Renormalization + cooperative game theory for multi-scale agent orgs

To model emergent organizations/alliances that aggregate across many scales, treat an org as a RENORMALIZATION of its members: the coarse-graining operator must output a same-typed agent so ONE solver runs at every level. Add a set-valued characteristic function v(S) with Shapley value (benefit = marginal contribution, not headcount) and the CORE for stability — a violated sub-coalition core constraint IS internal strife, an empty core IS fragmentation. Make v NETWORK-valued (counterparties reachable through members) so Shapley auto-prices brokerage rent. Represent membership as a normalized simplex (loyalty budget = 1, self as slack) for partial/overlapping/limited allegiance for free. Compute insight: cost tracks OBSERVATION not world size — local coalition games stay ~10-player (Shapley is exponential only in coalition size), keep only the attention cone live, sample the rest on demand.