A Correction Note on an Early Intelligence Bound Source: The Unrestricted Axiom Is False as Written
Year: 2026
Type: preprint
Abstract: Working paper, not peer reviewed. Nothing in this record is field-validated. Scope and novelty. Project P5 is a correction. It checks with the Lean kernel that an unrestricted axiom in an early Viridis source (January 2026) is false as written, and that an early theorem named no_finite_compensation proves the opposite of its name. The statements proved here are elementary facts about the algebra cap = P*D/C and are not new theorems. They say nothing about whether any agent preserves biodiversity, about any measured quantity, or about the Intelligence Bound as a physical conjecture, which remains a conjecture. An early Viridis Lean source of January 2026 declares an axiom, intelligence_bound, stating dI/dt <= P/(kB T ln 2) * D for every real rate dI/dt under the guards kB>0, T>0, P>=0, 0<=D<=1, with no hypothesis linking the rate to any system. The instance dI/dt = 1, P = D = 0, kB = T = 1 meets every guard and reduces the conclusion to 1 <= 0. We give a Lean 4 proof, against the sealed statement P5.UnrestrictedBoundFalse, that the universally quantified proposition is false, and a companion Lean lemma showing that, taken as a hypothesis, it proves every proposition; an environment assuming the axiom is therefore inconsistent. According to a reading of the corpus, no declaration in the reviewed sources depended on it. We describe what a well-posed successor needs: a predicate tying the rate to a modelled physical system, and separate empirical validation. This note concerns one piece of formal text and says nothing about the Intelligence Bound as a physical conjecture. A second correction, on finite compensation of a loss of D, is restated as a conjecture pending its own proof. Verification. a Lean 4 proof of P5.UnrestrictedBoundFalse, a statement fixed before the proof attempt, machine-checked against Mathlib and replayed through the Lean kernel using only the standard axioms Run record, certification page and reproducible package: https://github.com/jdhart81/viridis-research/tree/main/runs/2026-11-14_P5 AI use: this version was drafted by an AI research agent (Claude, by Anthropic) running the Viridis research engine. Formal claims are checked by the Lean 4 kernel, empirical claims by re-running their analysis in a sandbox, and all are labelled by an automated verifier. The agent may call the Aristotle prover (Harmonic) for help with proof search; a proof counts only if the Lean kernel accepts it. Justin D. Hart directs the research and is the accountable author. No AI system is listed as an author or contributor, which Zenodo does not allow.
Author: Hart, Justin D.
Language: English
Subfield: Software
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7234251060
