The second edition of the games, held in Beijing, put humanoid robots through sprinting, high jump, kickboxing, soccer, and a set of far less glamorous household and workplace tasks, laundry, hotel room service, library shelving, and firefighting.
The Highlight Reel
Robots beat the human 100-meter and standing high jump records, a real leap in whole-body control since last year's event
Some sprinting robots couldn't stop after crossing the finish line, crashing into barriers, breaking apart, or catching fire
Only 3 of 12 teams completed a full firefighting and hazard challenge requiring valve shutoffs and fire suppression
Many competitors were still human-teleoperated, and even the autonomous ones often leaned on cloud computing rather than onboard processing
The Records Are the Least Interesting Part
You can simulate something like 100,000 scenarios for hammering a nail, but that's just one task from millions of tasks in the world.
Kyle's read matched that skepticism almost exactly: the sprint records make for great viral clips, but a robot engineered to run in a straight line and nothing else isn't a general-purpose machine, it's a single trick built to win one event. The laundry-folding, shelving, and firefighting events, unglamorous and often unfinished, are the ones that actually measure whether these machines can do useful, unscripted work.
The Real Takeaway Isn't About Robots
Kyle tied the whole episode together with a broader theme worth sitting with beyond the robotics story itself: whether it's Carvana's sales data, Nissan and Honda sharing R&D, or a robot learning to sprint, the right move is treating a competitor or a faster-moving trend as something to study and draft off of, not something to resent. The instinct to learn from what's already one or two steps ahead is the actual skill on display across every story today, robots included.
Want the full conversation?
🎙️ Listen to today's Automotive State of the Union episode for the complete discussion.
A robot that can win a 100-meter dash and a robot that can safely stop running are not yet the same robot.
Investors have already poured more than $6 billion into humanoid robotics in 2025 alone, betting heavily on a future these games only partially previewed. The sprint times are the headline. Whether a robot can reliably do the boring stuff, folding a towel, shelving a book, shutting off a valve, without a human quietly steering it from the sidelines is still the actual scoreboard that matters.
