We introduce Ω_KEKS, an algorithmically defined constant measuring the probability that randomly generated formal statements cross the absurdity threshold—the point where attempted seriousness becomes indistinguishable from parody. Building on Chaitin's Ω (the halting probability), we prove that Ω_KEKS is (i) uncomputable via reduction from the halting problem, (ii) exhibits a phase transition governed by absurdity threshold τ, and (iii) admits empirical estimation via Monte Carlo methods. We establish Ω_KEKS ≈ 0.08 ± 0.03 for typical academic discourse, implying approximately 10% of formal statements inadvertently cross into self-parody. The paper includes formal proofs, empirical validation on real corpora (ArXiv papers, legal documents, LLM-generated text), and demonstrates its own thesis through self-referential analysis. This work connects algorithmic information theory to discourse analysis, providing a rigorous framework for understanding when formalism becomes self-defeating. Keywords: algorithmic information theory, Chaitin's Omega, halting problem, computability theory, phase transitions, academic discourse, self-reference, absurdity detection