趋势解读：Khosla-backed robotics startup Genesis AI has gone full-stack,，解读最新研究结论

Genesis AI , a startup that raised a $105 million seed round to build foundational AI for robotics, has unveiled its first model, GENE-26.5, and it comes with surprise hands. In a demo video, the company showcased various advanced tasks performed by a set of robotic hands it has designed in-house. “The model has always been the goal, because a better model means better intelligence,” Genesis cofounder and CEO Zhou Xian told TechCrunch. But the company soon realized that it needed control over the hardware. “So we decided to go full stack,” he said. Other well-funded companies operate at the intersection of AI and robotics — such as Physical Intelligence and Skild AI . Zhian also acknowledged that “there’s probably 50 or 100 robotic hand companies out there.” But he and his cofounder Théophile Gervet hope that building their own will give them the upper hand. The key difference is that Genesis’ hand has the same size and shape as a human hand — rather than the two-finger grippers many robotics companies have been using — reducing the gap with real-world conditions. “That lets us collect a lot more data than was previously possible, to train a model that can do many more tasks,” said Gervet, a former research scientist at Mistral AI who is now Genesis’ president. Of all the physical manipulation tasks showcased in the video below, Gervet’s personal favorite is cooking, because it proves that the robot has been able to complete a long series of difficult tasks, such as cracking an egg and slicing a tomato. But Genesis has also tasked its robots with preparing smoothies, playing the piano, and solving Rubik’s cube — a robotics gimmick . Other tasks, such as lab work, are closer to what could be the commercial applications of Genesis’ technology. But what happens behind the scenes is just as important: the startup has also developed a sensor-loaded glove that works as a real-life double of its robotic hand, collecting data that can more readily be used. “Our idea was that if we could design a robotic hand that tries to mimic a human hand as much as possible, we can instantly unlock huge amounts of human data without having to worry about what people call the ‘embodiment gap’ in robotics research,” Xian said. Others have tried their hand at that problem; the main novelty is how Genesis combines this with its model. The current version is named GENE-26.5 for May 2026, but Xian expects there will be many iterations, thanks to the simulation it has developed. “The real bottleneck for the iteration speed of the model is evaluation. So this helps us speed up model training a lot,” he said. Beyond simulation, though, data will be key to training models that can help robots perform more tasks. That’s also where Genesis’ glove could come in handy. Gervet said that, unlike clunky data collection devices that get in the way, it is just as light and easy to wear as the security gloves already used in many industries, while relatively cheap to make. “We’re in talks with a lot of customers right now, and a lot of the value of a glove would be that, for the first time, you can wear the data collection device when you’re doing your daily job, whether it’s a lab technician for pharma or for manufacturing,” Gervet said. This would also be complemented by ‘egocentric video data’ — people filming themselves doing the task. Still, it remains to be seen whether workers would be happy to wear the very gloves and cameras that could train robots to replace them, and whether they will get extra pay for that training. That will be between Genesis’ customers and their employees, Gervet suggested. “We haven’t nailed the details yet,” he said.