# Product Requirements — Cosmogonic Quantum Mechalogodrom **Canonical capability authority:** [CQM-25](./alife/cqm-capability-profile.html) version **1.0.0**, adopted **2026-10-22**, is the binding project framework. Feature plans and capability assessments identify affected [stable axis IDs](./alife/cqm-capability-profile.json), subject/scope, measures, controls, falsifiers and pinned source/run provenance, including negative results. The [adoption contract](./MODULE-CONTRACTS-2026-06-26.md) governs taxonomy changes. Evidence states are not calibrated capability scores; canonical `assessment.score` fields remain null. The nine-axis 155-system comparisons, rankings and ratings below are historical specialist context and cannot transfer to CQM-25. The separate 14 Butlin indicators and causal experiment receipts retain their original scope; framework adoption does not certify consciousness, sentience, AGI or superintelligence. [Provisional author estimates](./alife/cqm-capability-profile.html#provisional-estimates) are a separate, explicitly labeled layer with their own [rubric and source data](./alife/cqm-capability-estimates.json). They retain source provenance and limitations, confer no measured capability, and do not populate canonical `assessment.score` fields. No overall or domain mean, ranking or aggregate verdict is published. **Living document. Solo-owner working brief, not team ceremony.** One page that states what this product IS, who it is for, what "done" means, and the hard constraints — so scope creep and overclaim are caught against a written intent. Architecture lives in [ARCHITECTURE-2026-06-26.md](./ARCHITECTURE-2026-06-26.md); this is the _why/what_, not the _how_. > **Provenance.** This file was added 2026-07-02 as one of five PM artifacts, then deleted 2026-07-06 > by `f37ef878` ("Delete 24 redundant markdown files per consolidation plan") — which removed three of > the five (PRD, TEST-STRATEGY, PERFORMANCE-TARGETS) while keeping > [RISK-REGISTER-2026-07-02.md](./RISK-REGISTER-2026-07-02.md) and > [SECURITY-ARCHITECTURE-2026-07-02.md](./SECURITY-ARCHITECTURE-2026-07-02.md). The PRD is not > redundant with either: the register tracks what could go wrong, the security doc tracks the trust > boundary, and neither states product intent — the thing scope creep is measured against. Restored in > place per the living-docs law (same file, facts re-grounded, no dated fork), with the three links > that had rotted to deleted docs repointed at their live successors. ## 1 · Product statement Cosmogonic Quantum Mechalogodrom is a **deterministic, browser-native, multi-substrate Artificial-Life and cognition testbed** — a "petri dish for NHSI digital biologics and proto-sentience research." It runs a seeded cosmos of up to 50,000 entities, an apex "Super Creature" cognition stack instrumented against published consciousness theories, and the Tsotchke MIT quantum/scientific-computing corpus as its non-transformer substrate. It is a **research artifact and a demonstrable experience**, not a product for sale and not a claim of sentience. ## 2 · Users & jobs-to-be-done | User | Job | | ----------------------------- | ---------------------------------------------------------------------------------------- | | The owner (0thernes) | Explore emergence/consciousness hypotheses with real math; ship a defensible artifact. | | A-Life / cognition researcher | Read code-grounded receipts; reproduce a claim from a single seed; compare vs the field. | | A curious visitor (Pages) | Watch a living cosmos in the browser; read honest, non-overclaimed explanations. | ## 3 · Success metrics (what "good" means) - **Honesty:** every load-bearing number is measured (`file:line` or gate receipt); zero overclaims survive an adversarial audit. Butlin framing stays "8/14 met + 6/14 partial, NOT sentient." Capability improvement claims must satisfy the relevant CQM-25 measures, controls and falsifiers with pinned evidence; canonical `assessment.score` fields