S10 MATERIAL MIDPOINT RULES — REPRODUCTION

This compact distribution accompanies section S10.7 of v1.45. It retains the
protocol, astronomical inputs, all rule definitions, the exclusion witness
and rank lower bound for every rule, full h=-7 census, and verification record.
The previously computed numerical results are unchanged.

The generated results/all_field_rule_witness_scores.npz array and the
approximately 807 MB right-ascension cache are not transport files. Regenerate
them before running the full verifier. Neither is an independent input.

Python environment used: 3.12.14, NumPy 2.3.5, SciPy 1.17.0.
Commands from this directory:

python -m pip install -r requirements.txt
python generate_family.py
OPENBLAS_NUM_THREADS=1 python support/build_cache.py --output cache/ra.npy
OPENBLAS_NUM_THREADS=1 python census.py screen --cache cache/ra.npy
OPENBLAS_NUM_THREADS=1 python census.py full --cache cache/ra.npy
OPENBLAS_NUM_THREADS=1 python verify_results.py

results/family_summary.json reports all four overlapping families.
results/rule_screening.csv retains the witness and bound for all 139 rules.
results/full_rule_062.csv contains the complete h=-7 two-grid comparison.
results/verification.json is the retained numerical execution record;
it is not a claim that this editorial revision repeated the census.

The family was formulated with the prior Cygnus results known. The protocol
is an execution specification, not a public prospective registration.
Source extracts and their earlier-version names are retained for provenance.
SHA256SUMS.txt in the package root covers these supplied files.
