October 12, 2026
At the World Humanoid Robot Games 2026 in Beijing, a robot called Tiangong Ultra ran 100 metres in 8.64 seconds, faster than Usain Bolt's world record, according to Robohub's coverage of the event. Robots also boxed and played football at the same event. It's a great headline. It's also, for anyone evaluating robotics as a business investment, close to irrelevant.
Sprinting is a controlled, repetitive, single-axis task on a flat, known surface. Laundry -- the example Robohub's piece specifically points to -- requires identifying a garment's type and condition, handling a deformable object that changes shape every time you touch it, and adapting grip and folding strategy on the fly. That's a completely different engineering problem, closer to the frontier of robotic manipulation than locomotion. My take: the gap between 'can sprint' and 'can do laundry' is the single most useful lens a business reader has for evaluating any humanoid robot demo they see this year. If the showcase is athletic or choreographed, ask what happens when the object on the table isn't exactly where the robot expects it to be.
This matters because the commercial pitch for humanoids is overwhelmingly about unstructured, variable work -- warehouse picking, elder care, hospitality, home assistance -- not stadium events. A robot that can box but can't reliably fold a towel isn't close to deployable in any of those settings. Investors and operators should treat flashy competition footage as marketing, not a product roadmap milestone. For companies thinking about where actual automation ROI shows up today, it's worth looking at workflow automation built around software and process, not physical dexterity -- that's where the reliable gains are being made right now, while humanoid manipulation catches up.
Three other stories this week, none of them humanoid, are arguably more consequential for where humanoid robots eventually get their capabilities from. New Atlas reported on a worm-inspired robot from NYU's Tandon School of Engineering designed to pump water across rugged coastal terrain by shifting its weight, and on a flea-inspired hopping robot called DirectHop from the University of Washington built for small, efficient jumps. Separately, New Atlas covered a fiber-thin flexible motor from EPFL in Switzerland designed to actuate soft robots without a rigid motor housing.
None of these are humanoid robots, and I won't pretend otherwise. But each is solving a piece of the puzzle that humanoid platforms genuinely lack: soft, compliant actuation (the EPFL fiber motor), efficient locomotion over irregular terrain (the worm robot), and energy-efficient dynamic movement (the hopping robot). Humanoid robotics has largely been a story of rigid actuators and rigid torsos chasing human-like movement with brute force. Biomimetic research like this suggests the better long-term path might be borrowing actuation strategies from soft-bodied and small-bodied animals rather than scaling up rigid humanoid joints. If that bet pays off, the humanoid robots of five years from now may move and grip more like the flea or the worm under the hood than like the sprinting robot in Beijing.
For a business reader, the practical takeaway is patience paired with skepticism. The underlying science -- soft actuators, efficient locomotion, better manipulation -- is advancing steadily and genuinely. The showcase events are advancing much faster than the underlying capability, because a sprint is easy to stage and a towel-fold is not. If you're evaluating a vendor pitch, ask to see the boring task, not the highlight reel. We've made the same point before about why today's robotics headlines keep missing the humanoid story entirely -- the pattern holds again this week.
If your business is weighing a humanoid robot pilot in the next year or two, what's the one mundane task you'd actually want it to prove it can do before you'd trust it near your operation?
Sources