How we build it, in the open.
Thinking about Physical AI — long-form writing on the architecture, data, and systems that make physical intelligence work, from inside the deployment.
Three essays on why autonomous facilities are an architecture problem — the model, the data, and the system. More perspectives are in review and will follow, dated as they publish.
Intelligence, even when it works, does not produce autonomous facilities by itself. Between a robot that can perform a task and a facility that operates without humans, there is an entire layer of system engineering the field has historically underbuilt.
Models, by themselves, are no longer the dominant constraint on what robots can do reliably. The constraint has moved upstream — to the experience the models are trained on, and the systems that produce it.
Generating an action is not the same as deciding on one. The field's response has been a steady architectural shift — from reactive action generation toward predictive decision making — with the world model at its center.