NTL Research — The Design Theory Behind It

These are the research documents that shaped NTL. They sit alongside the polished protocol docs and are intentionally rawer — they capture the why behind the architecture, not only the what. Read these when you want to understand why NTL is designed as an actual neural network rather than a messaging protocol, why the signal is a new concurrency primitive rather than a message type, and why IPv6 is a hard requirement rather than a preference.

#	Document	What it covers
01	The Signal Primitive	A new concurrency primitive. Nodes, signals, synapses as the NTL triad.
02	Machine Learning at the Transfer Layer	PyTorch for applications; NTL for networks. ML as infrastructure.
03	Neural Network as Base Layer	The routing engine is a neural network, executable on NPU silicon.
04	Network Transport & IPv6	Why GSPN’s fabric layer cannot live on IPv4 NAT.
05	The Twelve Principles	The full neural repertoire beyond PyTorch’s five.
06	First Steps	The concrete path from now to research preview.

NTL is a transport layer and specifies storage only as an interface. For complementary work on graph-native storage, see siafudb.org — one ecosystem backend among several, not a dependency.