feat(a-sounder-constitution): typed civic contagion simulator — rights as transition constraints

[hyperpolymath]

A Sounder Constitution

A new research proof-of-concept in a-sounder-constitution/: a Plague Inc.-style civic simulator where the "pathogen" is a constitutional doctrine, norm, fear, right, or type error.

The one idea: rights are not resources (sliders, budgets). They are type constraints on legal state transitions. An unsound constitution can produce fast, stable order by permitting domination. A sound constitution makes some coercive moves literally unrepresentable — slower, but far more resilient.

This sits squarely in the incubator's remit: the load-bearing claim is type-theoretic, it ships an Idris2 formalization alongside an executable model, and it has a graduation path toward typell.

What's here

Path	Role
src/constitution.js	the type checker — under a sound constitution Person -> Property, Emergency -> PermanentPower, etc. are UNREPRESENTABLE; liberty restrictions need a discharged proof obligation. Unsound permits every coercive move and books hidden legitimacy debt.
src/engine.js	a deterministic typed cellular automaton — regions, mobile doctrine "gliders", neighbour diffusion, exogenous shocks, outcome classification. Coercive effects are routed through the type checker, so the same strategy diverges purely by constitution.
web/	zero-dependency canvas simulator; the live type-checker log makes the thesis visible (✗ unrepresentable vs ✓ permitted (domination)).
sim/compare.mjs	headless, quantitative sound-vs-unsound demonstration.
tests/	node --test suite (12 tests) pinning mechanics and thesis-level behaviour.
formal/Constitution.idr	the static certificate: the sound Person type has no Property inhabitant (soundNeverProperty), sound standing transitions are rank-monotone (soundMonotone), and lawful coercion demands proofs.
docs/, README.adoc, MOTIVATION.adoc, 0.1-AI-MANIFEST.a2ml	design, type-theory, glider catalogue, near-miss/novelty survey, graduation path.

The result (hold strategy fixed, vary only the constitution — npm run sim)

STRATEGY A · reach for coercion (emergency, surveillance, caste)
mode     t→order  safety  legit  domin  resil  blocked   outcome
unsound      32     67%     0%    93%     0%        0   Caste stability (order by domination)
sound     never     45%    52%     0%    32%    23778   Contested / transitional

STRATEGY B · build legitimacy (rights, virtue, due process)
sound        40     71%    90%     0%    86%     1674   Rights-preserving resilience

Unsound buys order by domination (legitimacy/trust → 0, huge hidden debt); sound refuses the coercive moves at the type level (tens of thousands "blocked") so domination never happens. In Reconstruction, the rights patch only "takes" under a sound constitution — under unsound the pre-existing caste can't be undone.

How to run

cd a-sounder-constitution
npm run sim      # headless thesis demonstration
npm test         # 12/12 tests
npm run serve    # python3 -m http.server 8080  →  open /web/

Verification done in this environment

• ✅ node --test — 12/12 pass
• ✅ node sim/compare.mjs — produces the contrast above
• ✅ all web module URLs resolve over HTTP; ui.js passes node --check
• ⚠️ formal/Constitution.idr is written to compile clean (%default total, no believe_me) but is not yet run through idris2 here (none on PATH) — flagged as pending in formal/README.adoc, per the repo's blocker convention.
• ⚠️ The web UI was not opened in a live browser in this environment (no headless browser available).

🤖 Generated with Claude Code

https://claude.ai/code/session_015w8C1xaGwiDcHHjfuxHBd6

---

Generated by Claude Code