For years, the artificial intelligence industry focused almost all its attention on machine brains, language models, reasoning, image generation, autonomous agents, but there is one problem that remains extremely difficult to solve.

Getting a machine to physically interact with the world like a human being does, because writing text is one thing, breaking an egg without crushing it is something completely different and now a startup called Genesis AI believes it has found a solution for this. The company revealed Genie 26.5, a new artificial intelligence model focused on human-level physical manipulation.

But the most interesting thing may not be the software itself, but in the way they decided to teach the robots. Genie 26.5 is betting on what many specialists call the holy grail of robotics, perfectly uniting the brain and hands. To achieve this, the company created a highly dexterous human-scale robotic hand and a system capable of capturing human movements with extreme precision.

The released videos show tasks that until recently were considered virtually impossible for autonomous robots, including preparing complex meals using both hands simultaneously, organizing strings of delicate wires, manipulating liquids in laboratory experiments and even playing piano at a speed comparable to real human movements.

And the most important detail is perhaps the small movements, a large part of current robots still fail in simple tasks because the physical world is full of unpredictable micro variations, an object slips, changes position, offers different resistance and in robots it blocks, humans correct that almost instantly without realizing it, machines do not yet, that is why Genie 26.5 decided to attack exactly that bottleneck.

The company developed a glove equipped with touch sensors capable of creating an almost direct mapping between the human hand and the robotic hand, in short, a human operator executes the task normally while the system records movements, pressure, microcorrections and physical behavior with an absurdly precise level of detail, but there is another even more important layer behind all that, in addition to physical tests, Genesis AI is also creating hyper-realistic simulated environments where artificial intelligences train other artificial intelligences.

The goal is to reduce the so-called simulation-reality gap, one of the biggest problems in modern robotics, because a robot can seem perfect inside a simulator until it touches the real world for the first time. And now creating a single advanced robot no longer seems enough. The goal is to make machines learn at scale.