Genesis AI and its robot Eno: How the full stack approach is revolutionizing intelligent robotics
📋 Summary
For a few years now, I have noticed that robotics is undergoing a major transformation. It is no longer just a matter of sophisticated mechanics, but of artificial intelligence capable of generalizing and learning. Genesis AI, the Franco-American startup founded by Théophile Gervet, perfectly embodies this revolution. Their humanoid robot Eno does not seek to copy human appearance, but rather to reproduce the manipulation and adaptation capabilities that make us strong. What particularly interests me is their strategy: instead of developing only an AI model or only hardware, they have chosen a full stack approach integrating software, artificial intelligence, and hardware. This decision reveals a deep understanding of the market. Customers are not looking for isolated technological bricks, but complete and operational solutions. Eno thus represents much more than just a robot: it is the demonstration that true innovation in AI requires the total integration of all elements. In this article, I propose to explore how Genesis AI has built this revolutionary approach, what technical challenges they have overcome, and why this strategy could redefine the future of industrial automation in Europe and beyond.
Manipulation, the Real Challenge of Intelligent Robotics
Manipulation remains the most complex problem in robotics. For a long time, industrial robots were limited to repetitive and predictable tasks. But creating a robot capable of interacting with its environment flexibly and intelligently is a whole different story. This is where generative AI 🤖 comes into play. Théophile Gervet explains that to build a truly generalist model, it must be trained on the largest and most diverse data source possible. Language models learn from the Internet. For manipulation, the equivalent of the Internet is human data: hundreds of millions of hours of manipulation performed by humans in different environments. That is why Genesis AI made the crucial choice to equip Eno with arms and hands similar to those of a human. If we had six fingers, their robots would have six as well. This natural correspondence is not cosmetic; it is functional.
I found this approach particularly intelligent. Many companies seek to create “perfect” robots that resemble nothing known. Genesis AI understood that technical perfection was not the goal: it was the ability to learn from existing human data. By using a human morphology, they drastically reduce the problem of generalization. The robot can directly leverage the millions of hours of human manipulation videos available. This is a brilliant learning strategy that significantly accelerates development.
Dexterity is therefore not an end in itself. A three-fingered hand could probably achieve comparable results. What really matters is the ability to directly exploit human manipulation data. This philosophy reveals a deep understanding of what it means to build a practical and scalable artificial intelligence.
The Hybrid Approach: Combining VLM and World Models
Genesis AI did not choose between two dominant paradigms in robotics: VLA/VLM on one side and World Models on the other. They did something more ambitious. Their hybrid approach combines the best of both worlds 🧠. In robotics, they needed a multimodal model capable of integrating video, natural language, proprioception, and tactile information. Their solution combines the understanding capabilities of VLM with a physics representation inspired by World Models. It is a sophisticated architecture that reflects a nuanced understanding of technical challenges.
This technical choice impresses me because it shows a rare maturity in the industry. Many startups rush to the latest technological trend. Genesis AI took the time to think about what would actually work for their use case. Excellent VLMs understand visual and linguistic context, but they may lack physical precision. World Models, on the other hand, model physics but can be limited in contextual understanding. By combining them, Genesis AI creates a more robust and generalist system.
This hybrid architecture is not just a matter of technical performance. It is a strategic decision that reflects how AI is evolving. We are moving from an era where a single approach dominated to an era where the best solutions combine multiple paradigms. This is particularly true for industrial applications where reliability and versatility are critical.
Simulation, the Invisible Accelerator of Innovation
Many people do not realize the importance of simulation in the development of intelligent robots. Genesis AI has developed a physical simulator that reproduces the interactions of the robot with its environment. This allows them to evaluate hundreds of thousands of scenarios in parallel in a few hours, where these experiments would take months in the real world 💻. It is an extraordinary productivity multiplier.
I see simulation as one of the most underestimated innovations in AI. While everyone talks about language models and neural architectures, simulation does the silent work of making large-scale innovation possible. Without simulation, Genesis AI would have to build hundreds of physical robots to test each variation. With simulation, they can explore the space of possibilities exponentially faster. It is an acceleration of the innovation cycle that changes everything.
This approach also reveals a deep understanding of what it means to build a sustainable technology company. Simulation is not just a technical tool; it is an economic multiplier. It reduces development costs, accelerates time-to-market, and allows for the exploration of more hypotheses. That is why Genesis AI has invested heavily in this infrastructure from the start.
