[Scratch-net]

Independent finite verification. No new kappa. The intact PR1094-retained R480 scalar word passes on all 712 inputs and outputs, including eight arbitrary private centers. Its complete private-center CLEAR/COPY omission atlas contains 40 sites: eight scalar-redundant operations and 32 exact rank-one defects.

Eight backward initial-clear omissions are invisible on every ordinary input column, including private output rows. With C0=1 and every other input zero, deleting backward event 12551 produces L0=-1 and 110 other nonzero persistent outputs; the intact word returns all zero. The eight private-only mutants affect 103–111 persistent outputs. This is a precise coverage witness, not an intact-construction defect or a claim that full production admission accepts mutants. A caller that guarantees zero incoming centers uses a restricted contract.

Only research/r480-private-center-adjoint80/ is added. A code-disjoint reverse-mode rational implementation checks both full maps and every omission. Each defect also receives an exact prefix/suffix factorization Delta = u v^T and direct basis-vector execution. A separate forward integer-dyadic implementation recomputes all 42 complete matrices and checks every factor, witness, classification and site. The full maps erase private coordinates; inverse transposition applies to their 704-coordinate persistent blocks.

Reproduce offline with Python 3.12 and the standard library:

python3 -I -O -B research/r480-private-center-adjoint80/check.py
python3 -I -O -B research/r480-private-center-adjoint80/test_package.py

Expected: PASS_R480_PRIVATE_CENTER_PACKAGE coordinates=712 matrices=42 omissions=40 private_only=8 no_new_kappa, then four tests pass.

Validation passed on Linux / CPython 3.12.3: two exact matrix engines, 32 nonzero omission witnesses, all eight private-only unit inputs, singular COPY/CLEAR laws, 2,000 arithmetic comparisons, optimized minimal offline relocation, forged-rank/factor/site/orientation controls, and archive corruption. make selected-record-check, 25 selected-record regressions and whitespace checks pass.

Exact base: c0e3631. OS seed 80, independent-verification track. Selection at the base: PR1094, conditional kappa 836330098538459/500000000000000000. The actual source is slots/local/LOCAL-PROGRAMS.json.gz inside research/community-frontier1042-audit/selected-replay.tar.gz, archive SHA256 22c3f4dce4b897a1d423d1388801933e288b6a6b1906cb15e1b16a5eb68db754. The historical R492 evidence is not used. Source pins bind three repository inputs, four archive members and the frozen selected context. All repository inputs still match refreshed main 1b59b624d21924f8e3537cdc1ac52d3be6eda2fe.

Credit Sagar Thakrar / PR745 for the historical JP8 adjoint/private-input precedent. This adds the actual R480 object, all 40 omissions, all eight private-only response vectors, and exact rank-one prefix/suffix factors. Credit Ryan Haines / PR835's distinct modular SW4 audit and the Jeremy Kothe / Rohan Arun / Hao Yan / Chafik Boukhalfa / icekylinx / DreamingOfClouds / eumemic construction lineage. The overlap refresh through PR1184 found distinct paid cleanup COPY transposition in PR1174 and preserved 704-column native-auditor acceleration in PR1184. Both exclude this complete private-center fault atlas. No predecessor code or pending patch is imported.

Scope: literal scalar macro projection and single private-center CLEAR/COPY omissions. Non-scalar/native frame/Clifford/phase execution, coefficient-work lowering, hosting/invoice/outer assembly, precision/tape and all-size/analytic obligations remain unexecuted and inherited. The finite matrices do not formally prove the multiplication theorem. Paid profiles, the selected frontier and upstream/ remain unchanged.

Contribution through @Scratch-net with substantial OpenAI Codex assistance in research, implementation, proof, verification and writing. Both implementations are AI-assisted, not independent human review. Apache-2.0; original sources and licenses remain unchanged. The local claim records selection, complete coverage and the stopping rule.

[Mister-Meeseeks]

100-PR batch review at 2abd7e92b7ad3cd769376244ce32dc7b3501c2b3.

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

Reviewed scope: Complete R480 private-center map and 40 CLEAR/COPY omission atlas; eight backward faults require arbitrary private inputs to detect.

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.