3. Menlo's Solution

Menlo is the full stack for the open robotics ecosystem: open hardware, a modern robot OS, and the economic infrastructure to turn skills into a real marketplace.

Open Robot Hardware: Asimov

Asimov is Menlo's open, licensable reference hardware, built to absorb the best ideas from the community and become a standard embodiment for the ecosystem rather than a closed, proprietary robot.

Right to repair, right to customize. 3D printed, CNC, or off the shelf parts, a detailed public assembly manual, and fanatical docs and simulation support so developers can fix and modify Asimov as fast as they iterate on software.
Built with manufacturing partners, not instead of them. Menlo works with a growing network of manufacturing and R&D partners worldwide, spreading supply chain risk and growing a global ecosystem instead of vertically integrating manufacturing in house.
A living roadmap. Asimov has moved from Asimov 0 (legs only, to learn locomotion) to Asimov 1 (the educational humanoid kit) to Asimov 2 (a deployment ready, bimanual robot), with each generation shaped directly by community feedback.

A Modern Robot OS: Asimov OS

Autonomy, in Menlo's model, is a collection of skills, not one monolithic model. Asimov OS is the software that runs and operates Menlo robots. It hosts skills, which are closed, tested units of control, and orchestrates them locally on the robot, so any coding agent can compose them into behavior.

This is deliberately different from robotics middleware like ROS: Asimov OS is the layer above it. Developers get typed, versioned calls like navigate_to(x, y) or grasp(object), never a raw topic to wire up themselves. High level mission reasoning can live in the cloud, with any coding agent; Asimov OS keeps the safety envelope, real time control, and skill runtime running locally on the robot, so a dropped connection never means a robot that falls over.

Enough skills is autonomy. Not an end to end foundation model that has to get everything right at once, but a growing, open catalog of capabilities where every new skill multiplies what every developer's agent can do with the robot.

Kepler: A Shared Understanding of Physics

Kepler is a world encoder that turns a robot's raw senses, video, audio, IMU, joint state, and touch, into a shared representation space: a "physics latent," or a universal primitive that developers compose skills from. It's what lets VLAs, world models, LLMs, and future model types all train against the same understanding of real world physics, instead of every skill developer reinventing their own.

Early experiments also suggest Kepler has emergent safety value: without any additional training, the frozen encoder's prediction error alone is enough to flag an invalid or corrupted robot state. Longer term, this is the layer Menlo intends to build a safety net for the whole ecosystem on.

Menlo Store: The Economic Infrastructure

The Menlo Store is what turns 100,000 developers experimenting into a real economy. Robot developers publish skills (manipulation, locomotion, perception) and spaces (evals, environments, and assets) to the store; agents compose over those skills to handle long horizon tasks; and digital rights management protects the investment of the developers who built them in the first place.

Internal Capabilities

We take a deliberate approach to building internal capabilities that accelerate our own iteration cycles. Speed is a moat. By building Tokamak, Menlo Cloud, and other infrastructure in-house, we move faster and at a smaller burn rate than other companies.

Tokamak

Tokamak is our internal software factory for compressing iteration cycles.

Tokamak serves two purposes:

Accelerated development: Automates build, test, and deployment pipelines so developers spend time on code, not bureaucracy.
Closed-loop iteration: Rapid deploy → observe → fix. Intelligence gathered from every deployment feeds the next.

Menlo Cloud

Menlo Cloud is our private cloud for robotics training and development. We built Menlo Cloud because humanoid robotics requires specialized hardware that hyperscaler clouds don't offer, ARM-based edge compute boards for firmware testing, robot-specific acceleration, and hardware-in-the-loop validation at scale.

Menlo Cloud serves two purposes:

Specialized robotics infrastructure: Provides compute and edge devices optimized for firmware testing, sensor processing, and real-time control, hardware that public clouds simply don't stock.
Predictable unit economics: Eliminates surprise pricing from hyperscalers while enabling us to scale testing hardware as fast as we need it.