Corporeal Mimesis and Convergent Specification: How LLMs Simulate Embodiment

Large language models exhibit linguistic competences that the embodied cognition tradition predicted would require a body. Froese (2026) formulates the dilemma: ei- ther these models enact non-biological sense-making, or its long-taken hallmarks do not require it. His exit, “technologically mediated embodiment,” names the phenomenon without specifying its mechanism. This paper offers an alternative: LLMs achieve their competence through convergent specification, the fill-in-the-blanks operation ordinary communication itself requires, since meaning-making never depends on the overtly ex- pressed alone. The mechanism is mimetic, not inferential: natural language, as the tes- timonial enactment of millions of situated, embodied exchanges, constitutes an ecology of convergences whose structure of mutual indication the model exploits by converg- ing across lexical, syntactic, pragmatic, and narrative layers simultaneously. Drawing on the philosophy of science of simulation and on non-representationalist enactivism, the paper distinguishes two routes to world-involving competence, remissional instal- lation and the mimetic route. Under a non-representationalist ontology, mimesis and embodiment are engagements with the same regularities at different phenomenal res- olutions, not dual conditions. The dilemma dissolves: no non-biological sense-making is posited, no refutation of embodiment is conceded. Sense-making remains with the embodied subjects whose exchanges produced the corpus; what the model achieves is a mimesis, itself enacted and continuous with the intersubjective acts of meaning whose convergences it exploits.