formal-scaffold: E-SUBSTRATE-1 + 5 pillars + 2 iron rules + load-bearing-5 triage

[AdaWorldAPI]

Summary

Two follow-up commits on top of merged PR #227. First commit triages the 21 freshly-landed epiphanies to a load-bearing five; second commit reclassifies the formal-scaffold reference from "parked" to "dormant with five cited pillars" after E-SUBSTRATE-1 surfaced as the fundament.

User directive (de):

"Formal-PDE-Scaffolding ist nicht reanimiert — es ist fundiert: mit Jirak als korrekte Noise-Floor-Zitation und E-SUBSTRATE-1 als substrat-geometrisches Fundament, auf dem Cartan-Kuranishi, φ-Weyl und γ+φ stehen. Status bleibt dormant; aber 'dormant with five cited pillars' ist ein anderer Zustand als 'parked without a paper track.'"

Commit 1 — Triage to load-bearing five (5d1126c)

EPIPHANIES.md cut from 406 lines of E-ORIG / E-MEMB entries to 5 load-bearing prominent entries + a compact Deposit log of the other 15 (one-liner each, cross-refs intact). Bodies removed from the wall; IDs still addressable.

Load-bearing five:

ID	Finding
E-MEMB-1 (promoted FINDING → ISSUE)	Python↔Rust slice layouts incompatible at the 10 kD membrane
E-ORIG-1	NSM + 144 verbs are orthogonal composition axes
E-ORIG-5	NSM is pre-sliced for the role_keys 10K layout
E-MEMB-5	18D QualiaColumn = sigma_rosetta projected onto the SoA
E-MEMB-9	to_aurora_prompt() IS a BusDto — I9 operational in Python

ISSUES.md gained E-MEMB-1 as P1 open issue (scope @integration-lead @truth-architect domain:membrane) per double-entry rule.

Commit 2 — Formal-scaffold reclassification (7ba303c)

New load-bearing entry — E-SUBSTRATE-1

VSA-bundling in d=10000 guarantees Chapman-Kolmogorov by construction. Saturating bundle addition is associative and commutative in expectation; Johnson-Lindenstrauss + concentration-of-measure suppress deviations at rate ~e^(-d). (Hamming-space, Bundle) is an abelian semigroup ⇒ K(2τ) = K(τ)² holds without proof obligation.

Consequence: Markov property is a geometric consequence of substrate choice, not a testable assumption. D7's implicit Markov reliance is grounded, not silent. CK consistency test reclassifies from falsification-gate to implementation-sanity-check.

Substrate-bound constraint: MergeMode::Xor or non-commutative binding BREAKS this guarantee.

[FORMAL-SCAFFOLD] four pillars → five pillars

Prior 4-pillar entry preserved as SUPERSEDED (APPEND-ONLY rule); new entry elevates E-SUBSTRATE-1 as fundament:

E-SUBSTRATE-1:   VSA-bundling d=10000 ⇒ Chapman-Kolmogorov by construction
                 (FUNDAMENT — four pillars stand on this)
Cartan-Kuranishi: existence via involutive prolongation
                  → role_keys widths (2000/2000/2000/900/70/60/30) ARE
                    the Cartan-character spectrum, not arbitrary design
φ-Weyl equidist.: maximally-irrational sampling, no aliasing
γ+φ transform:    coordinate regularizer, fewer prolongation steps
Jirak 2016:       Berry-Esseen rate under weak dependence (arxiv 1606.01617)

Status: DORMANT (not parked; five pillars cited; paper track not active; consult before inventing new significance claims / thresholds / substrate changes).

Coupled revival track

Three revival candidates reclassified from isolated features to one coupled experimental access path:

• Chapman-Kolmogorov test → implementation sanity check (regression)
• VAMPE + Jirak pair → bound-derived thresholds replacing hand-tuned σ / hardness / abduction values
• Learned attention masks → empirical discovery of Cartan characters (if masks reproduce role_keys slice widths → intrinsic geometry proved)

Coupling acceptance: activate one ⇒ check other two for coherence with the scaffold in the same session.

CLAUDE.md — two new iron-rule invariants

I-SUBSTRATE-MARKOV: VSA-bundling d=10000 guarantees Chapman-Kolmogorov semigroup. Do NOT replace bundle with XOR or non-commutative binding for transition paths. MergeMode::Xor is legitimate for single-writer deltas (I1) but NOT a Markov-respecting transition kernel.

I-NOISE-FLOOR-JIRAK: Bits are weakly dependent by construction. Classical IID Berry-Esseen is WRONG. Use Jirak 2016 (arxiv 1606.01617, Annals of Probability 44(3) 2024–2063) for noise-floor / significance claims. Rate n^(p/2-1) for p ∈ (2,3], n^(-1/2) in L^q for p ≥ 4.

ISSUES.md — E-MEMB-1 substrate constraint

Updated the open E-MEMB-1 entry: any Python↔Rust bridge MUST respect E-SUBSTRATE-1. Identity-map between layouts would violate bundle associativity (they encode different algebraic structures over d=10000). Reconciliation must pick one canonical layout OR define a projector preserving bundle commutativity — a naive bit-by-bit remap is not acceptable.

Non-Goals (explicit per directive)

• No paper-track reanimation
• No revival-candidate rename to "roadmap items"
• No new crate, no new PR out of the scaffolding
• No scaffold propagation to unrelated PRs

The scaffold is available for future decisions, not forcing on current ones.

Test Plan

• Docs-only PR — no code changes, no build impact
• APPEND-ONLY rule honored: prior 4-pillar entry preserved with SUPERSEDED status; no body edits, only status updates
• CLAUDE.md iron rules land in existing "Substrate-level iron rules" subsection, no disruption to I1-I11
• ISSUES.md double-entry rule honored (E-MEMB-1 updated in Open section only)

https://claude.ai/code/session_01SbYsmmbPf9YQuYbHZN52Zh