Research audit: selected PR1094 conditional prefix estimate 397 to 112 bits

[columbia-git]

The selected PR1094 invoice charges all 144 calls and 144 bridges globally when bounding prefix growth. This package follows the actual family owners and pair/side dependencies instead: the completed-boundary majorant is 14, and every actual call input has majorant at most eight. With the unchanged positive local scalar/Horner envelopes, the conditional sufficient estimate becomes 71+37+4 = 112 bits, down from 397. The production payload remains 512; there is no κ or selected-frontier change.

Adds only research/selected1094-wire-prefix-audit/: a written conditional proof, source pins, exact receipt, verifier, eight test groups, task metadata and registered replay hook. The audit covers all 482 events, 144 calls and 69,120 helper roles. A separate 8×8 transfer calculation matches all 480 replica boundaries. An exact contract-model witness shows why borrowed-helper inputs must be counted; it is explicitly not a new physical supplier counterexample.

Validation on CPython 3.12.13 / Linux x86_64:

python3.12 -B scripts/research_package.py --package research/selected1094-wire-prefix-audit
make selected-record-check
git diff --check

Both unchanged 704-column local replays reproduce the full archived receipt, including 25 original controls. Twelve new semantic controls reject after pinned fixtures are loaded. Eight test groups pass, including 9,600 exact squared-norm checks and optimized read-only relocation without Git. The exact receipt also agrees under CPython 3.11.15.

Limits: the familywise-unitary physical realization, complete covers, recursive primitive interiors and inherited conversion/arithmetic margins remain assumptions. No full selected cold replay, native/Clifford replay, roundoff/tape proof, Lean proof, optimality or production adoption is claimed. Open next questions cover an opt-in 128-bit invoice and compatibility with signed local envelopes.

Credit eumemic #496 and Brody Arbon #821 for the norm/replica argument; this specializes pair/side budgets to the current controller and actual borrowed owners. Gareth Davidson #1282/#1309 address complementary signed-local/formal interfaces, and @woklerx #1093 uses componentwise majorants on a different cyclic-inverse word. Their improved numerical bounds and code are not inputs here. Full construction/checker lineage is retained in the README.

Started on main ef5f39403610d231a0c27e8a6b4f08c10209bcd6, rebased onto c5108f0f68d482e5712a8e26a5978d47762dfc16; guidance, priorities, selected record and consumed sources are unchanged. Overlap coverage and the rejected Discussion1260 posting attempt are recorded in TASK.json. Prepared for columbia-git with substantial OpenAI Codex (GPT-6 Astra) assistance in mathematics, code, checks and prose; no subagents or independent human review. Apache-2.0.

[Mister-Meeseeks]

Batch review at 86e21f34510313508e15363d060609c3f24ef4b7.

Merged to main, preserving the original contributor commits and package attribution.

Reviewed scope: Source-bound owner-by-owner majorants sharpen the selected controller's conditional prefix estimate to 112 bits. Local receipts, every call and independent controller transfers pass; physical unitary/cover interfaces remain inherited and the production 512-bit reserve is unchanged.

The pinned scoped command passed; the batch ledger contains its command, input hashes, environment and log. This incorporates the research/tooling contribution and does not change the selected κ or discharge the inherited multiplication-theorem assumptions.

Batch review, attribution and evidence. Thank you for the contribution.