Execute selected complex compact adapters with sequential repair

[jmbarrancoml]

I implement the selected complex supplier's compact-control movement with sequential record tapes: ordered rotations, compact interchanges, exceptional-record extraction, inverse destination keys, stable binary radix sorting, reinsertion and cleanup. I then compile its fixed Gaussian monomial recipes into columnwise controls and unit phases, including the inverse program.

The retained notes already give the algebra, guard and asymptotic formula. This package targets their physical execution boundary and the actual selected R480/T16 catalogue. Given a completed ordinary interchange with cost O(V b^tau), the written composition gives O(V((f log p)^tau+1)) on the stated spaced-axis/long-record domain. It keeps control preparation and repair costs explicit before absorption. Tests use a completed elementary swap implementation and retain its call charges; the faster ordinary contract can use pending PR2234.

Evidence includes 3,474 distinct monomials and 1,778,688 exact local-column checks, complete small record executions, physical Gaussian forward/inverse programs, and controls for omitted repair, wrong inverse keys and a missing tensor phase. The full repair experiment processes 8,388,608 records, including 6,291,456 exceptional records through 23 radix passes. In that experiment, address oracles prepare the raw S input and expected T output; the repair executes through the sequential interpreter.

The source binding checks both selected 24,767-event local programs. It distinguishes unary Clifford phases from fixed role scalars: 0:1000.right sends two zero columns to (245,245) with phase zero, while the role SCALE coefficient remains -1/2. The compiler follows the source's strict-upper quadratic convention even when a raw recipe stores diagonal/lower entries.

Validation: source-authenticated Python 3.12 normal and optimized bounded replays, plus the documented full repair experiment. The default registered hook runs the bounded checks; verify.py --large repeats the larger receipt. Reproduce from the repository root:

python3.12 scripts/research_package.py --package research/selected-complex-compact-adapters
python3.12 -I -B research/selected-complex-compact-adapters/verify.py --large

I leave the complete complex caller, recursive numerical precision, finite supplier admission and analytic transfer separate. I change no selected source or frontier, and claim no new guard or multiplication exponent. Reviewers should check PROOF sections 3–6 against the tape interpreter and the ordinary-callee contract.

Coordination: Discussion2238. Canonical main remains 0fdd7d82de03e43c9e5a1000dfbcd380d1b687d2, priority revision 4. The final text search covered 477 open PRs, 952 Discussions and 1,538 local records, with complete remote pagination and no warnings. It excludes arbitrary code diffs.

Contribution by Jose Manuel (jmbarrancoml), with substantial OpenAI Codex assistance. NOTICE.md and SOURCES.json retain Douglas Colkitt's compact-control construction, the OpenAI framework, prior carry/guard audits, Ali Romman's source loader and the selected native Clifford/gauge44/T16 lineage. I claim no independent human or proof-assistant verification.