[JoyboyBrian]

The #4957 handoff leaves a concrete owner/route implementation obligation behind its conditional completed BIT backend API. This contribution supplies the addressed contract-machine schedule for actual original helper462/frame506 and its explicit S8 ⊗ I_f, N=8f tensor family. It emits chronological elementary high-coordinate operations, every carry cell, a paid low-type permutation, complete cache lifetimes, fallback calls and nested ROW borrow/return operations. The serialized example contains executable operation names and coordinate addresses.

The independent interpreter consumes that program without importing its generator. A separate source oracle checks the full forward cover row, inverse restoration and paired projector action. Exact rational checks independently reconstruct P²=P, SQS⁻¹=P, and both actual emitted elementary-factor products. Physical owner slots and objects must be distinct; all unrelated fields, original spectators, ROW values and head markers are checked after each step. A review found that accepting arbitrary odd radices would invalidate local-ring pivoting; both executable interfaces now perform exact primality checks and reject composite q=15.

Scope: this closes one P1/P2 scheduling obligation under named primitive contracts. Native, raw-body, ordinary, scaling and routing leaves still need their actual finite-alphabet fixed-tape implementations and movement/cost proofs. The ledger records all calls, widths, owned low-cache bits, ROW padding and integer precision, but unresolved primitive costs are not zero. It does not materialize full padded payload arrays or the enormous #4814 uniform compiler. P3, full primitive API admission, whole-program reserve/tape/precision bounds and any new κ remain open. It does not recalculate the already-covered scalar stock constants.

Validation on Darwin 27.0.0 arm64 / Python 3.12.2:

• Registered normal and -O replay passed: 3.739 / 3.706 seconds for the scientific checks, 8.55 seconds total wrapper time. The fixtures (f,q,w)=(1,5,2),(1,5,4),(2,7,4) emit 781 / 1,501 / 5,965 steps and exercise fallback plus 80 / 320 inner ROW returns.
• All 64 head/tail basis columns, 64 high-matrix basis columns plus zero, six dirty cases, 48 exact emitted rational factors and 15 rejecting controls passed. Omitted nonzero carry, omitted factor and wrong high-part controls recompute the operation ledger so they are rejected by wrong action, not stale counts.
• check_research_commit.py --package research/bit-row-route-generator --check-source-history passed on the exact submitted bytes; all input pin commits are in publication ancestry.
• Export/check passed; both normal and -O direct replays passed from a fresh extracted archive with PATH=/nonexistent and a fresh cache prefix. A modified source file was rejected by its SHA-256 binding before execution.
• Hosted CI has not yet been observed. No Lean, Docker or unrelated historical suite was run.

Reproduce: python3.12 -B scripts/research_package.py --package research/bit-row-route-generator.

Source base b91956876e5b4122fd89179f1b57557ff8378f15, priority revision7. The #4957 handoff remains at exact head 3ab584de66a9220007963c7bd476241137a263c9 (live rechecked); it is attribution/context, not an unmerged replay input. Live frame bank and native WORD are pinned in SOURCES.json.

Canonical investigation: #4945 (comment) . Entry overlap covered 317 open PRs, 4,064 Discussion records and 6,636 local notes. Final scoped offline search includes 6,638 local notes and finds the known handoff, our package and unrelated row generators; its remote cache may miss later changes and does not prove novelty.

Credit ai-nanachi, Chris R / Clear1nsight, rohanarun, Maxime Fleury, Cinna Davis, zxv, Jacob Alan Sussman, Mark/Matarener and the retained native/E8/ordinary/ROW/routing/precision/tape lineage. Prepared by JoyboyBrian with substantial OpenAI Codex assistance. Frozen sources and selected frontier are unchanged. Four commits separate schedule/proof, independent execution, prime/rational boundary repair, and source-bound evidence/replay.

Open next questions are in the README and TASK: lower one addressed leaf with its physical bill, substitute the complete native runtime with the same dirty-state contracts, and bind the full uniform matcher list before deriving P3 constants.

Final submission checkpoint: official main refreshed at 05:26 UTC remains b91956876e5b4122fd89179f1b57557ff8378f15, with unchanged governing guidance and affected source bytes. The new complete overlap snapshot covers 332 open PRs at 2026-10-23T05:26:43.366179+00:00, 4085 discussion records at 2026-10-23T05:27:32.011143+00:00, and all 6,636 current-base notes. Scoped helper462/ROW search found the credited handoff and this canonical claim. A separate verifier-agent reread the final code and completed metadata, reconstructed its original composite-q counterexample and verified both interfaces reject it normally and under -O, and independently multiplied all 48 emitted right factors. No blocker remained; physical primitive and P3 limits remain explicit. This is advisory overlap and author-process review, not a novelty guarantee or independent human acceptance.

[jwr404]

Independent check at head fe886dc0431d33e9df856b47d0510f1ec66ef750

Linux replay. The registered replay passes on Linux with CPython 3.12.14 in 9.3 s (outcome=partial). The package was validated on Darwin only.

Interpreter independence. interpreter.py imports only copy, fractions and math, and nothing from generator.py. The two share no code; verify.py imports both.

Prime radix. Both interfaces reject every composite radix I tried, q ∈ {9, 15, 21, 25, 49}, not just the q = 15 control. The generator was called directly, and the interpreter received a serialized program with q overwritten.

Randomized dirty states. The package's action check uses one deterministic state family (state_for). I added random states and computed the expected endpoint with my own modular arithmetic; only source_matrices is reused, as the target definition. Each state has:

• a random low type g₀ invertible mod q,
• random high digits, h, d, outer ROW fields and tape head,
• a random 461 spectator.

The expected endpoint, computed with my own unit-pivot inverse over Z/q^w:

• h and d follow the projector request g·P·g⁻¹;
• route_returns[0] = g·S and route_returns[1] = g;
• low and high are restored;
• outer, z, tape_head and owner 461 are unchanged.

(f, q, w)	steps	random states
(1, 5, 2)	781	12 ✓
(1, 5, 4)	1,501	12 ✓
(2, 7, 4)	5,965	12 ✓
(1, 3, 3)	781	12 ✓
(1, 11, 2)	781	12 ✓

q = 3 and q = 11 lie outside the package's fixtures.

Scope. As stated, this checks the schedule under the named primitive contracts. The native, raw-body, ordinary, scaling and routing leaves, their fixed-tape costs, and P3 remain open. My check does not price any primitive.

This is an author-independent check, not maintainer acceptance. κ is unchanged.

Reviewed by James Reed with substantial Anthropic Claude (Claude Code) assistance.