AGInetwork is the runtime on which gOS instances coordinate, and it settles what they exchange: it routes a job to the nodes that can do it, prices the result and pays the contributors out of a staked pool. Expert models, symbolic programs, executable artifacts and learning events move across it with provenance fields for claimed source and production context. At runtime, those components can be assembled for a task rather than fixed inside one general model. This is the network layer of Compositional Intelligence: it gives sovereign machines a way to combine distinct capabilities while leaving each participant responsible for what it runs, keeps and shares.

Provenance is the load-bearing part. It carries the claimed source and context alongside the key-based origin a signature establishes, which recipients can weigh, test or dispute. Accrued trust and network position can then inform which signals propagate and how far, rather than replacing judgment with a universal score. Nothing here requires a shared canonical database, so different nodes can reach different conclusions from the same event.

This is where Cultural Intelligence becomes infrastructure. A group that can inspect the recorded source claim for a judgment, the relationships that carried it and where dissent stayed legible can accumulate knowledge without erasing the distinctions that made it useful. AGInetwork therefore supports a plural, open-ended ecology rather than one global view: memories remain distributed, communities can normalise signals differently, and reusable artifacts can cross between them with provenance attached.