[aliromman]

Two public transport intervals now use a direct rank-five move instead of separate rank-four and rank-one moves. Existing owned carriers deliver their intermediate exports, and an exact output basis correction restores every arbitrary input.

The complete conditional claim is κ = 0.003288299, up 0.000001285 (+0.03909%) over PR6453 at b1a3e525793eca35c56324aa7ffd0b714cf606d4. The new increment excludes all inherited gains.

Construction and complete cost

Fields 368 and 336 each have an intermediate SCALE -1. The construction keeps that sign change at the held frame, redirects one export through an existing carrier, and emits 569 dyadic scalar gates at the common output frame. This extends Rohan Arun's constant-value donor merge, PR6399, to scaled intervals.

Per native call, the paid change is 2H5 - 2H4 - 2H1. Across all eighteen calls it is 36H5 - 36H4 - 36H1, with every productive target fixed. The full invoice has 67,096 paid jobs, 13,652 productive fields and rank-mass surplus 760. The extra scalar work is charged; the local allowance is 137,218. The 29 owned labels, 45 lifetimes and peak storage 13 remain unchanged.

The unique raw crossing for this complete fixed invoice lies strictly in
(0.00328830463392494440038260, 0.00328830463392494440038261).
The claim is less than 5.635e-9 below it. This changes the native word and paid profile, outside the historical fixed-local packing model. It does not establish an optimum for other words or callers.

Only research/p104-scaled-transport-fusion/ is added: 18 files. The exact PR6453 export and all 26 original member pins are retained with their notices. Both live inputs are pinned to an ancestor commit and declared in VERIFY.json. Shared code and the selected frontier are unchanged.

Verification

Local CPython 3.12.13 normal and optimized registered replays pass, as do both isolated, read-only, Git-free, empty-PATH export replays. Coverage includes all 799 columns, 2,939 native maps with exact phases, 208 main adapters, 7,488 physical caller contacts, two rational price engines, compiler and outer inequalities, sixteen named controls and retained source controls. Both price engines reject the unchanged source at the new claim.

Metadata, source pins, committed bytes/modes and official export integrity pass at the publication head. Since the complete scientific replays, only README, task metadata and validation prose changed. The optional full-history check was not run because the checkout is shallow; immediate-base/source ancestry and every cold runtime input were checked.

python3.12 -B scripts/research_package.py --package research/p104-scaled-transport-fusion

Hosted CI is pending. It will be recorded here for the exact published head.

Scope, related work and credit

This is an author-checked conditional construction. Independent scientific review and all five framework assumptions remain: native_all_columns, owned_erasure, compiler, precision_tape and analytic_recovery. The large uniform word and rounded multiplier are not materialized.

Canonical main remains 7c49fe32c55de02c2c873655c804b16e4c3fef0a, priority revision 8. Live PRs and Discussion claims were checked before publication. PR6453 supplies the retained stack; PR6399 supplies the donor-merge precedent. The fixed-word frame searches and PR6524 replay optimization have different scopes. Searches do not cover unpublished work or every historical diff.

New construction and package: Ali Romman with substantial OpenAI Codex assistance; no subagents. Credit Rohan Arun, Erwan L., DreamingOfClouds, Cinna Davis, David Arias, Gareth Davidson, Jacob Alan Sussman, zxv, Cao Yifan and all retained source authors. Apache-2.0. Start review with PROOF.md, SPEC.json and the registered replay.

Canonical investigation and handoff.