Cameron asked for the hardest, most load-bearing structure so later agents can pile on. The flow law (signals/messages/money as nodes exchanging typed flows, with tap/inject/redirect) has one shared shape; build it once as new isolated modules. - src/flow.rs — FlowGraph: domain-agnostic topology (directed, typed, gated edges), deterministic BFS reachability from a root set under a caller gate predicate (the inject/redirect precondition), and the tap subscription registry (the player's senses; take = untap owner + tap player). Domains own node data; the engine holds only wiring + listeners. BTree* for deterministic iteration/serialization. - src/schedule.rs — Schedule<E>: deterministic scheduled-event queue on the tick clock, generic over payload. at/after/due(now) in (tick, seq) order, no RNG, no wall clock, serde-stable. Decouples when from what for message delivery, account cadence, buffered processing. Recurrence is reschedule-on-fire. 15 unit tests (reachability+gates, directed edges, disconnected components, tap registry, schedule ordering/draining/tie-break/retain/next_tick/ recurrence, serde round-trips). 94 total green; clippy clean both features. spec/flow-substrate.md documents the contract for downstream agents (how reach/messages/economy/detection each ride it) and the non-goals kept out on purpose. DESIGN.md flow-law section gains "Code expression". Also refines chargen.md archetypes into the diegetic-origin frame (Proposal [OPEN], five origins mapped to flow-law verbs). Defense: constitution "The flow law" (this is its code expression) and "Self-similar scale" (one interface at every scale — the engine is the shared shape so aggregates/new domains slot in without a rewrite). The no-dead-code guardrail is honored by shipping a spec that names the real consumers and requiring reach to be wired before flow-substrate.md is IMPLEMENTED; the engine is the shape the constitution already mandates, not speculative generality.