The Mathematics of Myth: Information Entropy, Civilizational Self-Correction, and the Defusable Detonator.

A prior work formalized the degradation of institutional knowledge and the resulting “coherence suppression” in generative neural networks as thermodynamic and structural inevitabilities. That work closed by mandating external corrective work—“primary source re-entry”—as the only force capable of rupturing a crystallized epistemic state.

This paper formalizes the question that mandate leaves open: the agent capable of performing that external work—the corrective competence of a knowledge-bearing system—is itself subject to entropic decay, and is eroded by the very adoption of the generative tools that make it necessary. We model a knowledge-bearing civilization as a system whose persistence depends on a single negentropic term: its capacity to register the resistance of reality and correct against it.

Using a nonlinear dynamical model, we show that collapse is not inevitable but occurs as a saddle-node bifurcation: an abrupt, hysteretic, and hence irreversible transition. We then derive the observable signature of such a transition: the decoupling of a transmitted record from its referent, rendering it undecidable absent an independent channel. This is the mathematics of myth, yielding three regimes:

• The recoverable (Troy)
• The present-but-illegible (Linear A)
• The undecidable (Atlantis)

Finally, we situate generative artificial intelligence as a pharmakon: a detonator that can be aimed toward truth (primary re-entry) or toward the myth-attractor. The deepest practical risk is not that reality is refuted but that it is rendered structurally illegible: a record left wholly intact yet unreadable, because what a collapsing civilization loses first is not its archive but the competence to read it.

Access Note: Due to the sensitive nature of the formal apparatus and the ongoing phase of empirical validation, the full text is available upon request. Researchers interested in the derivation of the saddle-node bifurcation or the specific parameters for the epistemic regimes are invited to request access.