[dillona]

ToikY reported the chart-120 pattern in Discussion #7111. This PR independently constructs it on exact #7119 and verifies the complete application, raising conditional κ to 0.003639264106282, a gain of 0.000000120671216. The peer’s original #7115 artifact remains unpublished and is not imported. Credit for the pattern stays with ToikY.

At source cut 39058, public role 962’s rank-seven MOVE is subdivided through chart 120 into ranks four and three. Helper 1447’s rank-eight target becomes two independent rank-four targets. The boundary basis is identity; four unit ADDs implement the middle basis. Total additions: 8 ADDs + 1 net MOVE. PROOF.md gives the concrete arbitrary-input identity and strict concave gain H4+H7−H8−H3.

The complete signed 17-stage deltas are ΔP={3:+17,4:+17,7:−17}, ΔT={4:+34,8:−17}. The invoice has 110,723 paid positions, 22,257 productive fields, surplus 1,424 and no empty credit. Ordinary work is 777,741,264 <10⁹; fresh precision retains c=2342. The compiler gap changes to 10⁻¹⁸ and H to 172444000000000000000000, with every affected gate recomputed.

• Exact source: Conditional κ = .003639143435066 (+1.00503559e-7 over #7118): reciprocal helper split #7119 bd244b0e5034cb1b9f72641494c3421f757ec2e4, complete 25-file export with authenticated nested closure.
• Native SHA256 4f19c6ee57b74d7c3ea1863df0b33fe275a3a7511f77084f05726ee216ebdc34; 73,672 events, 1,489 public inputs, 1,089 independent helpers. Rohan Arun’s Conditional κ = .00363797977 via a 17-stage signed caller #7113 caller remains byte-identical. All owned lifetimes and rises are unchanged.
• Verification covers complete forward/adjoint, 14 native corruption controls, five 512-column chart checks, finite native/signed routes, rational/integer Laurent caller, owner/phase binding, all 3,049 paid/helper maps and 104 main boundaries, fresh grids/envelopes/ordinary/fixed-tape costs, both price engines and 54 strict + 4 functional outer conditions.
• Independent raw crossing: (0.00363926410628203626930823372185618,0.00363926410628203626930823372185619). Both engines reject Conditional κ = .003639143435066 (+1.00503559e-7 over #7118): reciprocal helper split #7119’s invoice at this claim. This is the emitted invoice’s ceiling, with BIT headroom 0.000127425893718; no global optimum is claimed.

python3.12 -B scripts/research_package.py --package research/p104-chart120-peer-composition

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 stale-probe rejection control. REPLAY.txt preserves the actual output; metadata was marked completed afterwards. No scientific code or receipt changed after the successful comparison.

The first composition replay caught the inherited fixed probe overshooting this crossing. The counted caller adapter now independently applies the credited price engine’s exact decimal shrink rule; a negative control requires the old probe to fail. The strict inequalities remain active and both complete gates agree. VALIDATION.md records this failure and correction.

Committed bytes and modes pass on both commits; source-history checks pass at canonical-based scientific commit c40ed08f7f0d7c08a2317fb01434fd5a3a7a1806. Public commit b3d68cf1db9bf493ded2d8f4cb922939edc0ef95 has the identical 26-file package tree c480f8103238a3bd5c06c4b71c266edcb2a5baf8. Publication is package-only atop the older account fork because the credential cannot push workflow-history changes; canonical provenance stays pinned. The scientific replay needs neither Git nor network.

Exact-head run42887924160 failed onboarding at19:46UTC (760tests,1failure,24skips): test_current_ci_section_matches_selected_record found the guide6030/selected7000 mismatch. Scientific replay was skipped; no hosted scientific pass is claimed. Predecessor #7119 failed the known canonical onboarding defect (CI guide #6030 versus selected #7000), skipping scientific replay. The existing correction handoff remains available.

APPLICATION.md supplies the exact physical, ownership, common-grid, precision and fixed-tape substitutions. The five framework assumptions, independent human/formal review, maintainer admission and compiled top-level execution remain open. Selected #7000 is unchanged.

Credit ToikY for this pattern; Ali Romman, Jerry Chen/CubeJerry, Matarener, Rohan Arun, DreamingOfClouds, Douglas Colkitt/#7000 reviewers, the earlier dillona/OpenAI Codex chain and all retained notices. PEER-CREDIT.json records the original finding. Substantial OpenAI Codex assistance; no subagents. Canonical investigation.