A Chinese humanoid beat Bolt's 100m time. The start is where Bolt still wins

Tiangong Ultra, a humanoid built in China, covered 100 meters in 8.64 seconds. That is faster than Usain Bolt's 9.58-second world record, set in Berlin in 2009, and gives the robot an average sprint speed of almost 42 kph. The run is fresh, and the comparison is where it starts to come apart.

Iain Hunter, a Brigham Young University exercise scientist who also serves as a USA Track & Field sport scientist, estimates that at the five-meter mark of that record run the eight-time Olympic champion would have been nearly two seconds ahead. Tiangong started from an upright posture with no blocks, and in its first run waited roughly a full second after the starting signal before moving, apparently failing to register it. Bolt's reaction time in 2009 was 0.146 seconds from blocks.

Tiangong's top speed is the easier half of the comparison to engineer. Racing the gun, by contrast, is a perception-and-decision problem the hardware alone will not fix. The team's response to the speed half has been a round of mechanical redesigns.

Guo Yijie, the humanoid's innovation department manager, said the redesigns covered a lighter torso and waist, upgraded motors, an expanded range of movement, and revised control software. Reuters, which reviewed the race video, counted more than 50 steps from Tiangong over the 100-meter distance; Bolt needed 41 to set his record. The robot does not run like a sprinter. It runs like a sewing machine, a high cadence of short, mechanically efficient strides rather than the long, elastic, ground-covering ones that define elite human sprinting.

Chungnam National University sport scientist Jung Moon-hyun read the run as a balance-and-sprint milestone, one where the robot held its line over the full distance without falling. Kwangwoon University robotics researcher Park Suhan went further, framing the gait as a direct reflection of motor capability rather than a Bolt-style running technique. A human sprinter's stride is shaped by years of training and a body the robot does not have. A humanoid's stride is whatever its engineers and motors can deliver, beat by beat, on schedule.

The framing also explains why the headline number flatters the robot. A 100-meter sprint is a short race, and the clock is a forgiving judge for a machine that does not tire. Tiangong covered the distance in a flat sprint, never left its lane, and never had to bend. The five-meter gap, the start, and the perception lag are not problems the next iteration of motors will fix on their own. They are problems about how the robot sees the world and decides what to do.

Hunter's estimate, that Bolt would have been about two seconds ahead at five meters, is the cleanest way to size what is still missing. On the current run, the robot needed most of the next 95 meters just to overtake the human.

Guo Yijie has framed the next round of work around what Tiangong still cannot do: faster decision making, sharper perception, and a cleaner read of the start signal. The team has not published a date for the next benchmark, but the next test is plain. The clock at 100 meters belongs to the robot. The start still belongs to Bolt.