Everyone reposted the humanoid that ran the 100m in 9.32 seconds.
Nobody reposted the two robots that piled up on the very next bend and got put out with fire extinguishers. Same games, same week, Beijing.
That's the whole industry in one clip. Every demo sells you the peak - the single run where the trick lands. What none of them show is the failure rate: how often that same robot eats the floor the next four tries.
A humanoid that nails it once and falls four times isn't a product. It's a showreel with good editing.
So here's the spec nobody prints on the box: how many laps before it needs a human and an extinguisher?
Which humanoid out right now would you actually trust to run it twice in a row?
A ROBOT THAT COSTS LESS THAN A USED CAR JUST THREW A KUNG FU COMBO. AND AMERICA MADE SURE YOU CAN'T BUY IT.
Not a render. Not a Boston Dynamics teaser.
A Unitree G1 in an empty office. It drops into a horse stance, throws a full combo, slips a strike, spins, and never loses its feet. Knock it down and it picks itself back up. Sixteen thousand dollars. Off the shelf. In China.
The American home robot in the same weight class runs $20,000, hasn't shipped, and still needs a person in a VR headset to fold a shirt.
So on July 28 the FCC added foreign-made robots to its Covered List. New Chinese humanoids and robot dogs can't get the approval they need to be sold in the US anymore.
Read that back. The pitch used to be that American robots were better. Now the pitch is that the Chinese ones aren't allowed here.
The legs are solved. The hands aren't. And the fastest way to lose a race is to ban the guy who's winning it instead of catching him.
Wrote about this exact gap two weeks ago and here's the footage that makes it concrete.
Straight track, no obstacles, an entire engineering team watching live. The policy still couldn't stop the body in time and it broke at the waist on camera.
This is the difference between "the robot ran" and "the robot ran under these conditions, with this failure rate, and this is what it does when it's wrong." The clip only proves the first one. Nobody in the video can tell you the second from watching it happen once.
That's the actual skill gap right now. Not training a faster policy. Building the harness that catches this before it happens live, in front of the team that has to explain it.



