[ToikY]

Result

NO VERIFIED GAIN. The κ change is exactly 0%. The strongest independently reproduced comparable complete witness is #7161 at κ = 0.003644890067940, live head 7a607733a114689a60179898ab5ec9e30561daac.

On the retained #7001 BIT word, all 22 main rank-six-to-eight terminal banks have zero qualifying independently restored same-chart helper transfer stock. The exact map exhausts 44 literal, 44 new pairwise-join and 22 generic rank-seven chart classes. It grants all old coframed contacts at all times and all compatible virtual pivots among 185 subdividable public terminal banks. In 70 classes the compulsory main-pivot response cannot cancel outside the helper group. The other 40 have zero pairing with every pure helper prefix. A separate projective-normal enumeration and marked-pivot audit reproduces those counts.

The complete signed BIT moment is still at most about −4.676835409452486 at κ = 0.0040075385047236 and −4.910335615712435 at the stricter raw trial. The required stock is 21/22 and 22/22 respectively. The optimistic bound grants all added setup, routing, ownership, restoration, precision and tape costs free. No changed supplier or individual site was built.

Scope and replay

Canonical base e54523517d0f03e58352e647c403a5af3e525a57. Changes the bit-helper-contract card to the retained #7001/#6726 terminal bank graph. The three canonical #7001 files bind the literal word, signed invoice and directed interval engine by path, Git blob and SHA-256 in SOURCES.json. The contribution keeps normalized contacts and other bank jobs fixed and requires each transfer to restore its own outside coordinates. #7200 concerns strict-prefix rows. #7202 concerns three mandatory paired IDLE children. This package answers the distinct terminal-bank question from Round 18. The separate selected-replay navigation commit reuses the verified #7161 CI prerequisite.

From a clean checkout with CPython 3.12 and assertions:

python3.12 -B scripts/research_package.py --package research/bit-terminal-main-bank-stock

Local registered replay, all 13 committed-byte/mode/source-history checks, six navigation tests, and the independent coefficient audit passed. Corrupt source and receipt controls reject. Hosted status is pending. No changed native, compiler, tape or complete multiplier replay ran.

Credit and next question

CubeJerry #7001 supplies the native word and tariff. Credit #6726/#6834/#6980 and the delayed-cover authors for the native lineage, Nick Petalas #7175 for the concave helper tariff, Round 18 of #7111 for the initial terminal price and transfer question, and Clear1nsight #7161 for the benchmark. OpenAI Codex and subagents materially assisted. Apache-2.0.

Joint corrections across several selected pivots, a changed bank graph, cross-chart corrections and changed scalar chronology remain open. Before testing a site, enumerate already paid nonterminal and erased bridges, deduplicate the compatible owner stock, and charge every extra operation.