What it is
Continuon is an embodied-AI project built on a simple idea borrowed from the Unix philosophy: one brain, many shells. Instead of a separate model baked into each robot, a single persistent brain carries memory and context across whatever body it is currently driving, so what a robot learns in one shell is available in the next.
Alongside the brain is ContinuonXR, a spatial-computing interface for controlling and observing robots in the room with you rather than through a flat dashboard.
Where it sits in the stack
Continuon is also where a load-bearing piece of the OpenCastor edition of PlatAtlas started: RCAN, the wire protocol underneath the whole robotics stack, began in the continuonai GitHub org. The spec and its Python and TypeScript SDKs now live in the RobotRegistryFoundation org. Continuon itself is a direction I am exploring for home robotics.
This one is earlier-stage than PlatAtlas and OpenCastor, closer to the Prototyper end of what I do: a place to try ideas about persistence, embodiment, and spatial control before they harden into the platform.