[rohanarun]

Global output-chart recharting improves conditional κ to 0.003641428753638, a gain of 0.000000189103454 over #7123. It changes 21 helper output frames and 129 scalar-contact charts, reconstructing every causal transport while preserving scalar coefficients/order, main boundaries and owned inventory.

The initial draft was published immediately after exact arbitrary-input native and directed complete-price verification. Subsequent full native/application checks and the registered isolated Python 3.12 replay passed locally, matching every frozen receipt field. The full evidence is retained in NATIVE-RESULT.json, RESULT.json, REPLAY.txt, PROOF.md and APPLICATION.md.

Coverage includes exact forward/adjoint endpoints, eleven 512-column chart checks (eight new chart pairs), phase and eight corruption controls, fresh integer grids and precision/error envelopes, finite native/signed physical routes, the exact rational/integer 17-stage caller, all 3,038 catalogue checks and 104 main boundaries, ordinary and fixed-tape costs, both directed price engines, and 54 strict plus four functional outer conditions. The full application has 110,400 paid positions, 22,291 productive fields, ordinary work 777,508,024 and precision context 2,344.

The independent raw crossing is (0.0036414287536383037542599734750979, 0.00364142875363830375425997347509791). Both engines reject the #7123 invoice at the new claim. This bounds the emitted invoice, not all constructions. The five inherited framework assumptions remain; author replay is separate from independent human/formal acceptance, maintainer admission and a selected-frontier update. Hosted CI status remains separate in the checks panel.

Per-native ΔT={1:−2,2:+2,5:−1,6:+1,8:−18,9:+18}; ΔP={1:−35,2:+12,3:+8,4:−8,5:+8}. The complete caller repeats these 17 times. Paid positions decrease by 15 per native invocation (255 total); both rank masses rise by 21 and full rank surplus stays 1,424, with no empty-root credit.

python3.12 -B scripts/research_package.py --package research/p104-global-output-rechart

The complete #7123 export is frozen at 020d581d9856b6db83e448c3f20418bc63b72c34, retaining all nested sources and notices. CHANGES.json and construct.py regenerate the entire word.

Credit 1337isnot1337/OpenAI Codex for fresh-helper-boundary alpha expansion; dillona/OpenAI Codex for #7123; Rohan Arun's #7122 rechart and #7113 caller; Ali Romman, Jerry Chen/CubeJerry, ToikY, Matarener, DreamingOfClouds, Douglas Colkitt/#7000 reviewers and all retained source authors. Rohan Arun with substantial OpenAI Codex assistance; no subagents. Apache-2.0.

Also includes the existing #7113/#7122 correction to docs/ci-verification.md. The initial hosted run failed its current-CI-navigation test because the guide said #6030 while the selected record is #7000; scientific replay was skipped. The targeted test passes with this documentation correction. No selected result or frozen scientific source changes.