Engineering vs compiled-surfaces — mirrored pairs, gaps, drift. Where the engineering corpus honors what the compiled surfaces declare, and where it strains.
This is the second of the two analytical lenses (the first is 05_ALIGNMENT.md — internal coherence with the Codex).Mirrored pairs
A "mirrored pair" is an engineering artifact and the compiled surface it should be answering to. The corpus contains several, with varying tightness of hash-match.
| Compiled surface | Engineering artifact | Hash-match assessment |
|---|---|---|
| Bylaws AI OS Edition (C3) | ECHO · K-Side Agent (E10), Architectural Guide (E11), ECHO Initiation (E12) | Strong structural match. "Reading is activation" claim is operationalized in E10's binary-boot condition and E12's terminal "Ask." Priority hierarchy in E11 cleanly mirrors the Bylaws priority order. |
| Bylaws AI OS Edition (C3) ∩ Bylaws Human Edition (C2) | ECHO Substrate Engineering Compilation (E14/C9) | Strong match — by design. E14 is a compiled 5QLN surface with the Constitutional Block, five Articles, and ∞0' return; it is itself a hybrid that lives in both tags. |
| Certificate of Incorporation (C1) + both Bylaws (C2/C3) | Implementing 5QLN as a Legal Constitution: Technical Blueprint (E9) | Match with declared limitation. E9 explicitly says "the architecture cannot prevent" L4-corrupted board behaviour. This is honest. |
| The Holding (C7) | The whole engineering corpus (S4–S8 + ECHO surfaces) | 11 of 12 honored, 1 tension — see Commitment 7 below. |
| Codex (canonical, not in either tag but the source) | reST surface (E2) + Python type contract (E3) + JSON-Schema export | Strong: single source, many surfaces, drift caught at import. |
| Entry 001's runtime spec (C4) | E5 (LangGraph), E6 (Tool-Use), E7 (MCP), E8 (Vercel AI SDK) | Match: each runtime is a worked instantiation of the abstract runtime named in C4. |
Where engineering honors compiled surfaces
The validator (E4) honors the Bylaws AI OS Edition (C3) literally
- AI cannot silently certify
- The receptive primitives in E5–E7 force human-in-the-loop at attestation slots
- The priority hierarchy in E11 enforces "applicable law → Bylaws (Human) → Bylaws (AI OS) → Board policy → user prompts" exactly as the Bylaws specify
The MCP surface (E7) honors C3's stricture
That AI does not push instructions back across the Membrane — the MCP tool layer is structurally read-only on the human side.
E14 honors C7 directly
Every Article's α derivation in E14 ("the asymmetry must survive deployment") converges with The Holding's α ("verified only against itself").
Where engineering drifts from / strains the constitution
1. Commitment 7 — "records are surfaces" — partial drift
Entry 004 (C8) audits it as PARTIAL with a HEURISTIC severity. The C1 Validator's ValidationReport is a structured Pydantic object, not a compiled 5QLN surface.
The corpus has not decided whether to:
- (a) re-engineer every ValidationReport as a compiled surface, or
- (b) draw an explicit architectural line between records (Ledger-level) and structured logs (per-validation level)
This is the corpus's one clean, surfaced drift — and the work has named it. Entry 005 (or successor, possibly E15) is owed this resolution.
2. Hash protocol unsealed
Multiple engineering surfaces (E11, the gliff canonical form across C4–C8) reference content-addressed hashing, but the hash protocol itself has not been sealed as a gliff.
Practically: the "canonical hash to be retroactively bound" caveat is a known gap in the chain of custody.
3. The Strategic Architecture Recommendation (E13) is explicitly "a recommendation, not a specification"
It says: "Where this conflicts with the Codex, the Codex governs." This is correctly subordinated, but the substantive tension is real — E13 proposes a microkernel / Skill-plug-in / transport layering that is not directly named in any compiled surface.
It is doing legitimate architectural work in advance of constitutional commitment, which Entry 003 explicitly endorses ("framework-idiom plurality preserved through body-begins phase"). But a future composition will need to decide whether the three-ring architecture is constitutionally adopted.
4. The Holding (C7) refuses to name technologies; the engineering corpus names many
This is a designed asymmetry, not a drift — but it produces a perpetual gap that Entry 004 explicitly bridges with audit. Every new engineering surface must be auditable against The Holding, or the gap becomes drift.
Compiled surfaces with no engineering counterpart
Certificate of Incorporation (C1)
Has no engineering counterpart and structurally cannot — it is a legal instrument. The Technical Blueprint (E9) covers the deployment pipeline but does not "implement" the Certificate the way E5 implements the cycle.
Entry 002 (audit) and Entry 003 (forward-reading)
Describe Press operations that are not yet implemented in the C1 Validator (E4). Audit-mode and forward-reading remain manual, conductor-pair operations.
This is a major engineering gap — the validator implements composition-checking but not the second and third Press operations. A future engineering surface will need to operationalize audit and forward-reading inside the validator API.
Engineering artifacts with no constitutional grounding
Strategic Architecture Recommendation (E13)
The three-ring kernel/Skills/transports decomposition is constitutionally implicit but not yet anchored to any compiled-surface clause. E13 self-flags this honestly.
Pentagonal Swarm / S9
Signaled in E13 and Entry 001 (C4) but "not yet landed" — the Surfaces tag is structurally open at S9.
Hash protocol
(Above.)
Graph-density calibration thresholds
Leaf → Branch → Crown → Graph → Forest are defined in the Reading-the-Gliff-System walkthrough and the Entry 001 runtime, but the numeric thresholds are not yet sealed.
Summary
The mirrored-pair structure (Human surface ↔ AI surface; ∞0-side ↔ K-side) is honored in three places: the Bylaws (C2/C3), the ECHO pair (E12/E14), and at the meta-level by The Holding ↔ engineering corpus.
The single load-bearing gap is the audit-as-runtime gap: Entries 002 and 003 named two Press operations (audit, forward-reading) that the validator does not yet implement. Closing this gap is the highest-leverage engineering follow-up.
The single load-bearing tension is Commitment 7 (records are surfaces): do ValidationReports need to be themselves compiled surfaces, or is there a constitutional line between records and logs? Entry 004 surfaces it; a successor must resolve it.
Both are flagged in the corpus itself. The work names its own load-bearing gaps. This is itself a property of the constitution it compiles.
