Audit paid-query composition at the actual QQ16 bridge

[Clear1nsight]

The actual ordinary bridge after F0/B1 needs current Y, which the earlier 96-output F17 projection does not supply alone. This package adds eight paid current-query outputs and constructs a complete local consumer of the resulting 104-field boundary. Exact routed algebra proves the bridge update for arbitrary arrays, including all owned dirt and endpoint frames.

The literal serial composition costs 208 H1 children, 416 native wrappers, 400 cover maps and 15,040 scalar primitive allowance, with 120 exclusive full-cover workspace fields and eight owner-private fields, all returned zero/identity. Exact fusion cancels the added query subtraction/reconstruction and leaves the existing 96 source bridge ADDs. Thus this repair closes the information gap but removes no additional source work for this bridge-only endpoint. Under the same conditional common-label native entry boundary, direct subtraction costs 768 scalar allowance; preserving and clearing the same 120 allocated dirty fields costs 1,008 and is checked against every final serial value/frame.

This is a scoped local construction and stop result. The original hosted template uses gauge44; actual QQ16 placement, inter-pair alignment, movement, storage, rounded precision, coverage and recursive supplier admission remain open. It is not a global cost rejection, a universal bridge lower bound, or an eight-new-family lower bound. The next viable step is a changed producer that deletes named source work, or an actual independent continuation that shares the h/query computation. No centre-retirement saving is counted again.

Validation: normal and optimized read-only replay; 106,496 exact signed native basis columns; four semantic word mutants plus wrong inverse-kernel translation; seven boundary tests including Git-free read-only export and source/word/typed-receipt corruption; registered package replay and baseline binding check all passed. Independent AI agent proof/code review found no blocker within scope, with exact reviewed revisions and subsequent hardening recorded. No human or formal proof-assistant review is claimed.

Final publication head: c8e72091d379972d510a409ede15eaf44eb7b6be. check_research_commit.py --check-source-history passed for all 18 package/input files, with canonical base 20ecf2891d80b5acf18d980b2164fb44ea7faa90 in ancestry. Official committed-package export and Git-free bundle integrity checks passed at that head. Canonical main was refreshed unchanged before submission; priority revision 6. Live overlap coverage is partial.

Package: research/qq16-paid-bridge-composition. Reproduce with python3 -B scripts/research_package.py --package research/qq16-paid-bridge-composition.

Follow-up of #3963 and #3952, coordinated in #3963 (comment). Preserve credit to the pinned demand/boundary audits, ywz65 #2512/#2707, Arham Tahir #2681, Ali Romman's QQ16/source lineage, alephnan #2346, thakrarsagar #2470 and retained geometry/native/supplier contributors. Exact provenance, original task card, remaining questions and substantial OpenAI Codex assistance are recorded in the package. Sibling recovery/budget results are not scientific or executable dependencies.

Hosted verification succeeded: run 39218677372, replay job 117949419851, CPython 3.12.15. Logs bind merge 0d93b29 of exact head c8e7209 into base 20ecf28, the package completion marker, all seven tests (16.214 seconds) and unchanged committed inputs. This is hosted package verification, not global supplier admission.