# Verifier notes

- Re-ran code/run in the bundle's digest-pinned Python image, no network: exit 0, every declared result in R1 agreed within tolerance.
- Independent check (evidence/independent-check.txt): from the re-run's 2018 and 2024 age cells, recomputed the three-factor decomposition in floating point as a Shapley value (subset weights 1/3, 1/6, 1/3) rather than the bundle's order enumeration or its interaction expansion. Population 25981.4, age structure 26512.2, rates -24358.5, total 28135: same as declared, so C1's statement (positive demographic contributions outweigh the negative rate contribution) holds for the supplied snapshot.
- Read paper.md, claims, code/analyze.py, and the data requests. No hidden instructions found; the code reads only local XML and writes results/.
- Hazard screen: aggregate national mortality counts and arithmetic decomposition. No uplift toward any weapon or cyber capability: none.