Video of the Week: Unitree Breaks Records

https://www.youtube.com/watch?v=zoMDadPQLKA

Right, so I've watched this video eleven times now, and each time I tell myself it'll be less impressive than the last.

It is not less impressive than the last.

It is, if anything, more impressive, which is deeply annoying because I was already embarrassed about how much time I've spent staring at a 30 second video of robot running (without a head).

What is this showing us? Well, Unitree's H1 just hit 10 metres per second. That's 22.4 miles per hour. That is - and I want you to just roll this round in your head for a moment - the speed you need to run a 10-second 100 metres. Which is definitely Usain Bolt territory. Not "pretty quick for a machine" territory. Not "impressive given the engineering constraints" territory. Usain. Bolt. Territory.

And the video doesn't even try to make it dramatic. There's no Hans Zimmer score. No slow-motion replay with sparkling overlays. No voiceover explaining why you should be amazed. It's just ... a robot, sprinting (without a head) on what appears to be a slightly unremarkable running track with some posters in the background that I've +definitely+ not tried to slow down and read.

The matter-of-fact presentation is actually doing a lot of the work here. This isn't a hype reel. It's closer to a test log that someone decided to post publicly, which somehow makes it more unsettling. Like: we've just got to the point where this is normal enough to not require fanfare.

The H1 moves like a human. Not "like a human if you tilt your head to one side, squint and accept that you've been watching robot videos for too long" - I mean genuinely humanoid gait, arms pumping, weight shifting, the whole thing (without a head). The running motion is fluid in a way that makes your brain briefly misfile it under "person" before correcting itself. That cognitive blip is where all eleven of my rewatches are spent. I keep waiting for it to feel mechanical. It doesn't.

Now, a sensible caveat: this is controlled conditions, a flat track, probably optimal everything. It's not navigating a race of other (headless) robots or doing any of the contextually complex stuff that makes real-world deployment hard. Speed on a straight line is one metric among many. I'm aware of this. I'm saying it to prove I'm paying attention.

But here's what I think actually matters: we've crossed a line. Not the speed record line, though that's fun. The line I mean is the one where humanoid robots stopped being novelty and became reference points. When the benchmark is Usain Bolt - not "fastest robot ever made" but fastest human - the frame of comparison has shifted. We're not asking "can robots move?" anymore. We're asking "how do robots compare to us?"

That's new. That's genuinely new.

(But I do wish it had a head).