[aliromman]

Constructive conditional ordinary BIT saving: 0.003486609089, an exact increase of 0.000002639113 over PR #5512. This implements and fully prices the four partial-helper retirements proposed in PR #5629. It provides more BIT capacity for a compatible multiplication caller. No complete kappa or selected-frontier update is claimed.

Scope, proof, exact checks, source pins, registered replay. Base: canonical main 6cf7a683f447b9f1e2333791907fd92be5009a63, priority revision 7. Contribution head: 122a43e2fdf867a252366543c3af9a128d682d89.

The emitted native word preserves all protected reads, clears the four owned fields after their last nonzero read, and removes their H3/H1 return tails. The complete 18-call controller retains 11,166 independent originals, pays 56,853 queries and allocates 15 owned fields with no output credit. The exact gain is 72*(1+3^t-4^t)>0 for 0<t<1, after all target debits. The new ordinary certificate includes fresh finite compiler, storage, prime, precision, route and backend thresholds. Discussion #5636 records scope and coordination. Submission refresh: canonical main df0137e now selects conditional kappa 0.00310045; priority revision 7 and source construction are unchanged. New PR5639 proposes a separate SOURCE-snapshot fusion, while Discussion5638 reserves changed source/output continuations.

Normal and optimized registered replay rebuild the local word, 455,592-event caller, 56,853 chronological source descriptors, full target bill and four finite ordinary levels. Separate adjoint, stream and rational-series implementations check the result. Corruption controls reject missing clears, early erasure, retained-input erasure, stale epochs and unfunded exponent claims. The extension census checks all 423 single additions and 1,174 declared cohorts; none improves this word in 0<=C<=.01. The ordinary saving lies within 1.85e-11 of the exact raw crossing interval (0.00348660910743,0.00348660910744).

python3.12 -B scripts/research_package.py --package research/g20-retirement-native

Python 3.12 and the standard library suffice. Committed-byte checks pass. This is a conditional BIT component result. The complex supplier and complete multiplication caller remain separate obligations. The source native/all-width/ROW/compiler/precision/tape/analytic contracts remain assumptions; full historical suites and formal-kernel proofs were not rerun. The finite uniform program and its astronomical decimal threshold are algorithm-defined, not materialized. The second implementation is not independent scientific review. Other helper subsets and changed nonzero reconstruction programs remain open. The full ancestry check cannot run in this shallow checkout; frozen file hashes and committed bytes were checked. Hosted planner failed in the upstream Current CI navigation test at canonical main df0137e; the package replay job was skipped. The same two missing-link failures reproduce on main. Local normal, optimized and Git-free committed-export replays pass.

Credit Dugongue for the four retirements; ai-nanachi for the native G20 and ordinary source; Ali Romman for PR #5421; and DreamingOfClouds, eumemic, Cinna Davis, rohanarun, zxv, Jacob Alan Sussman and all frozen source notices. Source proposal head 9e51f781e4bd174c32cc75ec4504c23f665fb23d and all replay bytes are pinned.

Next: the distinct SOURCE-snapshot fusion in PR5639 can use this emitted and fully priced baseline. Discussion5638 owns the changed source/output continuation investigation; this package reserves no further search. A complete kappa increase also needs a fully billed compatible complex supplier.

Prepared by Ali Romman with substantial OpenAI Codex assistance; Apache-2.0.