SovereigNet: Deterministic WAD-18 Energy Execution and Safety Asset Class Verification via Offline Cross-Substrate Architecture
This preprint presents SovereigNet, an architecture for deterministic energy execution and Safety Asset Class (SAC) verification across multiple computational substrates. Numbers are handled with WAD-18, an integer-exact fixed-point format at 18 decimal places. The same arithmetic and decision rules are implemented in a reference C kernel, Solidity contract logic, an independent Python model, and client-side code, and these implementations are tested against each other so that the same inputs produce the same results.The paper formalizes the WAD-18 arithmetic (addition, subtraction, floor-defined multiplication and division) and the energy decision function. It covers threshold boundary testing at ±1 atomic unit, the SAC state-transition predicate with fail-closed behavior, and the Russell Recursive Refinement (R3) method for iteratively refining the implementation while preserving its invariants.Reproducibility evidence comes from a Code Ocean capsule that reports 19 of 19 validation steps passed. The steps cover kernel fingerprint regeneration, on-chain/C-kernel equivalence, WAD-18 arithmetic equivalence, energy threshold boundaries, routing with 256-bit extremes, a dispatch benchmark, client-side parsing and ledger reconstruction, end-to-end integration, and a zero-floating-point audit. The paper scopes these results explicitly. They show the configured test suite passed. They do not establish universal correctness, deployment security, physical measurement accuracy, or financial certification. The paper ends with the further verification work those stronger claims would need.
Authors
- Michael Aaron Russell (ORCID: https://orcid.org/0009-0001-3360-6709)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-10-27
- DOI
- https://doi.org/10.5281/zenodo.23697269
- Primary Topic
- Blockchain Technology Applications and Security
- Type
- preprint