Humanoid Robots Are Entering Public Spaces. The Haidilao Incident Explains Why That's Complicated.

A humanoid robot at the Haidilao hot pot restaurant in Cupertino — Apple's literal backyard — was asked to dance. It did, briefly, then smashed plates and scattered chopsticks across diners' tables. Three employees had to physically restrain it; one wrestled it with a neck strap while another, apparently unbothered, checked her phone. Haidilao's official statement was that the robot was "not malfunctioning." That response, more than the incident itself, became the story.

On X, the Cupertino moment spread quickly and split into two distinct readings. @SadCreatorTalks framed it as a safety incident — employees put their bodies between a malfunctioning machine and customers — while @gothburz treated it as dark comedy, the future of dining arriving in the form of a robot restrained by a neck strap. Both posts pulled engagement, but neither framing was really about Haidilao. They were about whether humanoid robots belong in public spaces at all, and right now, the skeptics have the better material. The incident sat alongside a genuinely impressive clip that @sciencegirl shared of a humanoid robot performing surface cleaning with fluid, human-like balance — the kind of footage that makes the Cupertino disaster look like a calibration problem, not a category error. Both things are true: the hardware is advancing fast, and it is also currently scattering dishware in California restaurants.

The deeper tension lives in what @tphuang laid out plainly in a post that got less attention than the hot pot chaos but deserves more: home-service robots with wheeled bases work reasonably well in single-floor apartments and will hit real walls — literal ones — in multi-floor homes. It's a practical observation that cuts through both the hype and the panic. The question isn't whether humanoid robots can look impressive in a demo; it's whether the deployment constraints have been thought through before the robots go public. McDonald's testing humanoid robots in a Shanghai pilot and Beijing hosting the first World Humanoid Robot Games with 500 bipedal competitors from 16 countries suggest the industry has decided the answer is yes — but the Cupertino incident is what happens when the answer turns out to be premature.

The most pointed skepticism isn't coming from technophobes. A German-language post on Bluesky cited an actual Science paper arguing that the real bottleneck for humanoid robots isn't AI at all — it's hands. Dexterous manipulation and haptic feedback remain unsolved in ways that no amount of large language model progress addresses. Tesla's Optimus gets the headlines; the gripper problem gets the journal articles. Elon Musk announcing a breakthrough in "robot-capable artificial intelligence" and reviving the Dojo supercomputer project feeds a financial press that is constitutionally incapable of distinguishing between a capability demonstration and a deployed product. The news coverage this week was almost uniformly optimistic about Tesla's robotics trajectory. Bluesky was not.

What AheadForm's Origin F1 — a bionic humanoid with what @Yuhang__Hu called "new skins, new souls" — represents is a different bet than Tesla's: that the aesthetic and customization layer of humanoid robots matters as much as the locomotion stack. It's a reasonable bet if you think the near-term market is companionship and service roles rather than factory floors. But the Haidilao incident is a reminder that deploying robots in spaces where humans are eating and socializing is a different class of problem than deploying them in warehouses, and the industry's current posture — ship it, respond to incidents with statements about non-malfunctions — is going to produce more neck-strap moments before it produces fewer.