Just posted to my Forbes column about one of the most important and least solved problems in robotics: robot hands.
For years, humanoid robots have looked impressive in demos while still struggling with the messy, delicate physical manipulation humans do effortlessly every day. That’s starting to change fast. Genesis AI just emerged from stealth with a demo that honestly feels like a major moment for robotics. Its new robotic hand cracks eggs, chops vegetables, mixes drinks, applies electrical tape to wires, separates stacked cups, and even solves a Rubik’s cube, all supposedly fully autonomous and running at normal speed.
But beneath the viral demo is a much bigger story about where robotics actually is today versus where investors and headlines sometimes think it is. A few weeks ago, I spoke with the founders of Kyber Labs, another company building advanced robot hands, and one quote from co-founder Tyler Habowski stuck with me: “There are literally zero robot hands deployed right now doing routine work.”
That gap between spectacular demos and real-world deployment may be the single most important challenge in humanoid robotics right now. Genesis AI is betting that the future belongs to robotics-native foundation models trained on massive datasets of human dexterity. Kyber Labs is taking a more pragmatic path: cheaper, force-sensitive hands deployed into narrow real-world workflows first, then iterating from actual usage.
The fascinating thing is that both approaches might be right. The destination is probably general-purpose robotic manipulation. But the path there may look a lot more like SpaceX-style iteration than perfectly polished demo videos.
One of my favorite lines from the piece comes from a pottery-class analogy: the students graded on quantity ended up making the best pots because they learned through repetition, failure and iteration. Robotics may work the same way.