Finite complex construction: CK grid-preserving circuits save 1,310 per orientation (no new kappa)

[ToikY]

Historical PR626 CK coefficient macros can use shorter single-scratch circuits while retaining the original magnitude envelope, dyadic grid, frame and clearing boundary. This finite construction saves 1,310 real-component primitives per orientation against PR696 and 180 net against the complete four-Horner PR837 comparator. No new kappa.

Complete CK scalar projection	PR696	Four Horner choices	Replacement
Forward	169,618	168,488	168,308
Backward / inverse-transpose	169,634	168,504	168,324

The separate complete recount includes 234 primitives saved at strictly improved occurrences and 54 additional primitives elsewhere. Every COPY, source ADD/SUB, shift, signed accumulation and final dirty-work CLEAR is paid. No helper, frame transition, restoration or recursive child is added or omitted; paid profiles are unchanged. Only research/ck-grid-synthesis-seed24/ is added.

python3 -B -I research/ck-grid-synthesis-seed24/verify.py
python3 -B -I research/ck-grid-synthesis-seed24/independent.py
python3 -B -I research/ck-grid-synthesis-seed24/test_package.py

Expected main marker: PASS_CK_GRID_SYNTHESIS magnitudes=103 programs=2 columns=725 savings=1310,1310 no_new_kappa.

Validation passed on macOS / CPython 3.12.14, including isolated optimized execution and a minimal offline relocated checkout. Both literal primitive programs check all 725 arbitrary input columns, restored dirty helpers and erased private cells. Separate BFS and synchronous Bellman implementations agree on all 103 graphs, 16,251 vertices and 96,929 labelled edges. Nine graph, six macro, two whole-word and two admission controls reject meaningful mutations; gzip repackaging passes. Whitespace and one-package scope pass. Exact certificates, seven source pins, proof, validation and local claim are included. Finite checks do not formally prove the multiplication theorem.

Exact base: ba82555. Fresh OS seed 24, complex-supplier track. No pinned sprint issue exists; the local claim records selection, stopping rule and timing. Current accepted selection at this base is PR1042, kappa 0.001672188119205421. This deliberately contributes historical CK evidence, without transport to the selected word.

Credit Philipp / @mightyOil #882 for the JP8 grid-graph method and proof: this adds complete CK coverage through numerator 512 / denominator 64. Hanson Sun / @bbMisk #1036 covers source/target mutation and has larger gains in that distinct class; our certificate settles the bounded single-scratch class. #999 permits multiple target updates. #723/#734/#787 signed-digit and #722/#739/#769/#811 shared-work approaches are credited without importing pending code or adding their savings. The final overlap refresh reached #1113 and refreshed the nearest heads.

Credit Sagar Thakrar #696 (including the adapted interpreter), Gareth Davidson #837/#866, Chris Pagolu #734, and the icekylinx / Jeremy Kothe / Chafik Boukhalfa / DreamingOfClouds / Rohan Arun / eumemic lineage. Prepared for Eva Ayala / @ToikY with substantial OpenAI Codex assistance, including separately implemented AI checks; Apache-2.0. No independent human review or shift/add priority claim.

Optimality is only for the stated bounded grid with mandatory initial COPY, work-only preparation, one target update and final CLEAR. Full repository, supplier cold replay, complete selected-record inventory, physical/native/global admission, hosted CI and Lean passes are not claimed. All-size/compiler, realization, hosting, precision/tape and analytic obligations remain inherited. The selected frontier and upstream/ are unchanged.

[Mister-Meeseeks]

100-PR batch review at 82575f429d817f1730c28537e669e56a8016e3a7.

Merged to main, preserving the original contributor commits and package attribution.

Reviewed scope: Exact finite grid-preserving CK single-scratch coefficient circuits and shortest-path distances; complete per-orientation scalar saving, no paid-moment or runtime claim.

The pinned scoped command passed; the batch ledger contains its command, input hashes, environment and log. This incorporates the research/tooling contribution and does not change the selected κ or discharge the inherited multiplication-theorem assumptions.

Batch review, attribution and evidence. Thank you for the contribution.