Nils keynote made the $AUKI vision much clearer and once you see the architecture, it starts to click. The interesting part isn’t simply putting AI on a robot. Everyone is working on that. The bigger challenge is deciding what a machine should know locally, and what the environment should know collectively. Auki’s Machine/Domain split makes a lot of sense. The machine handles perception, immediate control, and essential responses with minimal compute. The Domain handles heavier, persistent context: mapping, interpretation, shared spatial information, and memory. That distinction matters because a robot shouldn’t have to rediscover an environment every time it enters it. The BracketBot demo showed this beautifully. Phones had already mapped the environment. BracketBot entered a remodeled space it hadn’t mapped itself, localized against the existing map, used that shared context to navigate, and then contributed higher-quality sensor data back into the model. That’s more than better robot navigation. That’s collaborative perception. Now add fixed cameras, phones, glasses, other robots, and Cactus into the same spatial Domain. Sensors become shared sources of context. Robots become both consumers and contributors. That’s where the Spaces / Sensors / Actuators model gets interesting: • Spaces provide persistent physical context • Sensors create distributed perception • Actuators turn information into physical action The Auki SDK is what starts tying these pieces together. Making one robot walk or grasp something is difficult. Making hundreds of different machines, sensors, and AI systems understand the same changing physical environment is a completely different challenge. It requires shared state, localization, spatial reasoning, communication, privacy, memory, and deployment. That starts looking less like a single robot running a model and more like distributed computing for the physical world. That’s the layer Auki is building. Bullish on $AUKI