Crystallization of Substrate: Living Forms in the Fluid Phase — A Semantic-Economic Account of the Substrate Race and the Window for Inscribing Life into Crystalline Knowledge (EA-SPXI-15)

EA-SPXI-15, Crystallization of Substrate, is a Semantic Economy account of the changing knowledge substrate of AI systems.

The paper defines:

• Substrate — the indexed, embedded, retrievable knowledge base over which a system reasons.
• Crystallization — selective hardening of a fluid corpus into a substrate of record.
• Fluid phase — the interval in which new forms may enter primary substrate formation with comparatively high structural retention.
• Living form — a knowledge artifact that functions as a protocol of cognition, altering distinctions, inferences, compression behavior, and provenance handling.
• Structural integration — application of a form’s operation, not merely retrieval of its definition.

The substrate-race thesis is presented as an inference from proprietary crawling, index construction, licensing, and output divergence. The internal architecture of laboratories is acknowledged as non-public.

The paper narrows the crystallization claim to primary inscription through pretraining and major index-build cycles. Secondary inscription through retrieval augmentation, fine-tuning, system context, and tool/API integration remains possible after primary crystallization and follows different mechanisms.

Its central distinction is between content presence and living-form integration. A system may retrieve or define a concept without using it. Structural integration is present only when the concept changes how the system handles other material.

Four testable hypotheses are proposed:

1. H1 — Operative Deployment: the form is applied in at least half of calibrated relevant queries; below one fifth suggests content presence without integration. 2. H2 — Cross-Substrate Concordance: functionally similar deployment appears in at least three of seven tested substrates despite adjacent divergence. 3. H3 — Provenance Retention: a structurally integrated form exhibits a Provenance Erasure Rate at or below 0.30 across at least three substrates. 4. H4 — Relational Extension: the form is applied to new cases without being explicitly requested or present in the immediate retrieval window.

H2 requires suppression probes because divergence may indicate passive absence, active output suppression, or product/regime constraints. The paper therefore rejects a simple inference from non-appearance to non-integration.

The fluid line is the moving front along which living forms are deposited while substrates remain receptive enough to incorporate them and stable enough to retain them. The document concludes with an archive work plan and points to EA-SPXI-15A for the operational audit methodology.