Genesis AI, a robotics startup backed by Khosla Ventures and Eclipse, has revealed its first AI model, GENE-26.5, alongside a demo of in-house robotic hands capable of performing complex physical tasks. The announcement was made in May 2026, roughly a year after the company emerged from stealth with a $105 million seed round.
The demo video shows robotic hands cracking eggs, slicing tomatoes, preparing smoothies, playing piano, and solving a Rubik’s cube. Cofounder and CEO Zhou Xian said the company originally focused solely on AI modeling but concluded it needed control over hardware to advance its goals. “So we decided to go full stack,” he told TechCrunch.
A key design choice sets Genesis apart from many competitors: its robotic hand matches the size and shape of a human hand, rather than using the two-finger grippers common in the industry. Cofounder and President Théophile Gervet, a former research scientist at Mistral AI, said this allows the company to collect significantly more training data. Genesis has also developed a sensor-equipped glove that mirrors the robotic hand, enabling data collection from workers in real-world settings such as pharmaceutical labs and manufacturing floors. Gervet noted the glove is lightweight and relatively inexpensive to produce.
The company is also in talks with customers about deploying the glove for on-the-job data collection, though Gervet acknowledged that questions about worker compensation and data-sharing arrangements remain unresolved. Genesis additionally trains its model on internet video footage of humans performing tasks and uses a proprietary simulation system to accelerate model development.
Genesis operates offices in Paris, California, and London, with a team of 60 people split roughly 40–45% in Europe and 50–55% in the U.S. Backers include Bpifrance, HSG, former Google CEO Eric Schmidt, Xavier Niel, and researchers Daniela Rus and Vladlen Koltun.
Xian told TechCrunch the company plans to unveil a full-body general-purpose robot in the near future, describing the ultimate objective as building “the most capable robotic system.”
Source: TechCrunch