Contemporary AI systems produce coherent, context-sensitive language without biography, development, embodiment, or ego formation. This empirical fact destabilizes a presupposition shared across many theories of language: that linguistic structure is fundamentally grounded in subjects, whether as cognitive agents, social participants, or intentional speakers. I develop the inverse possibility. I argue that language is a structurally prior substrate of differential, rule-governed operations through which subject-positions are generated as contingent effects. Language supplies the differential form—substitution, recursion, constraint satisfaction, compositionality—through which tokens become positionally meaningful; agents activate and traverse this field in specific situations; constraint regimes shape how traversal appears at the surface. To investigate this claim in a way suited to a theory-forward social science paper, I introduce a replicable protocol of prompt-based probes and administer it across five publicly accessible AI systems (OpenAI, Claude, DeepSeek, Yandex, and Google/Gemini). The probes are organized into three families—Symbolic, Imaginary, and Real—using Lacan’s triad as a diagnostic instrument for mapping stabilization, centering, strain, and repair within symbolic systems. Psychoanalytic case literature enters as an archive of language under pressure that helps reverse-engineer stress points in symbolization. Across systems, outputs converge strongly on explicit rule-binding and procedure preservation under local rule fields, while limit conditions elicit patterned coherence repair and meta-discursive stabilization. These findings support the substrate model of language and clarify three regimes of traversal—human–human, AI–human, and AI–AI—each coupled to distinct constraint hierarchies and discourse economies.
Kambiz Behi (Sun,) studied this question.