机器人跑半马这事，我认真查了下数据，不是“标题党” (EN)

Last Thursday (April 19), a humanoid robot half-marathon took place in Beijing’s Yizhuang district. My social feeds exploded—some said “robots beat humans,” others called it “another overhyped PPT product.” My reaction: let’s hold off on conclusions and look at the actual numbers.

Here are the facts: over 70 teams competed on a roughly 21-kilometer closed course. I watched several videos from the event, and “Glory·Flash” from the Honor team did look impressive—the robot’s cornering and acceleration movements were smooth and visually quite close to human runners.

But here’s the problem: where does the “surpassed human records” claim actually come from?

I traced it back. The earliest source was a tech media outlet with a headline claiming “Glory·Flash Sets New Human Half-Marathon Record.” But reading further, they were comparing the robot’s best time against the women’s human half-marathon record—two completely different baselines. Basically, it was a clickbait comparison.

The actual numbers? According to official post-race data, the robot’s best performance was around 1 hour 40 minutes, while the men’s world record is 56 minutes. That’s still a significant gap.

That said, I’m not trying to say the robot competition was “fake.” Actually, I think its significance isn’t about “beating humans”—it’s about validating several key technical capabilities: long-duration battery stability, navigation across complex outdoor terrain, and multi-robot coordination systems. These are the areas embodied AI truly needs to crack.

One side note: I observed that several robots used a “teleoperation plus autonomy” hybrid mode—humans made some decisions in the background while robots handled execution. This might actually be the more pragmatic path forward.