Measurement Schema × ADR-074 Alignment Review¶

Archived 2026-06-01 — moved from docs/research/ to docs/archive/ in the docs consolidation (ADR-083). A finished one-off alignment brief; its load-bearing conclusion (the schema's architectural-independence property) is carried forward in the Reader's Guide and ADR-074.

Date: 2026-05-31 Status: Architectural reference brief — for ADR-074 follow-up consideration, light-paper positioning, or internal architectural alignment. Inputs: docs/decisions/074-pulse-as-multi-domain-judgment-layer.md; docs/research/foundations/measurement-schema.md; docs/research/measurement-schema-working.md decomposition pass findings.

Purpose¶

ADR-074 (Pulse as a multi-domain judgment layer) made an architectural claim that the schema supports multi-domain expansion without rebuild. The measurement schema exercise (16 swings + 16 repeats + 34 decompositions) strengthens that claim with a more specific architectural finding: the schema's design dimensions are mutually independent.

This brief surfaces the alignment so it can feed downstream work (ADR follow-up, light-paper positioning, internal architecture documentation).

ADR-074's original architectural claim¶

ADR-074 argued multi-domain expansion is corpus work, not architectural work. The supporting evidence was: - Cube enabled multi-domain demonstration - Behavioral signals (14) fire consistently across domains per experiment - Drivers, value tensions, sync mechanic, agent layer are domain-agnostic - Estimated effort: 4–6 weeks of corpus + taxonomy + UI work

The framing was largely absence of barriers — what doesn't need to change.

What the measurement schema exercise adds¶

The exercise produced 8 architectural findings promoted to canonical:

Three-category gap synthesis (architectural / aggregation / exposure)
Schema architecture sound + consumer discipline is the gap
Forward-applied lesson pattern
Central-extraction SPOF with substrate-side parallel
Q5-vs-principle-7 categorical distinction
Closure-paths-to-profile-shapes mapping
Closure-paths-compose-additively + Single-swing closures in mapping
Architectural independence of design dimensions (D20 + D24)

Finding 8 is the load-bearing one for ADR-074 alignment.

Architectural independence finding (D20 + D24 generalization)¶

The schema's design dimensions — closure paths, measured outcomes, signals, conventions — are mutually independent. Specifically:

Closure paths compose additively (D20 / CAPR profile): 5 concurrent closure paths needed; no architectural conflicts arise from their composition.
Measured outcomes compose additively (D24 / MAX profile): 6+ M-outcomes firing high simultaneously without interaction effects.
Conventions are orthogonal (Convention 6 + Convention 7 + SPOF dependency-graph): three orthogonal reliability dimensions covering sample size, staleness, and signal quality independently.

Result: the schema's primitives, conventions, and closures are orthogonal design dimensions.

How this strengthens ADR-074 ¶

ADR-074's "no rebuild needed" claim now has stronger architectural footing.

Original framing: absence of barriers (nothing fundamentally blocks multi-domain).

Strengthened framing: positive architectural property (the schema's design dimensions are orthogonal; adding new substrate axes, new domain coverage, or new signals doesn't require rebalancing existing dimensions).

Concretely, three implications:

(a) Independent implementation scheduling¶

The 5 multi-swing closure paths can be implemented in any order, in parallel, or sequentially without coordination. Q4a (mode-stratification) ships independently of Q4d (cube-C2 stratification). OQ-1 product decision doesn't block OQ-2 closure. This means the multi-domain expansion can ramp closures in any priority order without architectural debt.

(b) Adding new substrate is additive¶

If the cube adds new axes (e.g., cube-C3 for future content dimensions), Q4 stratification gains a new sub-task (Q4e) without forcing changes to existing Q4a–Q4d. The cube-doc-side independence parallels the schema-doc-side independence.

(c) Domain-specific signal calibration is local¶

When ADR-074's corpus expansion lands and signal recalibration is needed, calibration applies to specific signals (E5, E7, E_classify_driver) without cascading through unrelated signals. The SPOF dependency-graph (architectural finding 4) names which signals share a classifier risk; signals outside that dependency graph are insulated from recalibration ripples.

Implications for light-paper positioning¶

ADR-074 currently supports a multi-domain positioning claim. The architectural independence finding upgrades that claim from defensive ("nothing blocks us") to confident ("our architecture is composable").

Defensive framing: "Adding new domains to Pulse doesn't require architectural rebuild — most of what's needed is corpus expansion + small taxonomy work."

Confident framing: "Pulse's measurement architecture is composable across domains. Closure paths, measured outcomes, and conventions are mutually independent design dimensions. New domains add new substrate; the measurement schema accommodates without re-balancing."

The confident framing is supportable by the 34-decomposition exercise and the 8 promoted architectural findings. It's a stronger claim for external positioning.

Recommended follow-up¶

Three options, not mutually exclusive:

Option A — Lightweight: cite this brief in ADR-074 ¶

Add a "Follow-up findings" subsection to ADR-074 referencing this brief. Preserves ADR-074 as written; surfaces the strengthened claim for future readers.

Option B — Medium: ADR-074 addendum¶

Write a separate ADR (e.g., ADR-076 — Measurement schema architectural independence). Captures the finding as a first-class architectural decision rather than a follow-up note.

Option C — Heavyweight: ADR-074 revision¶

Revise ADR-074 to incorporate the strengthened framing directly. Higher cost, higher coherence between ADR-074 and downstream work that cites it.

Recommendation: Option A is cheapest and preserves history. Option B is right if the architectural independence finding becomes load-bearing for other work (e.g., agent-builder pitch, light-paper, multi-domain expansion roadmap). Option C is overkill unless ADR-074 is being actively revised for other reasons.

What this brief is NOT¶

Not an ADR — ADR decisions on follow-up format are CoachJ's call.
Not a light-paper draft — positioning copy work is separate.
Not an extension of the schema doc — schema findings are already in canonical.
Not an audit of ADR-074's other claims — focuses narrowly on the architectural independence implication.