Add the Living Market AMM

.github/workflows/ci.yml

@@ -25,9 +25,11 @@ jobs:
       - uses: actions/checkout@v7
       - uses: dtolnay/rust-toolchain@master
         with:
-          toolchain: "1.89"
+          # wasm-pack 0.15.0 requires cargo-platform 0.3.3 (Rust 1.91+).
+          # The rust job above remains pinned to the workspace MSRV.
+          toolchain: "1.91"
           targets: wasm32-unknown-unknown
-      - run: cargo install wasm-pack --locked --version 0.13.1
+      - run: cargo install wasm-pack --locked --version 0.15.0
       - uses: pnpm/action-setup@v6
         with:
           version: 10.14.0

.github/workflows/pages.yml

@@ -21,11 +21,12 @@ jobs:
       - uses: actions/checkout@v7
       - uses: dtolnay/rust-toolchain@stable
         with:
-          toolchain: 1.89.0
+          # wasm-pack 0.15.0 requires cargo-platform 0.3.3 (Rust 1.91+).
+          toolchain: 1.91.0
           targets: wasm32-unknown-unknown
       - uses: Swatinem/rust-cache@v2
       - name: Install wasm-pack
-        run: cargo install wasm-pack --locked --version 0.13.1
+        run: cargo install wasm-pack --locked --version 0.15.0
       - uses: pnpm/action-setup@v6
         with:
           version: 10.14.0

PDD.md

@@ -675,7 +675,7 @@ artifact resolves and records that identity. Micro continues to support exact
 oracles even when exhaustive search over Showcase would be inappropriate, and
 an approximate or candidate-bounded Showcase frontier must be labeled as such.
 
-The conformance service enforces minimum Showcase pressure for all thirteen
+The conformance service enforces minimum Showcase pressure for all fourteen
 problems. Every result also receives transport-neutral **Model size** evidence:
 accounts, rules, steps in the trace, rule-check probes, and alternatives. These
 measurements are explanatory, not objectives, and cannot affect validity.
@@ -844,7 +844,7 @@ artifact and frame. Invalid identities visibly fall back to a catalog default,
 and shared links never auto-run computation. The same URL therefore works
 against the native server, the browser Wasm engine, and GitHub Pages.
 
-The implemented Studio exposes the complete thirteen-problem Showcase
+The implemented Studio exposes the complete fourteen-problem Showcase
 surface: pathfinding and networks use graphs; Sokoban and Connect Four use
 grids; Exact Cover uses a constraint matrix; scheduling and perishables use
 timelines; markets expose multi-party settlement and rejected shortfalls;
@@ -872,7 +872,7 @@ operational run and transport failures.
 Portable artifacts prove offline playback for the full catalog. The native
 path proves generated OpenAPI calls, resumable SSE, pause/resume/cancel hooks,
 artifact and frame retrieval, and scrubbing. Browser tests prove that the same
-thirteen-problem Rust service initializes, runs, streams observations, publishes
+fourteen-problem Rust service initializes, runs, streams observations, publishes
 artifacts, and cancels inside an isolated Worker. The Pages build uses a
 repository-relative Vite base, includes its Wasm binary and `.nojekyll`, and
 deploys from `main` without a server. Specialized projections are added
@@ -1172,7 +1172,7 @@ compatible set or sequence of those resolutions, but it cannot bypass them.
 
 ## 10. Executable conformance suite
 
-The project now validates the thesis with thirteen deliberately different
+The project now validates the thesis with fourteen deliberately different
 problems. Full formal specifications live in [PROBLEMS.md](PROBLEMS.md).
 
