Semantic Gravitational Attractors

Content-addressed systems make an important move: they replace positional identity with identity derived from a representation. When that representation is first normalized under declared rules, many variant inputs can converge upon one canonical form. This paper develops semantic gravitational attractors as a bounded architectural metaphor for the systems that deliberately create and govern that convergence. The attractor is not a physical force, a universal semantic equivalence oracle, or proof that meaning can be reduced to bytes. It is a policy-and-infrastructure arrangement: a versioned canonicalizer, a declared equivalence regime, domain-separated hashing, governed admission, lineage, constrained transformation, and evidence together route admitted representations toward stable identities. The convergence is only as sound as the distinctions preserved by the regime. The paper therefore replaces the strong claim that "equivalent meaning collapses" with a testable, regime-relative claim. It distinguishes false merges from false splits, identifies cases for which canonicalization is undefined or refused, separates byte identity from reference and standing, and states the environmental closure required for deterministic replay. It also integrates canonical identity with MCIR, GMB, the universal GSM, governed project lineage, semantic topology, binding, debt, receipts, and semantic computing infrastructure. Semantic gravitational attractors can reduce duplication, identity churn, and unsupported divergence. They can also make a mistaken equivalence more durable, erase a material distinction, or centralize an illegitimate canonicalization law. The resulting architecture is valuable precisely when it keeps those risks inspectable and revisable.