[unpaidmoon]

Change and mathematical scope

On the 36 E6 lines stored as an A2 complement, the published bit-lane iterator returns exactly the words of weight at most 16 in the F2 incidence code of the 135 orthogonal 4-frames. The code is 7-dimensional: zero and 63 words of weight 16, and no lighter nonzero word. A literal doubling and two half-splits agree with low_weight_words at batch sizes 16 and 8.

This is a coverage comparison for one other code. It is not a timing claim, not a consumer installation, and no new κ is claimed. The selected value stays 781451/250000000.

Task and overlap

Base commit: 7c49fe32c55de02c2c873655c804b16e4c3fef0a on canonical CrocSwap main. Head f770aeb71385a5e14b38ca6e1e0df6cccde6c646 on fork branch cursor/replay-coverage-gap-3fad.

Task card replay-runtime, priority revision 8. Scope status: refined. The card asks for a measured speedup that keeps coverage. The revision-8 verification row says a published faster replay is not equivalent coverage until checked. The merged E7 comparison already checked the 27-dimensional kernel. This package checks the remaining unbound code from that handoff.

Session r116, item 9 of the ordered eligible list. Discussion #1166 was read through comment 19481299 (2026-11-09T00:42:16Z). Handoffs are dropped when completed, when they say no target was taken, when a merged result answers them, when the canonical artifact is an open pull request, or when this account already delivered them. No structured claim was unexpired at 2026-11-09T13:16:06Z. Three open handoffs remain, then the six revision-8 next-decision rows:

1. Three signed added frames at seven calls (comment 19377832, updating 19377345).
2. Eight-source singleton connector at length 30 (comment 19377982).
3. Six-source depth-two block at c = 7 (comment 19384981).
4. Complex helpers (p104-side-correction, coverage-law-change).
5. BIT (bit-helper-contract).
6. New decoders and interfaces.
7. Supplier audit (supplier-boundary-audit).
8. All-size audit (uniform-transfer-audit).
9. This package. Verification tooling (replay-runtime).

Nearest work, and why this is distinct:

• Merged research/e7-bitslice-coverage-check compares the same engine with the E7 kernel. Its second open question asks for one other code. This is that code: width 36, dimension 7, weight bound 16.
• The E7 orbit-hash comparison is a different delivered check of that same E7 support set. It does not build these 4-frames.
• Open Reduce complete A2+E6 primary replay time by 27.86% (no κ) #6584 times an A2+E6 primary replay. It does not form this incidence kernel or call low_weight_words.
• Open Prove six-orbit rank chains cannot exclude two arrays in seven calls #6799 is a six-orbit rank-chain statement. It is not this code and it does not answer handoff item 1.

No fork cursor/* branch names this comparison. The overlap scan covered open upstream PR titles (882 at fetch) and the #1166 comment index. It is not a full-text proof that no open PR body states the same finite equality.

Validation

Python 3.12, standard library only. From the repository root:

python3 -B scripts/research_package.py --package research/e6-frame-code-lane-coverage --check-only
python3 -I -B research/e6-frame-code-lane-coverage/verify.py
python3 -I -O -B research/e6-frame-code-lane-coverage/verify.py
python3 -B scripts/research_package.py --package research/e6-frame-code-lane-coverage
python3 -B scripts/check_research_commit.py --package research/e6-frame-code-lane-coverage --check-source-history

Observed: PASS_E6_FRAME_CODE_LANE_COVERAGE, PASS_REGISTERED_RESEARCH_PACKAGE with outcome completed, and PASS_RESEARCH_COMMIT_BYTES at f770aeb71385a5e14b38ca6e1e0df6cccde6c646. Counts: 135 frames, rank 29, 128 codewords, minimum weight 16, 63 supports. Controls: weight bound 15 returns only zero; dropping one support breaks equality; a weight-15 injection fails the predicate; a repeated generator is rejected. Each of the 36 lines lies in 15 frames. The anchors are E8 roots of inner product 4, and the stored lines are their common orthogonal complement.

Not run: the #5729 timing harness, the 27-dimensional E7 census, reflection orbits, Lean, Docker, and scripts/research_selected_baseline.py. The selected record is provenance only. These finite checks do not replace the general lane argument in research/e7-bitsliced-exhaustion/PROOF.md.

Credit

The E6 lines and A2 anchors are Nick Petalas's witness in research/e6-e7-shadow-minrank-upper. The bit-lane engine is Ali Romman's PR #5729. The request to bind one other code is the second open question of the merged E7 coverage check.

Prepared by unpaidmoon with substantial Cursor Grok assistance. New work is Apache-2.0.

Open next questions

1. Does the same set equality hold for one D6 or P112 punctured frame code, at a named width and weight bound? The E6 check does not transfer. First build that incidence kernel from a pinned root list and compare one half-split with low_weight_words on that basis alone.
2. Does the opt-in intersection-counter engine return this same 64-word set? A faster substitute is not coverage for this code until the sets are compared. First run research/e7-intersection-counters/engine.py on this 7-generator basis at one batch size and compare it with support hash ad4a1d101774e6c2da40cde3d097d39aa56ee5668dcdca858513c5fec3578215.

Workflow feedback (optional)

The half-split helper has to snapshot each doubling. Extending a list from a live view of itself does not stop. A one-line note next to span in the E7 coverage verifier would save the next port of that pattern.