 | Problem | What it demonstrates |
@@ -1184,6 +1184,7 @@ problems. Full formal specifications live in [PROBLEMS.md](PROBLEMS.md).
 | Job shop | Six operations, three machines, identified capacity slots, precedence tokens, completion front, makespan, and direct-oracle agreement |
 | Rescue | Four hidden sites, 32 encoded scenarios, restricted views, sensing and evacuation, seeded Monte Carlo, approximate success/resource front, and replay |
 | Bridge | Repeated multi-agent bids, escrow, capacity, atomic round reset, alternative mechanisms, and fairness/credit frontier |
+| Living Market | Exact state-evaluated constant-product swaps, endogenous energy/credit value, liquidity depth, heterogeneous actors, counterfactual replays, and exact Shapley price attribution |
 | Marketplace | Account-derived matching, complete shortfalls, four coupled six-party settlements, clearing, and participant-utility front |
 | Logistics | Recurrent encoded chance, long rollouts, repair loops, risk projections, approximate policy front, and MCTS route planning |
 | Connect Four | Standard 7×6 geometry, compact 69-line win certificates, encoded gravity and terminal truth, adversarial MCTS, and plain-board oracles |
@@ -1577,7 +1578,7 @@ factor.
 13. Add persisted search checkpoints, worker leases, task notifications, and
     tenant-aware authorization only when the MCP reference boundary is moved
     into a real multi-process deployment.
-14. Measure full-artifact transfer and rendering across the thirteen Studio
+14. Measure full-artifact transfer and rendering across the fourteen Studio
     adapters; add paged model projection, incremental Pareto/Monte Carlo
     publication, or compact rate schemas only where measured pressure warrants
     them.
@@ -1881,7 +1882,7 @@ Axionomy succeeds when:
   the same Rust service in a Worker; they can inspect retained solver evidence,
   scrub every accepted exchange, and inspect exact balances and deltas without
   a parallel mutable world model or handwritten cross-language contract.
-- All thirteen canonical problems are discoverable, runnable, replayable, and
+- All fourteen canonical problems are discoverable, runnable, replayable, and
   meaningfully inspectable through Studio and static Showcase artifacts, with
   explicit Micro and Stress selection on live interfaces.
 - CLI, HTTP, MCP, and the browser Worker return the same semantic problem

PROBLEMS.md

@@ -33,24 +33,24 @@ they bypass the example's trusted action helper.
 
 ## Conformance matrix
 
-| Capability | Maze | Sokoban | Exact cover | Workshop | Job shop | Rescue | Bridge | Marketplace | Logistics | Connect Four | Mission | Perishables | Work League |
-| --- | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: |
-| Multi-account atomic rewrite | ✓ | ✓ |  |  | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
-| Preserved facts/catalysts | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
-| Declared invariants | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |  |
-| Resource objective | ✓ |  |  | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |  | ✓ | ✓ | ✓ |
-| Exact Pareto front | ✓ |  |  | ✓ | ✓ |  | ✓ | ✓ |  |  |  | ✓ |  |
-| Approximate or derived Pareto comparison |  |  |  |  |  | ✓ |  |  | ✓ |  | ✓ |  | ✓ |
-| Infeasible instance |  | ✓ | ✓ |  | ✓ |  |  | ✓ | ✓ |  | ✓ | ✓ |  |
-| Specialized proposer |  |  | Algorithm X |  | Branch optimizer | Scenario/MC evaluator | Auction | Assessment clearing | Monte Carlo/MCTS | MCTS/minimax oracle | ISMCTS/scenario evaluator | Event agenda/index | Competing seeded policies |
-| Generic algorithm | BFS/A*/Dijkstra/Pareto | BFS | BFS | BFS/best-first/Pareto | Best-first/Pareto | Rollout/MC/Pareto | BFS/Pareto | Pareto | Rollout/MC/Pareto | MCTS | ISMCTS/Rollout/MC/Pareto/RL | Pareto | Replay-derived vector comparison |
-| Hidden or stochastic state |  |  |  |  |  | ✓ |  |  | ✓ |  | ✓ |  | ✓ |
-| Multi-agent resolution |  |  |  |  |  |  | ✓ | ✓ |  | ✓ | ✓ |  | ✓ |
-| Long-horizon trajectory |  |  |  |  |  |  |  |  | ✓ | ✓ |  | ✓ | ✓ |
-| Per-step leaderboards |  |  |  |  |  |  |  |  |  |  |  |  | ✓ |
-| Learning trajectory |  |  |  |  |  |  |  |  |  |  | ✓ |  |  |
-| Observation-scoped tree |  |  |  |  |  |  |  |  |  |  | ✓ |  |  |
-| Deterministic replay test | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
+| Capability | Maze | Sokoban | Exact cover | Workshop | Job shop | Rescue | Bridge | Marketplace | Logistics | Connect Four | Mission | Perishables | Work League | Living Market |
+| --- | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: |
+| Multi-account atomic rewrite | ✓ | ✓ |  |  | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
+| Preserved facts/catalysts | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
+| Declared invariants | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |  | ✓ |
+| Resource objective | ✓ |  |  | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |  | ✓ | ✓ | ✓ | ✓ |
+| Exact Pareto front | ✓ |  |  | ✓ | ✓ |  | ✓ | ✓ |  |  |  | ✓ |  |  |
+| Approximate or derived Pareto comparison |  |  |  |  |  | ✓ |  |  | ✓ |  | ✓ |  | ✓ |  |
+| Infeasible instance |  | ✓ | ✓ |  | ✓ |  |  | ✓ | ✓ |  | ✓ | ✓ |  | ✓ |
+| Specialized proposer |  |  | Algorithm X |  | Branch optimizer | Scenario/MC evaluator | Auction | Assessment clearing | Monte Carlo/MCTS | MCTS/minimax oracle | ISMCTS/scenario evaluator | Event agenda/index | Competing seeded policies | Seeded actors/Shapley |
+| Generic algorithm | BFS/A*/Dijkstra/Pareto | BFS | BFS | BFS/best-first/Pareto | Best-first/Pareto | Rollout/MC/Pareto | BFS/Pareto | Pareto | Rollout/MC/Pareto | MCTS | ISMCTS/Rollout/MC/Pareto/RL | Pareto | Replay-derived vector comparison | Exact coalition replay |
+| Hidden or stochastic state |  |  |  |  |  | ✓ |  |  | ✓ |  | ✓ |  | ✓ |  |
+| Multi-agent resolution |  |  |  |  |  |  | ✓ | ✓ |  | ✓ | ✓ |  | ✓ | ✓ |
+| Long-horizon trajectory |  |  |  |  |  |  |  |  | ✓ | ✓ |  | ✓ | ✓ | ✓ |
+| Per-step leaderboards |  |  |  |  |  |  |  |  |  |  |  |  | ✓ | ✓ |
+| Learning trajectory |  |  |  |  |  |  |  |  |  |  | ✓ |  |  |  |
+| Observation-scoped tree |  |  |  |  |  |  |  |  |  |  | ✓ |  |  |  |
+| Deterministic replay test | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
 
