Executable Certification: Verification Code and Machine-Checked Proof as Accretive Layers in Orthogonal Review
DOI:
https://doi.org/10.68366/g7xtyx81Keywords:
formal verification, Lean 4, orthogonal review, reproducibility, falsifiability, machine-checked proof, Chrysene FormalismAbstract
A previous paper in this journal established orthogonal review: a procedure in which review stages are assigned deliberately different validity questions rather than collapsed into a single judgment. This editorial announces an accretive extension of that procedure. Two artifacts are now published beside qualifying papers — a verification-code dossier, in which declared quantitative claims within scope are recomputed by a program that asserts rather than prints, and a formal-proof dossier, in which declared discrete content of the paper is machine-checked by the Lean 4 kernel. The methodological claim is that these layers supply a second kind of orthogonality. The three review disciplines of the earlier architecture are independent in the question they ask but identical in how they fail, namely by forming a wrong opinion; a verification harness and a proof kernel have no opinion to be wrong, and signal detected disagreement by a failed check. Their reliability remains conditional on correct encodings, adequate checks, and the trusted implementation. The two new layers are partial rather than total predicates on a manuscript, and their domains are fixed by status labels together with the declared checking scope; we state those domains precisely and show that neither contains the other. We report the defects the layers have found in two published papers and in one unpublished dossier — the last caught in pre-production by an independent checking system — and we grade those findings, since several are conservative and one was not. We also report one false positive generated by the review layer itself. For a journal that applies mathematical rigour to questions such as biogenesis, where the experimental loop is slow and in places not yet runnable, machine-checked internal consistency is the external constraint available before experiment, and its principal value is that it converts a falsification register into quantities an experiment can measure. The layers do not establish empirical chemical claims; numerical checks reach declared real-valued quantities, while the present Core Lean tier is limited to its stated discrete domain. They do not substitute for adversarial review; what they replace is testimony, by an artifact the reader can execute.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Charles D. Schaper, Ph.D.

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.