The Grammar of Structural Intelligence: Core Primitives, Derived Dynamics, and Layer Routing

This paper consolidates the architecture of Structural Intelligence by organizing its expanding vocabulary into a canonical grammar of primitive terms, derived dynamics, diagnostic states, failure modes, repair modes, and domain routes. As the SI corpus has developed across AI evaluation, psyche, institutions, public life, predictive runtime, field ontology, presence, structural debt, occupancy, and repair, it has gained expressive range but also faces a new architectural risk: if every concept is activated at once, SI can become a powerful vocabulary without a stable dependency order. This paper does not add a new SI layer. It compresses and orders the existing framework so that field, difference, gradient, relation, boundary, constraint, local structuration, coherence, contact, answerability, presence, steering, debt, occupancy, and repair can be used with clearer jurisdiction. The paper proposes a dependency chain in which field precedes local structuration; local structuration precedes coherence; coherence must meet contact; contact tests answerability; answerability depends on steering and presence; refused contact becomes debt; captured steering becomes occupancy; and repair is proven only when future behavior changes. The contribution is architectural rather than empirical: it makes SI easier to teach, score, falsify, route, and revise without overactivating the full corpus in every case.