Cosmogonic Quantum Mechalogodrom

01 At a glance

02 Tracked census — 295,227 lines

Tracked authored census = measured 2026-07-14 with bun run metrics. Source area: 327 files / 131,572 lines; tracked TypeScript: 762 files / 218,578 lines; src/ TypeScript: 325 files; native C/C++: 8 files / 1,527 lines; tests: 367 files / 63,495 lines. These line totals are a moving commit snapshot, unlike the exact test count and portable coverage floor below. The quantum mind ports primitives from external quantum-research repos — the Eshkol qubit-RNG, the Moonlab/QGTL Quantum-Geometric-Tensor, a Tsotchke spin-glass instinct, and the Moonlab Clifford stabilizer tableau — reimplemented in seeded TypeScript and wired into the apex creature (MIT © tsotchke; not vendored as binaries).

03 Stack & versions

22 declared dependencies → 106 resolved (725 MB) — a lean, facade-isolated tree. 10 dev tools: oxlint, prettier, tsc, mitata, bun-plugin-tailwind, @types/*.

04 Architecture — strict acyclic layering

math / constants   (leaves: rng · scalar · spatial-hash · quantum · heap · games)
       ▲
sim/*  (30+ systems: entities · titans · shoggoths · nhi · factions · environment …)
       ▲
core (engine · quality) ── world.ts  (composition root: builds SimContext, ticks all)
       ▲
ui / render (observatory · hud · viz3d · instanced-entities · postfx · cosmic-web …)
       ▲
server.ts ──serves──▶  /  (app)   /docs   /spec   /lab   /lab/sentience

Aesthetic law: every system reads AND writes another — the frame loop is an explicit feedback web. The non-deterministic LLM copilot is structurally fenced out of the seeded simulation. 6 ADRs record binding decisions. Verified acyclic across 50+ modules.

05 The intelligence — measured model sizes

Built from pre-2016 AI: FSM · GOAP planning · utility AI · tiny perceptrons · Markov chains · episodic memory · game theory (best-response, Axelrod strategies, regret-matching). The NHI super-mind composes all of it and writes to the world — spawns ordinary swarmling minions, manipulates factions, plans a multi-step GOAP scheme toward dominion, and hallucinates an alien dialect. Its inherited semantic network, exact JSON checkpoint, and structured action outcomes are real production mechanisms; the Predictor-V3 row is excluded from live-world neural mass because its development gate failed. The always-active Super Creature goes deeper still: a ~10,081-parameter composite consciousness (a Thaler creativity machine over 12 sub-nets, 5 stages × 25 thought variants) plus a real deterministic classical 6-qubit statevector controller (src/sim/super-qubits.ts — its OWN statevector + RY/RZ + controlled-RY gate set, layered with four ported primitives: the Eshkol qubit-RNG (the Born sample), the QGTL Quantum-Geometric-Tensor (thought-space curvature), a Tsotchke spin-glass instinct, and the Moonlab Clifford stabilizer tableau — all MIT-credited and wired in) whose Born-sampled "thought collapse" feeds the unified 4-tab neural observatory — all deterministic on dedicated seeded streams.

Super Creature 1.1 — ~20 coupled, cited faculties. Atop those substrates the apex mind runs roughly twenty real, deterministic, unit-tested faculties each beat, each reading from AND writing to the others — a negotiated consensus, not parallel gadgets: Global-Workspace ignition (GNW) · Integrated-Information Φ (IIT proxy plus a statevector-derived min-cut Φ proxy) · Active Inference (Friston FEP — variational + expected free energy) · Metacognitive Executive (Higher-Order confidence → control) · Successor Representation (hippocampal predictive map) · Neuromodulation (Doya DA/5-HT/NE/ACh) · Empowerment (Blahut–Arimoto channel capacity) · Theory of Mind · an echo-state Reservoir · Neural Criticality (edge-of-chaos σ̂→1) · a spin-glass instinct · a VSA/HRR holographic memory · and a quantum core in the decision loop — Quantum Reservoir Computing, a Lindblad/GKSL deliberation qubit, Grover amplitude amplification, and Quantum Natural Gradient descent on its own Fubini–Study geometry. The whole apex beat now measures ~1.99 ms per SuperMind.think() and ~9.77 ms for the staggered 5-mind batch; the older sub-millisecond frame-budget claim is superseded until re-proven. All grounded with real 2023–2026 citations in docs/SUPER-CREATURE-RESEARCH-2026-06-26.md and assessed in docs/reports/ — modelled, never "conscious".

