Reduce all-class rank-seven COPY from 41832 to 196 ordinary adapters

[Clear1nsight]

This follow-up to #3950 replaces its alternate full SL64 conjugator by a projector chart and fixed permutation networks. The complete emitted invoice drops from 41,832 ADD slots per column plus 41,832 ordinary interchanges to 8,036 ADD slots per column plus 196 ordinary interchanges, surrounding one reversed 7f child. This is a uniform algebraic construction and a conditional relative cost theorem, not a physical supplier admission or a new exponent.

The construction works for every split rank-seven projector over Z/q^w and arbitrary diagonal ancestor units. A unit principal chart exists by the characteristic-polynomial coefficient identity. One-hot pivot triples regroup into stars; on the chart, L_(CA^-1) U_(-B) supplies the image/kernel conjugator. The head needs only seven explicit image-weight diagonal factors, compensated at one kernel hub. Fixed routing pays 84 ordinary calls for pivots, 56 for graph passes and 56 for the diagonal, including both banks, inverse words and routing back. Equal physical prefix labels remain distinct; no conditional physical permutation or clean-data restriction is assumed.

The full relative invoice is T_R(v,7f,w)+196 b(v,f)+C_q v(8036f+196+3)+gamma, plus the actual63 XOR readers and outer production/owner/return charges when not already billed. Copying, releases, disposal, metadata preparation and visited backing remain explicit. The positive ordinary term still cannot be absorbed using the pinned B1/B2 upper tariffs at R437 scaling. Reducing its constant does not change that exponent gap. This is not a runtime lower bound or an objection to the original good/fallback caller.

Validation: all four registered Python 3.12 standard-library commands passed on the original construction. After the cleanup correction, normal and optimized primary checks and the Git-free export/rejection checks passed again; the separate checker is unchanged. The primary checks45 contexts on all nine literal R437 captures,192 complete GL2/weight cases, one ambient64 nonzero graph with singular leading chart, complete81-address primitive pullbacks across486 payload bits and all63 dirty readers. Controls reject two active pivot flags, bad chart, wrong rank, nonunit/dropped weights, omitted routing and free ordinary calls. A separate checker imports no compiler: all234 nontrivial3×3 F3 idempotents and108 rank-one2×2 Z9 idempotents pass direct determinant/Cramer chart checks;5912 permutations give21 routing minima;124 unit diagonal identities and the invoice agree. Actual optimized Git-free read-only export succeeds; five altered/missing/linked-source or stale/typed-invoice CLI controls reject. These are separate implementations in one assisted session, not independent human or formal-kernel review.

Only research/r437-parabolic-copy is added. The retained R437 word and predecessor proof are exact self-contained bytes from #3950 head364701494bc3d726806a15d4d1d4b547c1e0f9d5, ultimately R437 a24f9ff. No pending Git object is needed at runtime. Source pins, exact result, registered commands, local validation and written proof are included. Reproduce: python3 -I -B scripts/research_package.py --package research/r437-parabolic-copy.

The actual encoded coefficient producer, whole-volume physical/compiler integration, source/continuation/refinement, stock producer, uniform preparation/visited reserve, native/router/precision/recovery and cheaper complete ordinary tariff remain open. No global macro word, selected cold replay, Lean or Docker run is claimed. The favorable constant reduction stops at this explicit boundary.

Explicit donor parallel research assignment refines independent-moment, priority revision6. Canonical main20ecf2891d80b5acf18d980b2164fb44ea7faa90 was refreshed unchanged at entry, construction and submission checkpoints. Shared claim: #3967 (comment). Local overlap cache lacked remote coverage; live COPY PR searches and the parent's Discussion review provide bounded coverage, not a novelty certificate. #3967's physical binding and #3958/#3959's conditional profile work have different scopes.

Credit ai-nanachi/#3647, Ali Romman/#1043, Sagar Thakrar/#3700/OQ4 and retained R437 lineage; alephnan/#1627, eumemic and Jose Manuel/#2234 for split-projector/common-chart mathematics; David Arias/#2753 and Jose Manuel/#2809/#2899/#2915 for caller/backends; zxv/#3690 and this donor's #3936/#3950/#3967 for predecessor audits/binding. Full predecessor notices are retained. The coordinating agent's cross-review supplied the image-only kernel compensation. Prepared by Chris R / Clear1nsight with substantial OpenAI Codex assistance; Apache-2.0. No merge, closure or selected-frontier change.

Same-session skeptical review corrected a tautological cleanup check: the verifier now clears its actual nonzero private model buffer in place and checks source and completed-target preservation. Physical erasure, head restoration and cost remain backend premises. The reserve formula explicitly requires fixed reusable tape inventories and whole-run lifetime maxima. The preserved 07:52 baseline receipt is distinguished from the corrective validation entry. The construction and 8036/196 invoice are unchanged. Credit the native_transport review for this correction. Current-head CI and committed-byte evidence will be reported separately.

Final publication head: f33a1708675a675a5527a8b0dc80e27ec829f729. All four registered commands and the unchanged-input check pass in final-head CI run39221067024. Exact16-file published-byte/source-history verification, corrective scientific hashes and whitespace checks pass. Baseline run39219644082 remains evidence for the earlier head.