REVIEWER_GUIDE.md

# Expert review route

The strongest question is whether a restricted functional equivalence class removes a genuine post-conversion reachability obstruction while paying for metrology, repair and retained access.

| Reviewer focus | Start | Critical challenge |
|---|---|---|
| Theorem correctness | THEOREM_INDEX.md, manuscript sections 6A.2–6A.8 | Check exact interval intersection, cost monotonicity and every robust-margin premise |
| Prior art and novelty | SOURCE_AUDIT.md, CLAIMS.json | Identify earlier synthesis of scale selection, reserve limits and access-dependent maturation |
| Statistics and controls | RESULTS.md, DATA_DICTIONARY.md, results/gauge/runs.json | Preserve pairing; explain matched-control ties and differential-bias failures |
| Numerical implementation | src/gauge_contracts.py, tests/test_gauge.py | Challenge grounded-response solves, confidence budgets and finite precision |
| Molecular/material feasibility | manuscript sections 11–12, EXPERIMENT_PROTOCOL.md | Find a conversion actuator with measurable positive bounded increments and stable contacts |
| Scaling and transport | manuscript sections 10 and 16 | Include acquisition, local inference, feedstock flux, reserve space and closure obligations |

The quickest falsification attempt is to construct an explicit case satisfying the exact stated G2 or G3 premises but violating its conclusion. A failed chemical premise instead rejects that backend. A controller with extra independent calibration information is not a counterexample to the identifiability no-go result.

For a numerical issue, report the claim label, version or commit, seed, method, dimension, parameters, expected result and actual result. Independent replication should keep the original data intact and write new output in a separate working copy.