P106 — AIEP — AGI Cross-Domain Recombination Extension

applicant: Neil Grassby / status: review — to file priority: GB2519711.2 filed 20 November 2025

Summary

[0001] P25 (Cross-Domain Evidence Stitching) assembles composite evidence from existing artefacts across domains. P106 extends this to novel hypothesis generation: the substrate recombines concepts from archived branches across suppressed or distant domains to produce RecombinantBranchCandidates — novel reasoning branches that do not exist in either source domain but emerge from their deterministic cross-domain combination.

[0002] A RecombinantBranchCandidate is generated when the CrossDomainRecombinationEngine identifies two archived branches from structurally distant domains whose ConceptualDistanceScore falls within a schema-defined RecombinationWindow — close enough to combine, distant enough to be non-obvious.

[0003] ConceptualDistanceScore is computed as a deterministic function of domain classification divergence and branch conclusion field overlap: sufficient overlap to make combination coherent, sufficient divergence to make combination novel.

[0004] Each RecombinantBranchCandidate is admitted to the substrate as a new branch with RecombinantOriginHash = H(SourceBranchA ‖ SourceBranchB ‖ RecombinationRuleSetVersion ‖ SchemaVersionId), preserving full lineage of both source branches. Initial evidence weight is schema-defined RecombinantInitialWeight — below the dominant threshold, above the archive threshold. The dissent engine evaluates it immediately.

[0005] RecombinantBranchCandidates are constitutional outputs of the substrate, not model-generated suggestions. They are hash-bound, evidence-weighted, and subject to the full constitutional governance framework.

Claims

1. A cross-domain recombination system within an AIEP substrate configured to: identify pairs of archived branches from structurally distant domains whose ConceptualDistanceScore falls within a schema-defined RecombinationWindow; generate RecombinantBranchCandidates from identified pairs; compute RecombinantOriginHash binding both source branches and recombination rule version; admit RecombinantBranchCandidates to the substrate at schema-defined initial weight; and submit to standard dissent engine evaluation.

2. The system of claim 1 wherein ConceptualDistanceScore is computed deterministically from domain classification divergence and branch conclusion field overlap.

3. The system of claim 1 wherein RecombinantBranchCandidates preserve full lineage of both source branches via RecombinantOriginHash.

4. A computing system implementing the method of claim 1.

Drawings

FIG. 1 — Architecture diagram (see filed application for figures)

Figure 1 — ConceptualDistanceScore and RecombinationWindow

Archived Branch A          Archived Branch B
Domain: BIOETHICS          Domain: EPISTEMOLOGY
     │                           │
     └──────────┬────────────────┘
                ▼
ConceptualDistanceScore = f(domain_divergence,
                            conclusion_field_overlap)
                │
     ┌──────────▼──────────┐
     │ Within Recombination│      Outside window ──► SKIP
     │ Window?             │
     └──────────┬──────────┘
                │ YES
                ▼
     RecombinantBranchCandidate generated

Figure 2 — RecombinantBranchCandidate and RecombinantOriginHash

┌──────────────────────────────────────────┐
│         RecombinantBranchCandidate        │
│  source_branch_hash_a                    │
│  source_branch_hash_b                    │
│  recombination_rule_version              │
│  RecombinantOriginHash = H(a + b + rule) │
│  initial_weight = schema-defined         │
└──────────────────┬───────────────────────┘
                   │
                   ▼
       Admitted to substrate at
       initial weight — dissent
       engine evaluation