Just for those unfamiliar with LORAWAN, LORA WAN is a combination of AI's "low rank adaptation" (LORA) approach with phsysionerual networks where by the WAN network grid is mirrored by an identically structured neural network. By performing a LORA on the neural network model of the physical network the network flow and efficiency can abstractly be "fine tuned" to a particularly observed use case, then deployed physically as node executor policy. Using it generally entails (1) system-wide packet-dump A LA NSA to build a training set (not saying whether or not the NSA does this sort of thing) then (2) fine tuning against the LORAWAN model and finally (3) deployment of the AI engine for each note to manage packet forwarding ,routing, and queueing policies.