[dillona]

A fully charged residual helper split raises the conditional claim to κ = 0.003639042931507, an increase of 0.000000100498024 over #7117. On that exact native word, one paid rank-five MOVE is subdivided into ranks four and one, and two rank-one exits are retimed. This permits another independent helper9 target to become helper8+helper1. The signed 17-stage caller is retained and freshly checked.

The full invoice changes by ΔP={1:+17,4:+17,5:−17}, ΔT={1:+17,8:+17,9:−17}. It has 110,689 paid positions, 22,223 productive fields and surplus 1,424, with no empty credit. Source-unit ordinary work is 777,622,672 <10⁹. Fresh precision raises c from 2336 to 2338. Native scale 1125/F24 and caller scale 675 are derived separately and charged.

• Exact source: Conditional κ = .003638942433483 (+5.6433774e-8 over #7116): paid helper meeting #7117 092255ac8e9ed07ea59fee87c9711a5d72b10750; its full 25-file export and nested predecessor sources are frozen and authenticated. Every source native byte is reconstructed before the edit.
• Source cut 45209/chart 246/pivot 236. Role 789 pays the subdivision; roles 472 and 1359 retime neutrally. Added 24 ADDs + 4 SCALEs + 1 net MOVE. All 54 owned lifetimes, peak 43 and 23 rises of mass 56 remain unchanged.
• New native SHA256: cbce61015e22007f785474e2af2462a7d97ead1e105d71cd2d9cbc0a449580f3; 73,628 events, 1,487 public fields and 1,087 independent helpers. The caller is byte-identical to Rohan Arun's Conditional κ = .00363797977 via a 17-stage signed caller #7113.
• Complete forward/adjoint endpoints, 16 native corruption controls, all 512 columns of four affected chart maps, full finite physical native/signed routes, and the fresh complete 17-stage application pass. All 3,048 paid/helper maps and 104 main boundaries receive signed-Pauli/absolute-phase checks. Owner, grids, exact prices, ordinary work, precision, fixed-tape costs and 54+4 outer gates are included.
• Both exact engines reject Conditional κ = .003638942433483 (+5.6433774e-8 over #7116): paid helper meeting #7117's invoice at the new claim. Independent raw bracket: (0.00363904293150729230234892845394584,0.00363904293150729230234892845394585). This is the emitted invoice's ceiling. The retained BIT cap leaves 0.000127647068493 headroom; no global word/caller optimum is claimed.

python3.12 -B scripts/research_package.py --package research/p104-funded-residual-split

Both complete direct CPython 3.14.7 runs passed. The final registered isolated CPython 3.12.16 replay passed both complete commands, matching every frozen native and composition receipt field. REPLAY.txt preserves the actual output; metadata was marked completed afterwards. No scientific code or receipt was changed after the successful comparison.

Committed bytes and executable modes pass on both commits. Source-history checks pass at canonical-based scientific commit 9cdbdc1def1ee4ca94b27874b42173be0e47ebce. Public commit 3f6ff8c34b22af12b40c5a8bcee585cd1aafdb9c carries the identical 25-file package tree ece2a48ff8c2b84d90ddcbfb15dbd836984df5da. Publication is package-only atop the older account fork because the credential cannot push workflow-history changes; canonical provenance remains pinned. The scientific replay needs neither Git nor network.

Exact-head run42879681891 failed onboarding at18:56UTC (760tests,1failure,24skips), and skipped scientific replay. test_current_ci_section_matches_selected_record reports the same guide6030/selected7000 mismatch. The predecessor #7117 workflow failed the canonical onboarding test because the CI guide names #6030 while the selected record names #7000; its scientific replay was skipped. The existing correction handoff remains available. No hosted scientific pass is claimed.

PROOF.md gives the concrete arbitrary-input identity and the positive concave gain H8+H5−H9−H4. APPLICATION.md supplies the exact component argument, including 18,479 independent stage/helper inputs, 867 native banks, 42,261,280 scalar services and 25,465,969,408 elementary monomial records. The five framework assumptions remain. Independent human/formal review, maintainer admission and compiled top-level execution are not supplied. The selected frontier is unchanged.

Credit Ali Romman for helper bases and paid meetings; Jerry Chen/CubeJerry for retiming and native/compiler engines; Matarener for discovery kernels/pairing; Rohan Arun for the 17-stage caller and application; DreamingOfClouds, ToikY, Douglas Colkitt/#7000 reviewers, preceding dillona/OpenAI Codex contributions and every retained notice. ToikY's pending chart-120 rank-eight lead is distinct and unclaimed here. Substantial OpenAI Codex assistance; no subagents. Canonical investigation.