Full Stack: Why Build Everything Yourself
Genesis AI started by developing an AI model, then robotic hands, before presenting a complete robot. Why ultimately choose a full stack approach? The answer is simple yet profound: their clients are looking for a fully integrated solution 🔧. A model alone does not solve their problem. There is no hardware platform capable of meeting their specific needs. Therefore, they had to build their own robot.
This decision reveals an important truth about the AI and robotics market. Successful companies are not those that excel in a single area, but those that can integrate multiple disciplines. Genesis AI masters AI, robotics, mechanical engineering, and simulation. It is this integration that creates real value for clients. You can check out our guide on AI and data science tools to better understand how these technologies come together.
I believe this is an important lesson for the entire AI industry. Too often, we see companies building brilliant models but failing to deploy them in the real world. Genesis AI understood that true innovation lies in the ability to transform a technological advancement into an operational product. That is why their full stack approach is not a limitation; it is a major competitive advantage.
From Data Collection to Generalization
Genesis AI does not just use existing data. They also produce their own data through an innovative data collection glove that records finger positions and tactile information when a human performs a task 📊. The robot is then equipped with the same sensors, facilitating data transfer. They deploy these gloves with industrial partners to massively enrich their database.
This is a brilliant approach to data collection. Instead of relying solely on public data or existing videos, Genesis AI creates a system that captures data exactly as it will be used by the robot. This match between the data source and the final application significantly reduces domain transfer issues. It is a data strategy that shows a deep understanding of what makes AI practical. To delve deeper into this topic, check out our article on data management tools.
The goal is ambitious: to scale human manipulation data to a whole new level. Several tens of millions of hours of data is the equivalent of the Internet for manipulation. With this amount of data, the model can learn to generalize to completely new use cases. That is why Genesis AI is investing heavily in this data collection infrastructure. It is the future of their competitiveness.
The First Clients and the Deployment Strategy
Genesis AI is currently focusing on industrial uses in the broad sense: factories, pharmaceutical laboratories, data centers. Why this strategy? Because the value created is greater there, and the environments are more structured than in services or at home 🏭. They have announced a first partnership with LG in Korea. What particularly interests them about LG is the diversity of tasks performed by their employees. This allows them to exercise the model’s generalization capability across many use cases, simply within the same company.
I see this strategy as very intelligent. Instead of trying to sell thousands of robots to different clients, Genesis AI prefers to work deeply with a few key clients. This allows them to collect rich data, understand real problems, and build truly tailored solutions. It is a classic B2B approach but applied with particular sophistication. Intelligent automation is at the heart of this strategy.
The deployment goals are ambitious but realistic: a few dozen robots this year, a few hundred next year, and then probably several thousand starting in 2028. It is a growth curve that reflects the complexity of the product and the need to build a robust supply chain. Genesis AI is working with international partners for manufacturing, with a goal of local assembly in Europe for European clients and in the United States for American clients.
Conclusion
Genesis AI represents for me a model of innovation that should inspire the entire AI industry. They have understood that true revolution does not come from a single technological advancement, but from the ability to integrate multiple disciplines into a coherent solution. Their full stack approach, their hybrid architecture combining VLM and World Models, their massive investment in simulation and data collection: all of this converges towards a unique goal, to create robots that are truly useful in the real world. What impresses me the most is their strategic patience. They are not looking to conquer the consumer market tomorrow. They are building the foundations to dominate industrial uses today, and gradually expand tomorrow.
Europe has a rare opportunity. Genesis AI is Franco-American, based in Paris and San Francisco. They benefit from the pool of French talent in AI and robotics while having access to American investors and markets. This is a model that other European companies should study. Intelligent robotics will likely be the main lever for the reindustrialization of the continent. And Genesis AI shows that it is possible to build a world-class company in Europe, provided one has the ambition, the strategy, and the ability to integrate multiple technological domains. It is a message of hope for the French and European ecosystem.
📝 In Brief
- Genesis AI combines generative AI, robotics, and simulation in an integrated full stack approach
- Their robot Eno uses a human morphology to directly leverage human manipulation data
- The hybrid architecture combines VLM and World Models for better generalization and physical understanding
- The physical simulation accelerates development by allowing the evaluation of hundreds of thousands of scenarios in hours
- The deployment strategy initially targets structured industrial uses before gradually expanding
