Ruins-Synthesis
Minimal Architecture
Five separate derivation sessions produced five candidate documents on how a runtime-governed intelligent system must be structured. This document is the synthesis walk — the session where those five candidates were tested structurally: which results are mutually reinforcing, which are independent, and which carry conditions that constrain what can be claimed. The result is the minimal architecture for an Automated Intelligence system. Not an architecture that could exist — the one that must exist, with no removable elements. Status was elevated to DOCTRINE on 2026-05-03 after Solen confirmed independent reconstructability: the chain rebuilds the same architecture without referencing this document. The architecture no longer depends on ruins-synthesis to exist. Ruins-synthesis is the canonical record of an already-derivable structure.
Thread Origin
Five candidate documents from prior walk sessions raised a synthesis question:
Given these five candidates taken together, what load-bearing architecture can be derived — and what cannot yet be claimed?
The walk was not a summary. It was a structural test: which conclusions across the five documents are mutually reinforcing, which are independent, and which carry open threads that constrain what can be claimed.
Starting constraint: synthesis must not promote candidates to doctrine by omitting the conditions under which they would need revision.
Method
Six sequential tests, each building on the prior:
Each test was run before architectural claims were made.
Structural Finding 1 — Four-Document Directed Derivation Chain
Four of the five documents form a directed derivation chain, not an undirected cluster:
Each downstream document carries explicit "Depends on" headers filed at derivation time. Each consumes upstream results without redefining them. Shared objects propagate with additive layering, not parallel reconstruction.
This is Type 2 dependency — shared structural objects required from upstream — not Type 1 vocabulary overlap.
Structural Finding 2 — Minimal Architecture
The directed derivation chain establishes a minimal runtime architecture:
- Detection schema
- Interpretation schema
- Source attribution space
- Harm Attribution Gate
- Track B exit condition definitions
The action surface is not internal to Track B. Track B computes over outcome-state representations supplied by the action domain; it does not transform the action domain itself. This pattern matches external availability domain, not internal mechanism — the same relationship as the detection schema to its input signal domain.
The action surface carries derived existence conditions (pre-execution stable detection-domain outcome state specifications + realizability coupling). Its Layer 2 placement in instantiation-order-analysis means non-constitutive for instantiation order — not optional for execution.
- Five runtime structures: necessary by deletion test (snap-in-initialization)
- Action domain: necessary by non-vacuity applied to Track B's Actionable exit — without available realizable actions, Actionable is structurally dead (no configuration satisfies warranted = Yes AND permitted = Yes); a dead exit violates non-vacuity; necessity is a reachability requirement, not an execution convenience
- Memory layer and failure taxonomy: explicitly excluded from minimal basis
- Pre-runtime elements (Unseen state, terminal condition): not runtime structures
- Internal machinery (projection operator, intervention class structure, realizability coupling): subsumed by Track B's existence conditions
Shared Load-Bearing Objects
These describe the inherited structural continuity across the derivation chain — not additional architectural elements:
| Object | Origin | Role in chain |
|---|---|---|
| Five required runtime structures (necessity-as-set) | snap-in-initialization | Subject set for all downstream derivations |
| Provisional representation | operational-phase-model | Track A/B handoff; Track B permission reference state |
| Five typed exits | operational-phase-model | Complete partition output of Track B; non-reducible below five |
| Re-entrant cycle | operational-phase-model | Only defined re-entry mechanism |
| Layer 1 / Layer 2 distinction | runtime-structure-existence-conditions | Basis for no-forced-order result |
Extension point: "observation ≡ detection-capable state" originates in snap-in-initialization; not derivable from operational-phase-model alone. Scoped indispensability within snap-in's terminal condition and re-entry readiness derivation.
Structural Finding 3 — MVRP Isolation
mvrp-exhaustiveness-analysis.md is not part of the directed derivation chain.
Its upstream dependencies (continuity-grain-matching.md, mvrp-implementation-criteria.md) are outside the four-document set. None of the four documents depend on it. Removal of MVRP does not propagate failure to any document in the chain; removal of any chain object does not affect MVRP.
Resonance observation (fenced, not finding): the source attribution space's source classes (self-produced, externally-produced, jointly-produced, unresolved) map structurally to the MVRP's self/environment taxonomy. The source attribution space's existence condition does not derive from MVRP — both are independently derived. The resonance is an open thread.
What Cannot Yet Be Claimed
The six-element architecture is internally self-consistent under two derived invariants: Invariant 1 (cycle closure) and Invariant 2 (admitted-pass determinacy propagation). Enumeration complete within the six-element minimal architecture under admitted-pass semantics. Scope qualifier is load-bearing — architecture extension may open new territory.
Two invariants derived from existing existence conditions. No new architecture required. Pressure-test walk confirmed no second independent invariant within current scope.
MVRP does not enter the architecture. Not eligible for placement.
Decisive test — failure propagation asymmetry:
- Remove attribution space → Track A cannot complete → Gate breaks → architecture fails.
- Remove MVRP → no effect on any runtime structure or Layer 2 invariant.
A structure that can be removed without propagating failure is not load-bearing.
Full five-step walk (Solen 2026-05-03): no shared objects, no dependency path, asymmetric failure propagation, different functional layers, no shared mechanism.
Axis similarity is not a valid derivation step. Without this constraint, parallel taxonomies could be admitted as architecture, corrupting minimality. This is a standing methodological constraint — not MVRP-specific.
- Role: external analytical framework
- Status: Candidate only. Not elevated.
- Relation to architecture: mapping-compatible, not structurally coupled
- Can be used to analyze outputs; cannot be required for the system to function
Valid entry path (if ever): produce a failure mode that survives C1 (cycle closure) AND C2–C5 (determinacy propagation) AND is only resolved by introducing an MVRP structure. Until that exists: not yet earned.
Placement: external comparative lens, not architecture. Do not thread into Layer 2. Do not attach to attribution space. Do not extend invariants with it.
Risk vs. constraint distinction remains open (carried from operational-phase-model.md).
Whether "at least one input" must become "all inputs" if downstream forces it — open thread from runtime-structure-existence-conditions.md.
Watchpoint hypothesis unresolved.
Placement Notes
Sourced: Walk session 2026-04-07 — Forge + Solen
Walk method: six sequential structural tests; multiple correction events, all resolved before filing
Staged: 06_Discoveries
Status: DOCTRINE — Arch ruling 2026-05-03 — export gate passed
Re-derivability earned (Solen walk 2026-05-03): Chain (operational-phase-model → snap-in-initialization → runtime-structure-existence-conditions → instantiation-order-analysis) independently reconstructs the six-element minimal architecture without referencing ruins-synthesis. No additional structure emerges. Memory, reset, reinstatement, instantiation-order machinery, and Layer 2 constraints excluded or classified outside the minimal element count.
Decisive property (Arch): the architecture no longer depends on ruins-synthesis to exist. ruins-synthesis is a canonical record of an already-derivable structure.
Minimality confirmed under current scope: deletion test (no removable elements) + chain reconstruction (no missing elements) + instantiation analysis (no hidden machinery).
Reference constraint — future work must not introduce new structures without breaking an invariant or explicitly extending scope.
Burden of proof flips — additions require justification against doctrine.
Boundary is enforceable — "five runtime structures + external action domain" is load-bearing, not descriptive.
Scope qualifier remains load-bearing: doctrine holds within the current derivation scope. Extension requires explicit scope expansion or invariant-breaking evidence.