When perception becomes motion, model accuracy is no longer the control question.
A camera that detects is an observation system. A camera that causes a robot to move is an execution system. The second is a different product, a different risk, and a different kind of evidence.
Industrial and edge deployments are crossing that line. Once a model’s output can become a physical effect, the question is not only whether the model was right. It is whether the system was authorized to cause that motion, on that object, at that moment — and whether anyone can show it later.
Nothing the system learns on the floor should enlarge what it is allowed to cause. A successful detection is not a grant. A confident model is not a grant.
Hood’s public claim here is architectural, not a deployment claim. Physical autonomy requires identity, authority before action, control at the boundary, and evidence afterward. Hood's physical-operations work is building; no live industrial-control deployment or safety certification is claimed. The detailed implementation remains private.
Hood Research. Category commentary. Not a specification and not an implementation guide.