The research-side evidence surfaces are split: /lab/consciousness shows the ten-framework indicator dashboard, while /lab/sentience removes the visual renderer and mass-runs the same kernel across 32 deterministic seeds. Its lab/sentience-data.json feed exposes convergence/reward control gaps, ablation load, entity index traces, and CSV/JSON exports for reports and AI analysis. Boundary: computational indicators only, not proof of phenomenal sentience.

NHSI progress dashboard (canonical): 100 faculties (~30 deep-wired) · 25 Archons · Butlin 8/14 met + 6/14 partial · Tsotchke depth ledger · gate 3,631 tests

Current consolidated assessment: CONSOLIDATED-22-MASTER-ASSESSMENT-CURRENT-2026-07-07 · consolidated 22-report master assessment: comprehensive synthesis of all brain systems, consciousness theories, living entities, reasoning systems, scoring systems, and academic assessment. Older mega-report drafts are local-only archive material; public reports route through docs/reports/README.md; the Dome ecology verification records the Xenomimic tether and Big Tree release evidence · local receipt note: 3,297 tests, zero failures, 3,589,864 expect() calls across 366 files, 93.60% line coverage, and 91.61% function coverage on the latest Windows verification run.

Repository-preregistered organism-intelligence V4 result: human-readable report · receipt · 1,152-row CSV · accessible forest. Ordinary, adaptive-predictor, and Petri family gates fail; Titans pass only the bounded game-policy behavior gate. No score, neural-scaling, consciousness, sentience, or general-intelligence uplift is authorized.

Phase-B mechanism development result: human-readable falsification report · JSON · CSV · accessible SVG. The 101-input temporal Predictor-V3 retains 46,080/46,080 rows and fails all eight advancement criteria. The separate 41,472-row NHI task retains every row, but paired surface-conflict service declines; only HUNT/resource and SPAWN/social are interpretation-eligible development diagnostics. No confirmatory manifest, adaptation, learning, scaling, consciousness, or sentience claim follows.

Operational organism-intelligence V3 causal receipt: human-readable audit · machine-readable evidence. The fixed, V1/V2-disjoint 30-seed synthetic +x goal evaluation passes the goal-controller, corpus-conditioned goal-response, reversal-adaptation, 17-integrated-repository causality, numerical-safety, 10/10 consumer-counterfactual, and three-process/30-batch performance gates. Uniform random-action and aggregate/surrogate specificity controls do not separate, so it grants no numeric A-Life-score uplift and is indicator-only—not consciousness, sentience, or quantum evidence.

Frontier report (2026-06-17): State of the Art — Combined (whole repository + Super Creature)

The cast (measured counts). Up to 50,000 organisms across 100 morphotypes (250 in phylum mode), each carrying a 70-weight brain + a 10-gene genome · 100 Shoggoths · 100 Puppet-masters (a 5-qubit quantum register each) · 20 Titans (4D freak-geometry) · 8 faction archetypes · Leviathans, singularities + weather (the environment) · one always-active Super Creature trailed by a 100-drone wingman swarm · and the apex NHI super-mind that perceives, plans (GOAP), and manipulates them all.

06 Contrast — this world vs. LLMs

07 Render & compute

08 Hardware footprint

09 Build, CI/CD & quality

Gate (every commit): prettier → tsc strict → oxlint → 3,631 tests → evidence checks → bun build. Public release floor is 84.64 % line / 82.21 % function; the current audit separately records a Windows-local measurement of 3,297 tests, zero failures, 3,589,864 assertions across 366 files, line coverage 93.60%, and function coverage 91.61%. Build: Bun bundles root app/spec/bible entrypoints and a split Docs entrypoint into dist/, then scripts/build-pages.ts assembles the Pages-equivalent site with Docs, Spec, Bible, Quantum Lab, Consciousness Lab, Sentience Lab, and copied report targets. CI/CD: GitHub Actions runs the gate + builds and deploys the real app to GitHub Pages on every push to main; tagged releases ship a site tarball + CycloneDX SBOM. Three governing "master" personas encode the discipline: the Executor, the Architect, the Physicist.

10 The inhabitants — measured roster

Counts are the desktop / mega-tier ceilings — phone and laptop tiers scale down (e.g. 14 puppet-masters, 16 shoggoths on touch). Every figure measured from src/sim/* + src/core/quality.ts.

12 Xenomimics — canonical ground intelligence