Born from the imagination of a group of science fiction and robotics enthusiasts, the Asimov project applies the open source model to the manufacture of humanoid robots. He even offers a “DIY” kit to build his robot.
Building your own humanoid robot is now possible. With the Asimov project, the start-up Menlo Research, based in Singapore, is developing a robot based on an open source model. All of the part plans are published on the GitHub platform, with development being done in collaboration with a community of enthusiasts. And a “DIY” kit has just been revealed.
Robotics DIY version
Unlike players like Tesla, Unitree or Figure AI, who develop robots in a classic industrial logic, Menlo Research is betting on an open approach. The start-up specializing in research and development in AI is notably at the origin of the open source agent Jendot.AI. With Asimov, she applies open source logic to hardware.
The project was born from the minds of science fiction and robotics enthusiasts. Star Wars robots were a major source of inspiration, particularly the scene where Anakin builds C-3PO in his garage. “This feeling of curiosity, the fact of tinkering with something and building with your own hands, sums up the spirit of Asimov well,” explains an Asimov spokesperson to Journal du Net (the team wanted to express themselves collectively, editor’s note). Asimov started as an open source humanoid project. We wanted to create something that people could build, understand and take apart. The right to repair, the right to modify, this is at the heart of our vision of robotics.” Quickly, the choice fell on a robot with a humanoid appearance, capable of evolving in environments designed for humans. The owners of the robots will be free to configure them according to their needs or desires.
From the first stages of the project, the community was invited to collaborate via a forum and a Discord group which now brings together more than 2,000 members, mainly engineers and developers. And every two weeks, an event called “Community Friday” is held to review designs and allow community members to provide feedback. These can range from minor details, like reporting a problem with a hip joint or suggesting improvements to how parts fit together, to larger contributions, like suggesting entirely new approaches. “Ideas come from different countries, backgrounds and levels of expertise,” explains Asimov’s team. “This diversity is essential, because robotics is not a single-discipline problem.”
This collaboration gave birth to a DIY kit called “Here Be Dragons Edition”, priced at $15,000. It will be delivered in the coming months. The version of the robot thus proposed will measure 1m20 and weigh 35 kilos, and will have 25+2 degrees of mobility, characteristics close to those of Unitree’s G1.
The goal is to get robots into the hands of developers, researchers, and builders so they can start experimenting and contributing. “We are less interested in selling a single robot than in creating conditions for a whole generation to build with them,” explains Asimov’s spokesperson.
An accessible robot
Still in the spirit of democratizing humanoid robots, the Asimov project aims to reduce the total cost of operating a humanoid robot to less than $30,000 per year, including maintenance and software. The team is betting that this cost reduction will come through the development of an ecosystem with open supply chains, multiple manufacturers and standardized designs.
Another priority is enabling parts to be replaced quickly and cheaply, which is currently a major problem for humanoid robotics. “Many systems are designed for ideal conditions,” explains the Asimov spokesperson. “We try to do the opposite: design in anticipation of failures and degradation. A robot that only works when everything is perfect is not a robot that can really be deployed.” The plans for each part are published on the open source GitHub platform, which allows them to be manufactured with a 3D printer.
The software layer is at the heart of the project, because it is the only persistent element. A major challenge is the transition from simulation to reality, often problematic in the development of humanoid robots. To answer this, the team designs the firmware and control algorithms as a unified system from the start, integrating the imperfections of reality. “It’s tempting to think that photorealistic simulations are enough, but reality is messy in a way that’s difficult to model,” the spokesperson acknowledges. Dust on a camera, electrical interference, worn parts after months of use: so many micro-failures that no simulation perfectly captures.
Beyond the technical and economic aspects, Asimov intends to contribute to a broader reflection on humanoid robotics and its place in our societies. “Many companies focus on functional demonstrations and capabilities, which is important, but that’s only part of the story,” said the spokesperson. “When you create something that evolves in the world in a form similar to ours, it naturally raises questions: what does it mean to build in our image, what responsibilities does that entail and what kind of relationship we want to have with these systems.”
