[dillona]

The funded split in #7117 moved role 231 into the rank-eight boundary group 3→246. On exact #7120, this permits a further independent helper 8→3+5 split with the complete paid profile unchanged. The new full conditional claim is κ=0.003640102187396, gain 0.000000838081114 over #7120.

At source cut 31847/chart 259/pivot 231, the exact prefix/suffix witness has a=e448ᵀ, b=e448+e1205/16−e995/16, c=e231ᵀ−4e707ᵀ, d=e449/8−e707/4. The boundary/cut bases have 7/6 instructions and add 20 ADDs + 6 SCALEs, with no added, removed or retimed MOVEs. PROOF.md gives the arbitrary-input replacement argument; APPLICATION.md supplies the complete component substitution.

The signed 17-stage deltas are ΔP={}, ΔT={3:+17,5:+17,8:−17}. The full invoice has 110,723 paid positions, 22,274 productive fields, surplus 1,424 and no empty credit. Ordinary work is 777,785,464<10⁹; fresh precision gives c=2344, height 405 and mass exponent 1172. Native 1125/F24 and caller 675 grids are independently derived and charged. Regenerated gates use gap 1/100000000000000000 and H=16487800000000000000000.

• Exact source Conditional κ = .003639264106282 (+1.20671216e-7 over #7119): compose ToikY chart-120 pattern #7120: b3d68cf1db9bf493ded2d8f4cb922939edc0ef95, complete 26-file export with every nested source and notice. Every source native byte is reconstructed.
• New native SHA256 0a27f1224fdb196602d52d4170bc12b8bde0b9818809e73ecdbdaaa5d3b7adbc;73,698 events, 1,490 public inputs, 1,090 independent helpers. Rohan Arun’s Conditional κ = .00363797977 via a 17-stage signed caller #7113 signed 17-stage caller remains byte-identical. All owned lifetimes/rises remain unchanged.
• Full arbitrary-input forward/adjoint, 15 native corruption controls, three 512-column chart checks, complete finite native/signed routes, rational/integer Laurent caller, owner/phase binding, all 3,050 paid/helper maps and 104 main boundaries, fresh grid/envelope/ordinary/fixed-tape costs, both prices and 54 strict+4 functional outer conditions pass. A probe at the lower crossing must fail; both complete gates agree.
• Independent raw crossing: (0.00364010218739649072410851912941669,0.0036401021873964907241085191294167). Both engines reject Conditional κ = .003639264106282 (+1.20671216e-7 over #7119): compose ToikY chart-120 pattern #7120’s invoice at this claim. This is the emitted invoice’s ceiling; BIT headroom is 0.000126587812604, with no global optimum claimed.

python3.12 -B scripts/research_package.py --package research/p104-unlocked-rank8-split

Both complete direct CPython 3.14.7 runs and the final registered isolated CPython 3.12.16 replay passed. Every frozen native and composition receipt field matched, including the crossing-probe rejection control. REPLAY.txt preserves the actual output; metadata was marked completed afterwards. No scientific code or receipt changed after the successful comparison.

Committed bytes and executable modes pass on both commits; source-history checks pass at canonical-based scientific commit ee7fa6751c8c1dbdeeb01daa80754217d5ae30b7. Public commit 1670baa28f5a9631286591068d7cc2934d2c2d37 carries the identical25-file package tree d84224c4654a28016a7e9f0a8133d71ee3d374d0. Publication is package-only atop the older account fork because the credential cannot push workflow-history changes; canonical provenance stays pinned. Scientific replay requires neither Git nor network.

Exact-head run42891566648 failed onboarding at20:08UTC:760tests,1failure,24skips. The same test_current_ci_section_matches_selected_record guide6030/selected7000 mismatch skipped scientific replay. No hosted scientific pass is claimed. Predecessor #7120 exact-head run42887924160 failed canonical onboarding (guide6030/selected7000,760tests,1failure,24skips), skipping scientific replay. The existing correction handoff remains available.

The five framework assumptions, independent human/formal review, maintainer admission and compiled top-level execution remain open. Selected #7000 is unchanged. The bounded five-group screen produced three legal nominations; only candidate 2 was built, with the shortest nominated bases. DISCOVERY.json gives exact coverage and distinguishes local search evidence from the public witness. No exhaustive rational cost optimum is claimed.

Credit Ali Romman, Jerry Chen/CubeJerry, Matarener, Rohan Arun, ToikY for the retained chart 120 pattern and signed predecessors, DreamingOfClouds, Douglas Colkitt/#7000 reviewers, the prior dillona/OpenAI Codex chain and every retained notice. Substantial OpenAI Codex assistance; no subagents. Canonical investigation.