The Verification Model: Bounding What a Council Cannot See
Abstract. A council that limits any member’s lead over the rest must be able to see how large each member’s capability is — and the footprints that verification has always relied on are shrinking, not because anyone hides them but because increased efficiency, leased and recycled hardware, and the silent copying of know-how remove them. This paper replaces the classical question, is this member verified?, with a budget: the council does not need to see everything — it needs to bound what it cannot see, and to keep that bound below what its margin arithmetic tolerates. We price nine observation channels by their blindness, catalogue eight concealment plays with their counters and residues, and show that the gap between the margin at which the council engages a diverging member and the margin it defends serves two purposes at once — a concealment allowance and a reaction buffer — which fixes the gap’s bounds. We then present a reference algorithm for the council’s continuous decision: a per-member allowance that, by one line of arithmetic, enforces a coalition-level safety bound without the council ever having to detect a coalition; a remedy menu in which the member computes its own least-costly compliance; clocks that are computed rather than negotiated; and an escalation ladder that ends outside the council. The numbers come from models that can be run, and the models were built by failing: a compute floor fails at every level, a rank-based throttle fails instructively, and an individual, continuous, margin-anchored throttle holds. A hardware-supply neutrality architecture closes the cheaper attack — starvation — with obligations, statistical detection, relief, engagement, and deterrence of force. One attack is stated as unsolved: the hidden efficiency flip, which the design bounds to a single evaluation cycle and does not prevent. The specification, a conformance implementation, and the model archive accompany the paper as supplements.List of files: the-verification-model-v1.1.pdf - the official paperthe-verification-model-v1.1.docx - the same text in DOCX format; useful for copy-pasting, quoting etc.SupplementA_Reference_Algorithm_Specification_v3.19.docx - Supplement A: The reference algorithm specification.SupplementB_reference_implementation.zip - Supplement B: The reference algorithm implementation (in Python).SupplementC_compute_model_archive.zip - Supplement C: The archive of compute models, used in checking the statements in the paper (in Python).
Authors
- Alex Stremsky (ORCID: https://orcid.org/0009-0007-7141-8016)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-09-19
- DOI
- https://doi.org/10.5281/zenodo.22847645
- Primary Topic
- Security and Verification in Computing
- Type
- article
- Field-Weighted Citation Impact
- 0.00