Language as Substrate: AI as Informant for a Regime Account of Language

Language is often treated as grounded in a subject—cognitive, social, or intentional—but recent large language models force a reconsideration of that grounding assumption. In this paper, I develop the inverse claim as Language as Substrate: language is a structurally prior field of differential, rule-governed operations—substitution, compositional assembly, iterability, and constraint satisfaction—through which subject-positions and “voice” become legible as contingent effects of traversal. Using Lacan’s triad (Symbolic/Imaginary/Real) as a diagnostic vocabulary for mapping rule-binding, centering, strain, and repair, I introduce a replicable set of prompt-based probes administered across multiple publicly accessible systems and read the resulting outputs as discourse artifacts produced under identifiable constraint regimes. The findings show strong convergence in explicit rule-binding and invariance under local rule fields, alongside patterned stabilization moves under limit conditions (bracketing, meta-framing, scaffolded breakdown), supporting a regime account in which substrate-level symbolic operations remain robust while centering and repair reorganize with constraint profiles.