The Clumsybot and the Vending Machine We Already Have

There’s a video doing the rounds of a humanoid robot in what looks like a retail store, fumbling a shelf retrieval and making a bit of a mess. The kid nearby looks delighted. The internet, predictably, lost its mind.

The comments split pretty cleanly into two camps. One camp found it charming, almost endearing, the robot equivalent of a new employee knocking over a display on their first day. The other camp went straight to the existential: jobs, surveillance, the inevitable robot uprising. Someone made a Terminator reference. Of course they did.

But the comment that stuck with me was simpler than any of that. Someone just asked: why couldn’t this just be a vending machine?

It’s a fair question. And the honest answer is that for most of what these retail robots are currently doing, a vending machine is the better tool. More reliable, cheaper, already solved. The counter-argument someone offered, about fewer moving parts and software updates instead of physical rebuilds, was reasonable in theory. They then immediately admitted none of it applied to the current state of the technology. At least they were honest about it.

This is the gap that doesn’t get talked about enough in the breathless coverage of humanoid robots. We’ve somehow convinced ourselves that because large language models got very good very fast, physical robotics must be on the same curve. It isn’t. One commenter put it well: the difference between a model writing competent prose and a robot reliably picking up a spilled object is enormous. Boston Dynamics has been working on locomotion for decades. Fine motor control in unstructured environments, a cluttered retail floor, a kid wandering into the frame, is genuinely hard in ways that generating text is not.

There’s a neurological reason for that. Someone in the thread pointed out that the human brain devotes significantly more processing to physical movement than to conscious thought. All that hand-eye coordination, the pressure feedback from your fingertips, the way your cerebellum silently recalibrates every time you miss a catch so you don’t miss the next one: none of that feels like work because evolution spent a very long time making it feel effortless. We throw a basketball into a hoop from five metres and don’t think twice about it. A robot doing the same thing reliably, in varying lighting, on a slightly uneven floor, with a child nearby, remains a genuinely difficult engineering problem.

The thing is, I find this interesting rather than reassuring. The fact that physical manipulation is hard doesn’t mean it stays hard forever. There’s real research happening on treating robot manipulation as a generative problem rather than hard-coded motion planning, and the pace of progress isn’t nothing. I don’t know when it tips. Nobody does, and I’d be suspicious of anyone who claims otherwise.

What I do know is that the jobs conversation is more complicated than either “robots are useless, relax” or “we’re all done.” The warehouse automation picture is already further along than most people realise. Dark warehouses moving pharmaceutical pallets, sorting robots processing hundreds of thousands of packages with barely a human in sight: that’s not science fiction, it’s current operations. The retail humanoid fumbling a shelf retrieval is the visible, viral, slightly comic end of a spectrum that has a much less funny other end.

One comment that sat with me: what happened to the elderly gentleman who used to work at this store? It was meant lightly, I think. It didn’t land lightly.

The honest position is that I hold two views simultaneously and I can’t fully reconcile them. The technology is genuinely fascinating. The societal management of its consequences is, so far, not keeping pace. Both things are true. The robot recovering from its stumble and still completing the task is impressive. The question of what we actually do for the people whose livelihoods get rationalised away in the process remains largely unanswered, certainly at the policy level here and everywhere else.

The kid in the video thought it was brilliant, by the way. Maybe that’s the right response for now. Enjoy the stumble. Just don’t stop asking the harder questions while you’re laughing.