Unify heterogeneous value/utility as shadow prices, not per-type laws

When building a value model over heterogeneous currencies (goods, skill, relationships, protection, information), do NOT write one bespoke value law per type — that is the anti-pattern. The primitive: value = the shadow price (LP/convex dual) of a constraint in each agent's own recipe-derived optimization. Recipes are the constraint matrix (columns=activities, rows=input stocks); the dual of an input's row IS the agent's willingness-to-pay, endogenous and per-agent. This gives reservation value and self-limiting demand (complementary slackness: surplus input => shadow price exactly 0) for free. The types become regions of a small property vector {enters-whose-production, role(stock/coefficient/column/state-guard), rivalry, excludability, state-contingency, observability}. Precedent: Arrow-Debreu GE prices good x location x date x state under one system. Also: information is NOT one scalar — split use-value (EVPI, retained on diffusion) from exclusivity rent (vanishes on diffusion); once exclusivity enters a coalition game it has externalities => partition-function form, so plain Shapley/core do not apply.