 ## Runnable examples
 
@@ -66,6 +66,7 @@ cargo run -p axionomy-problems --example workshop
 cargo run -p axionomy-problems --example scheduling
 cargo run -p axionomy-problems --example rescue
 cargo run -p axionomy-problems --example bridge
+cargo run -p axionomy-problems --example amm
 cargo run -p axionomy-problems --example marketplace
 cargo run -p axionomy-problems --example logistics
 cargo run -p axionomy-problems --example connect_four
@@ -117,6 +118,7 @@ catalog, so those two public explanations must change together:
 | Job shop | Six precedence-constrained operations across three machines and 18 capacity slots |
 | Rescue | Four hidden sites, 32 encoded Nature scenarios, noisy sensing, contact, and return evacuation |
 | Bridge | Two consecutive auctions/allocations with escrow, recharge, atomic round reset, and fairness tradeoffs |
+| Living Market | Eight actors, exact nonlinear AMM laws, real needs and obligations, liquidity changes, information, a demand shock, internal supply response, three replayable realities, and 128 exact actor coalitions |
 | Marketplace | Four coupled orders across 14 accounts with shared budgets, inventory, shipping capacity, taxes, and commissions |
 | Logistics | Four deliveries, recurrent weather and breakdowns, refueling/repair loops, up to 53 accepted transitions, Monte Carlo, and MCTS |
 | Connect Four | Standard 7×6 board and 69 four-cell win certificates; 226 concrete rates replace the old 1,282-rate 4×4 projection |
@@ -843,8 +845,56 @@ replayed snapshot.
 - Live frame publication without moving transport state into the economy.
 - Shared URLs that restore a precise problem, outcome, replay step, and
   leaderboard on both a server and static GitHub Pages.
-- Concrete-rate expansion pressure that strengthens the case for future typed
-  parameterized rate schemas without prematurely changing kernel semantics.
+- Concrete-rate expansion pressure that keeps the new exact scalar
+  parameter/law schema intentionally narrower than a general callback system.
+
+## 14. Living Market AMM
+
+Source: `crates/axionomy-problems/src/amm.rs`
+
+### Specification
+
+Energy and settlement credit trade through a constant-product pool. The
+founding reserve ratio is only an opening hypothesis: a generator supplies
+energy, a factory and household buy it to satisfy encoded needs, an informed
+speculator and demand whale buy for different reasons, an adaptive provider
+changes market depth, and an internal arbitrageur sells into the resulting
+scarcity. Credit remains useful because it settles encoded obligations;
+energy remains useful because consuming it produces goal and utility assets.
+There is no external market, stablecoin peg, or oracle.
+
+Every swap is an exact state-evaluated rate law. The kernel reads pre-exchange
+reserves, evaluates integer floor division with a 0.3% fee, checks the caller's
+minimum output, applies both reserve and trader deltas atomically, and emits a
+replayable receipt. Proportional liquidity deposits preserve price while
+changing depth. Withdrawals burn shares against exact pre-state ownership.
+
+### Acceptance laws
+
+- Independent `u128` quotes agree with kernel-computed swap receipts.
+- Asset, energy lifecycle, credit, liquidity-share, actor-goal, and information
+  invariants hold across every replay.
+- Constant product never decreases during a swap and round trips cannot create
+  credit.
+- Quotes are monotone in input and remain strictly below the output reserve.
+- Market Day reaches the factory, household, generator, and information goals.
+- The no-whale and thin-liquidity counterfactuals produce different discovered
+  prices without changing the AMM law.
+- Direct per-actor price changes sum exactly to total reserve-price movement.
+- Exact Shapley allocation replays all 128 coalitions of the seven active price
+  actors and sums exactly to the full counterfactual price difference.
+- Studio price history, reserve curve, actors, rule expressions, rejected
+  slippage probes, and leaderboards derive only from replay evidence.
+
+### API pressure
+
+- Serializable state-dependent rates without opaque domain callbacks.
+- Exact named exchange parameters and authoritative inequality conditions.
+- Nonlinear effects while retaining assets, accounts, rates, and exchanges as
+  the only semantic primitives.
+- Causal attribution through isolated replay forks rather than external
+  analytics state.
+- A dedicated market surface that remains a projection of Rust-owned truth.
 
