Genesis AI, a startup that has raised $105 million to build basic AI for robots, has unveiled its first prototype, the GENE-26.5, and it comes with surprising hands. In a demo video, the company demonstrated several advanced tasks performed by a group of robotic hands that it designed in-house.
“Model has always been the goal, because better model means better intelligence,” Zhou Qian, co-founder and CEO of Genesis, told TechCrunch. But the company soon realized it needed to control the devices. “So we decided to go full steam ahead,” he said.
Other well-funded companies work at the intersection between AI and robotics, such as Physical Intelligence and Skild AI. Chu also admitted that “there are probably 50 to 100 robotic hand companies.” But he and his co-founder Theophile Gervais hope that building their own factory will give them the upper hand.
The main difference is that Genesis’ hand is the same size and shape as a human hand – rather than the two-finger grips used by many robotics companies – reducing the gap with real-world conditions.
“This allows us to 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 artificial intelligence company Mistral who is now president of Genesis.
Of all the physical manipulation tasks shown in the video below, cooking is Gervet’s favorite, because it proves that the robot was able to complete a long series of difficult tasks, such as cracking an egg and slicing a tomato. But Genesis also tasked its robots with making smoothies, playing the piano, and solving a Rubik’s Cube – a robotic gimmick.
Other tasks, such as laboratory work, are more akin to what could be commercial applications of Genesis’ technology. But what’s happening behind the scenes is just as important: The startup has also developed a glove equipped with a sensor that acts as a real-life double of its robotic hand, collecting data that can be used more easily.
“Our idea was that if we could design a robotic hand that tried to mimic the human hand as closely as possible, we could unlock huge amounts of human data instantly without having to worry about what people call the ‘embodiment gap’ in robotics research,” Zhou said.
Others have experienced this problem. The main novelty is how Genesis combines this with its model. The current version was named GENE-26.5 in May 2026, but Zhu expects there will be many iterations, thanks to the simulation he developed. “The real bottleneck to model iteration speed is evaluation. This helps us speed up model training a lot,” he said.
Beyond simulation, the data will be key to training models that can help robots perform more tasks. This is also where the Genesis Glove can come in handy. Unlike older data collection devices that get in the way, they are as light and easy to wear as the security gloves already used in many industries, while being relatively cheap to manufacture, Gervitt said.
“We’re in conversations with a lot of customers right now, and the big value of the glove will be that, for the first time, you can wear a data collection device when you’re doing your day-to-day job, whether it’s a pharmaceutical lab technician or for manufacturing,” Gervitt said. This can also be supplemented with “egocentric video data” – where people film themselves performing the task.
However, it remains to be seen whether workers will be happy to wear the same gloves and cameras that can train the robots to replace them, and whether they will receive additional pay for this training. Gervait suggested that this would be between Genesis customers and their employees. He added: “We have not worked out the details yet.”
Either way, they may decide not to share that data with the startup, the founders acknowledged. But the startup also has its own ways to build its own “human skills library” – and can even pay external partners to collect data. Its model has already been trained on “vast amounts of human video on the Internet,” according to a press release that did not mention compensation.
Combined with its own simulation system, this could help Genesis lower the costs of its technology for real-world applications like the one it demonstrated. “This represents an important milestone for their team and the broader robotics industry,” said Eric Schmidt, former Google CEO, who invested in the startup.
In July 2025, just a few months after its inception, the startup emerged from obscurity with a $105 million seed round co-led by Eclipse and Khosla Ventures, with additional backers including Bpifrance, HSG and individuals like Schmidt, but also Xavier Niel, Daniela Rus and Vladlen Koltun.
This funding helped Genesis increase its headcount. With offices in Paris and California, it has also expanded to London. “One of the main reasons we decided to be in Europe is because there is a huge density of talent across the entire continent,” Gervit said. Its 60-person team is split “40-45% in Europe and 50-55% in the US,” and the startup is currently hiring across all three locations.
Aside from hiring, the company also plans to unveil its first general-purpose robot soon, which Chu told TechCrunch will be a robot that covers the entire body, not just the hands. But he insisted that the road map remains the same.
“Our goal is to build the most capable automated system,” he said.
When you buy through links in our articles, we may earn a small commission. This does not affect our editorial independence.