remain null until calibrated. Unsupported claims are withdrawn when the source or experiment no longer supports them (see §6). - **Determinism:** identical seed → bit-identical run, always (machine-verified). - **Health:** `bun run check` green — 3,631 tests, 84.64% line / 82.21% func portable floor, 0 tsc/lint suppressions. (Receipt tokens; propagated by `scripts/sync-surfaces.ts` from `scripts/canonical-receipts.ts` — never hand-edit them here.) - **Experience:** 60 fps at the desktop tier; the app boots and self-recovers on GitHub Pages with no frontend secrets (see [BENCHMARKS-2026-06-26.md](./BENCHMARKS-2026-06-26.md)). - **Capability assessment:** CQM-25 is the primary reference; implementation and bounded-test evidence must retain their limits. **Historical comparison:** the nine-axis matrix records rubric breadth against a 155-system / 154-peer survey — breadth `3.8 / 5`, rank `#1/155`, population z `+3.236`, at a self-assessed peer maturity of `1.5/5` against a survey median of `4.0`. Rubric breadth does not establish general intelligence or quality, and the z is a **denominator-sensitive** statistic: growing the corpus moves it without any capability changing. See [NHSI-PROGRESS-DASHBOARD-2026-06-26.md](./NHSI-PROGRESS-DASHBOARD-2026-06-26.md) and [reports/2026-07-01-25-POINT-SCRUTINY-SCORECARD.md](./reports/2026-07-01-25-POINT-SCRUTINY-SCORECARD.md). ## 4 · Scope **In scope:** deterministic sim core; apex + pantheon cognition; Tsotchke substrate wiring; WebGL rendering; the observability/receipts stack; the docs/reports corpus; GitHub Pages deploy. **Out of scope (explicit non-goals):** - **Not a claim of consciousness/sentience/AGI/ASI** — those are research _targets_ and an architecture _program_, never achieved capability. The hard problem is untouched. - **Not an LLM/transformer** — the substrate is deterministic + Tsotchke quantum math, by mandate. - **Not a quantum-advantage claim** — the quantum math is exact classical simulation; a physical QPU would add speed/scale, not correctness. - **Not commercial / not multi-user / not a backend product** — single-machine, static-deployable, All-Rights-Reserved. No frontend secrets; managed cloud only if ever added. - **No world-first claim** — novelty is by integration, not primacy. ## 5 · Constraints (binding) Determinism (seeded `Rng`, no `Math.random`/`Date.now` in sim) · living-docs SSOT · full gate before commit · no PRs (local→`main`) · proprietary license · public/private boundary (see [SECURITY-ARCHITECTURE-2026-07-02.md](./SECURITY-ARCHITECTURE-2026-07-02.md)). These are enforced by the three MASTER files and `bun run check`. ## 6 · Open product bets (tracked, unproven) - **Coupling > count** — the emergence thesis, and still the headline unproven bet. `meanAbsCoupling ≈ 0.27` against a gate floor of `0.188` (`coupling-audit.test.ts`) and an honesty ceiling of `< 0.6`. It has climbed (0.167 → 0.183 → 0.197 → 0.270) but `0.27 ≪ 0.5`, the dense-binding threshold: the ~30 deep-wired faculties co-activate without forming a tightly bound mind. The repo's own phrasing stands — _a pile is not a mind_. See [COMPLEXITY-2026-06-26.md](./COMPLEXITY-2026-06-26.md) and [BRAIN-NEUROLOGY-CONSCIOUSNESS-ENGINEERING-ASSESSMENT-2026-07-06.md](./BRAIN-NEUROLOGY-CONSCIOUSNESS-ENGINEERING-ASSESSMENT-2026-07-06.md). - **Reproduction without selection** — heredity is live and lineage-verified (`entities.ts` `breedTraits`), but no birth or death path reads a heritable trait: birth is a flat trait-blind roll and lifespan is never inherited, so the population drifts rather than evolves. This is why the reproduction axis was corrected 4.0 → 3.5 on 2026-07-16 rather than defended. Closing it means making one reproductive quantity depend on a gene and gating it (`GATE-REPRO-SELECT`) — a product bet, not a bug. - **External validation** — no third party has reproduced a claim yet; the reproducibility artifact is the next product milestone (see [RISK-REGISTER-2026-07-02.md](./RISK-REGISTER-2026-07-02.md)).