[petalas]

Summary

augmented factor/target matroid has a characteristic-two representation
  -> preserve every old query and reader zero
  -> exact target survives over characteristic two
  -> actual complete-plane reader bill >= sum of target group sizes

Add a self-contained sharp floor for arbitrary finite target/term counts and external directions on both factors. A cheaper realization must fail the representation hypothesis on both sides. Full representability is sufficient for this proof, not necessary for selective support transfer. No supplier, new kappa or selected-frontier change is claimed.

The explicit rank-five, four-target, eleven-query near-regular family has a nonbinary U2,4 restriction and still transfers to GF4. This goes beyond PR3491's TU/target-span scope and PR3489's four-term pair scope. The separate non-Fano five-term investigation remains distinct.

Evidence

• Before: prior sufficient conditions restrict dictionaries or target/term counts. After: PROOF.md supplies the general representation implication and its transpose corollary.
• Two separate implementations agree on all 15,503 exact polynomial minors, 32,768 subset ranks and 16,384 source/reader requests. Six rational/GF4 operator pairs pass 76,800 entries and both support inclusions. Both checkers verify expanded I224. Three behavioral groups include six real optimized CLI corruptions.
• Registered CPython 3.12 replay and strict Ruff checks pass. At actual head c00f31594402e1769775a4fea2819ed9187cf9ee, the committed-byte/source-history gate covers twelve package/input files. Official export/check and all three Git-free scientific commands pass; all thirteen export files retain identical bytes, modes and inodes.

python3.12 -I -Werror -B scripts/research_package.py --package research/p112-matroid-reader-transfer

Credit Tekalign/PR3370, Philippe Van Velzen, Arham Tahir/PR3345, DreamingOfClouds/PR2111, the immediate query/reader predecessors and their retained P112/native/supplier lineage. Near-regular specialization is standard; its elementary use is reproved. Exact sources, comparison-only receipts, replay, limits and next questions are in the package. Independent human and formal proof-kernel review are unclaimed.

Merge Danger

Door: Two-way. One isolated research package.

Blast Radius: Research. Production tools, frozen sources and the selected record are unchanged.