Khosla-backed robotics startup Genesis AI unveils GENE-26.5, its first model, which can control robotic hands that it designed in-house to do tasks like cooking
TechCrunchAnna Heim
Context & Ripple Effects
Genesis AI emerged from stealth with a $105 million seed round to build foundational AI for robots, using synthetic data as part of its training approach. Its stated ambition was a model that could work across robot types.
GENE-26.5 turns that earlier platform ambition into a first disclosed model-and-hardware demonstration: Genesis AI is pairing its software with robotic hands it designed itself. Related coverage also shows Generalist pursuing models for high-dexterity physical tasks, making manipulation a visible competitive frontier.
First-order effects
Genesis AI now has a concrete first model to demonstrate against its prior foundational-robotics thesis, with control of its own robotic hands as the immediate proof point.
The company gains tighter control over the interface between model behavior and hand hardware, rather than presenting an AI system only as an abstract cross-robot platform.
Second-order effects
Other robotics-model startups working on dexterous tasks, including Generalist, face a clearer benchmark: not merely task completion, but integration of a model with hardware capable of performing it.
Genesis AI's synthetic-data effort becomes more consequential if it can improve performance on its own hands, linking its training-data strategy directly to an embodied product demonstration.
Third-order effects
If model developers increasingly build or co-design the manipulators on which they demonstrate, robotics AI may compete around integrated hardware-software stacks as well as broadly compatible foundation models.
The emerging contest is shifting toward general-purpose dexterity: companies will need to show that task demos translate across hands, tasks, and robot forms before claims of foundational capability are established.
The trend: Robotics AI startups are moving from foundational-model promises toward demonstrations of high-dexterity manipulation on tightly coupled hardware-software systems.
This a remarkable demo, and there is a moment in it that hit me, emotionally, in a way that reminds me of how @polynoamial has talked about “feeling the AGI” when you see a model exhibit skill in something where you take pride in your own ability. Do watch with sound on.
Robotics is a systems problem. Every layer matters, and every detail must integrate across the full stack. So bringing real scale to robotics requires building the full stack from the ground up: hardware, data, model, and simulation. Over the last year, we built a global team
Robots cracking an egg, slicing tomatoes, cooking an actual omlette, ... pretty unbelievable for 2026! The rate of progress in both hardware and software is awesome.
So lucky to live in this era, working on the most exciting problem with the best people at the best place: Genesis. It feels amazing that the ambitious plans can be achieved by having conviction, working the full stack from the lowest level up, and carefully getting every
4/ Rubik's Cube solving has been a long-standing challenging benchmark for robotic manipulation. The task requires fine-grained control under the geometric and kinematic constraints imposed by the cube itself. Prior state-of-the-art is still the single-handed solver from OpenAI' …
Incredibly impressive control stack! But the ambiguous wording around the hardware is unnecessary. If it's a vendor's gear, credit them or don't mention it at all. Better to just be transparent—even if you don't say it, anyone in the industry can see it.
Excited to share this step toward human-level robotic manipulation. Solving this takes more than better AI models — it's a full-stack system problem. Proud to have been part of the journey this past year. More to come.
The past year has been the happiest period of my life. Genesis is a place of miracles. We set countless ambitious goals. None of them seemed possible with existing technology. One by one, we made them real. The density of breakthroughs coming from such a small team, across so [vi…
For well over a decade I've worked on the cutting edge of robotics. I've seen every team that's claimed a breakthrough in generalized manipulation over the past few years. What @gs_ai_ just put out is like nothing I've ever seen. Genesis approached this problem from first