Research/r439 block1 schedule seed8

[Deftsalt-debug]

Finite bit-supplier construction; no new kappa. PR589's ten-wire phase-B block 1 (frame 507, B[128:163], including arbitrary dirty helper 465) uses 30 instead of 35 additions. Rebuilding its required literal reverse cleanup reduces the complete local word 24,305 → 24,295 native ADDs.

The two fixed positive factors attain exact minima 14 + 16 within their destination-once positive-addition classes. Exhaustive nomination and a separately implemented F2-elimination/integer-lifting verifier agree on all 2,048 states, 10,240 state/destination checks, and 8,464 feasible edges. This is a restricted-class optimum, not unrestricted XOR optimality.

The ordered forward integer map, all 661 register frame paths, 439 dirty helpers, nine retained copies, paid histogram and rank mass 5,129 are preserved. The reverse positive matrix decreases entrywise, and the complete reverse prefix falls 1,113,070 → 1,097,158. Forward prefix stays 49,347; payload remains 93 bits under the unchanged 104-bit cap. All preparation, reads and literal restoration are charged.

Reproduce

python3 -I -B research/r439-block1-schedule/verify.py

Expected:

PASS_R439_BLOCK1_SCHEDULE gates=35->30 native_adds=24305->24295 columns=661 rank_mass=5129 controls=13

Additional entrypoints: search.py reproduces the complete nomination; check_package.py checks minimal relocation and corruptions; native.py runs the unchanged C++17/Boost legality and 5/9/11-stage checkers.

Completed validation

• Exact block maps, all 1,024 local Boolean inputs, all 661 arbitrary-input columns, original E8/frame/H1–H5 interface, exported record and COPY/ERASE semantics, complete paid profile and prefix bounds; 13 controls.
• Fresh native legality and 5/9/11-stage checks on both words: 883 / 1,327 / 1,549 formal columns and 96 bridge-omission controls. Python/C++ counts and bounds agree; full normalized native receipts are retained.
• Optimized isolated execution, minimal source relocation with spaces in paths, five source/candidate/receipt corruptions, existing-output preservation, and a second minimal native replay matching the stored receipt.
• make selected-record-check, whitespace checks, four source files compared directly with pinned PR589 Git objects, and a format-patch applied to the exact base that reconstructs the committed tree and passes verification.

Linux Debian 13, Python 3.12.14, GCC 14.2.0, Boost 1.83. No full regenerating repository suite, cold supplier reconstruction, new hosted invoice/outer assembly, Lean or remote-CI success is claimed. Inherited all-size/compiler, realization, hosting, signed-decoder, precision/tape and analytic obligations remain. This is not a formal proof of the multiplication theorem.

Pins, overlap and attribution

Base: f567884aeb8851b6e794af1184119a3bfbe1e631; reviewed PR589 supplier: a5afc8c6169ea61ba7c51dfd3cade02cc7b13b46. Only research/r439-block1-schedule/ is added; selected inputs and upstream/ are unchanged. Source hashes, exact certificates, proof, scope and validation are included. OS assignment seed 8; no pinned sprint issue exists, so the claim is local.

Overlap refreshed through #876; main remains the exact base. Credit navinor's concurrent #873 at 117c48527dd4f066dad103121761154cf1bb39cd, which uses whole-block scheduling on the distinct eleven-wire frame-508 block. This changes ten-wire frame 507, certifies two fixed factors, and supplies native multistage evidence. No mechanism priority or composition is claimed. #844 changes block 4; #857 changes block 8; #728/#792 change A mini-nodes. No pending implementation or savings are imported.

Credit eumemic, DreamingOfClouds, Dylan Crow, Zhihao Chen, DaysSky and retained predecessors; related fixed-factor work by Yuzhou Wang (#789), commdata2338 (#790), Jerry Lin (#810), and Gautam Gupta (#844). Verifier structure adapts mtpsl's #691, and the native harness adapts Ali Romman's #688, with attribution and source pins.

Prepared for Krishnendu Purohit (@Deftsalt-debug) with substantial OpenAI Codex assistance, under Apache-2.0. Both new arithmetic implementations were written by one AI agent; no independent human or delegated-agent review is claimed.