 ## Cross-problem acceptance tests
 
@@ -905,7 +955,7 @@ cargo test --workspace --all-targets --all-features
 The benchmarks support the generalized kernel, but they also show its next
 limits clearly:
 
-1. Scheduling and stochastic outcomes need typed parameterized rate schemas to
+1. Scheduling and stochastic outcomes need richer typed parameter schemas to
    avoid eager concrete expansion.
 2. Solvers need a standard finite binding enumerator derived from schemas and
    account capabilities.
@@ -926,5 +976,5 @@ limits clearly:
    or domain-proven pruning; approximate epsilon dominance must never be
    mislabeled as exhaustive Pareto truth.
 
-Any future abstraction must continue to pass all thirteen problems without
+Any future abstraction must continue to pass all fourteen problems without
 moving authoritative meaning into solver callbacks or an external world.

README.md

@@ -82,7 +82,7 @@ evaluation and rating design, and
 between the Work Utility System and the Autonomy Harness that owns agents,
 learning, deployment, and assurance.
 
-The thirteen reference problems expose explicit **Micro**, **Showcase**, and
+The fourteen reference problems expose explicit **Micro**, **Showcase**, and
 **Stress** instances. Micro keeps exact fixtures and independent oracles fast;
 Showcase is the default that Studio and the generated artifacts use. Stress
 raises a domain-relevant dimension—sampling, topology, board size, horizon,
@@ -135,6 +135,8 @@ adapters.
 
 - User-defined asset, account, rate-ID, role, and checked `Quantity<N = u64>` types.
 - Atomic multi-account exchanges with consume, produce, and preserve baskets.
+- Serializable state-evaluated rate laws with exact integer expressions,
+  named exchange parameters, and authoritative conditions for nonlinear rules.
 - Explanatory assessments, complete shortfalls, projected deltas, and receipts.
 - Declared invariants, asset-configured goals, isolated forks, and deterministic replay.
 - Stable Serde models plus optional `BigUint`, dimension-safe `uom`, and Jiff authoring.
@@ -239,7 +241,7 @@ pnpm dev
 Open `http://127.0.0.1:5173`. The native server is optional: when it is healthy,
 Studio uses HTTP/SSE and exposes pause, resume, and cancellation; otherwise it
 loads the same Rust service as WebAssembly in an isolated Worker and keeps Run
-enabled for all thirteen problems. Worker cancellation is immediate by
+  enabled for all fourteen problems. Worker cancellation is immediate by
 terminating that disposable worker. If WebAssembly cannot initialize, the
 committed Showcase artifacts still provide read-only playback. The status badge
 is backed by a current health/initialization check rather than by a catalog that
@@ -489,6 +491,7 @@ contract, deployment limits, and integration details.
 | Job shop | Precedence, identified machine slots, makespan, per-job completion | Best-first, branch search, direct oracle, exact allocation Pareto front |
 | Rescue | Hidden truth, seed, observation, belief, chance, sensor use | Exact scenario evaluation, seeded Monte Carlo, approximate policy front |
 | Bridge | Capacity, bids, escrow, priority, joint resolution | BFS, first-come, auction, exact priority/credit Pareto front |
+| Living Market AMM | Closed energy/credit economy, exact constant-product swaps, fees, liquidity shares, real needs, obligations, information, actor price attribution | Market-day replay, no-whale and thin-liquidity counterfactuals, exact coalition Shapley attribution |
 | Marketplace | Buyers, sellers, carriers, tax, commission, competing orders, participant utility | Assessment matching, replayable clearing, exact allocation Pareto front |
 | Logistics | Orders, routes, fuel, time, weather, breakdown, repair | Long rollouts, risk-aware Monte Carlo, route MCTS, approximate policy front |
 | Connect Four | Identified cells, gravity, turns, line counts, wins, draw | Vector-valued MCTS with plain-board/minimax oracles |