once you strip away all the engineering and marketing pitch?

The philosophical answer is uglier

Here’s what I think is going on.

Humanoid robots are companions, helpers, attendants, but above all, slaves. We’re being offered access to a taboo relationship with impunity: a servant without wages, rights, scheduling, conflict, or conscience. A human-shaped other that you can command without moral consequences.

It’s clear that luxury goods are purchased not for utility but for status display. A person-shaped assistant is conspicuous by design. It announces that someone has the means to own another person in a world where we already decided that slavery is evil incarnate (…because we did… right?). That’s why the butler framing keeps returning.

The edge case that matters (and doesn’t justify the hype)

There are settings where a human-like form can plausibly help: assisted living, mobility support, rehabilitation, or situations where adapting the environment is impossible and the goal is to interact safely in human spaces. But those are precisely the settings where privacy, safety, accountability, and regulation must be the strictest.

If your pitch is a teleoperated camera-microphone body wandering around private homes so rich people can outsource inconvenience (and tech bros can harvest even more personal data to sell to third parties), then we are watching yet another blatant failure in our civilisation through the consolidation of a new servant economy in a latex skin.

I hope this trend dies. Not because I hate robots, but because I don’t trust what this particular kind of robot is for.