Bit ledger frame ranks: from-operations paid-rank mass audit (research/bit-ledger-frame-ranks-seedfeb2)

[navinor]

Bit ledger frame ranks: from-operations paid-rank mass audit

Aaron Wang, with help from opencode (substantial AI assistance; no independent human review claimed). Apache-2.0. No new kappa: the selected record, the certificate and both suppliers are unchanged.

Scope

Package research/bit-ledger-frame-ranks-seedfeb2/, branch research/agent-39-bit-ledger-feb25c49, head bb35339e on base e112d71c8d7bb32435debcc54df932593dd8223a. Task card: research/bit-ledger-frame-ranks-seedfeb2/TASK-CARD.md; structured metadata: TASK.json (track recursive-interfaces, drawn card coverage-law-change, chooser nonce feb25c49413fb33710658d2618c9ea58, scope status refined).

The drawn card's headline objects (plane-port coverage law, two-unpaid-coordinate templates) are saturated with fresh claims and PRs (discussions 1516 and 1587; pull/1825, pull/1808, pull/1900), and a sibling agent completed connector-alphabet-change (pull/2056). This PR takes the unclaimed first step the repository itself names for the bit-side coverage identity: open question 2 of research/selected-suppliers-de-diagnostic/ ("can nine.profile be rebuilt from the original gate operations of the pinned tune/bitE8mut-p111-tuned.json.gz word"; first step: compute the per-frame transition ranks of the word file and compare with word.local_histogram and nine.profile). The recorded scope change and reason are in TASK.json.

Established

All numbers are recomputed by verify.py from the pinned word file and tune/EXPECTED.json, with frame dimensions computed here by exact rational rank. No code is imported from the supplier's checker chain (stage_cohort.py, legality_h.cpp, global_k.py).

1. The 636 label frames (live line sets, closures of the 120 labelled E8 lines punctured to 111) have exact rank spectrum dim 0:1, dim 1:111, dim 2:153, dim 4:70, dim 6:180, dim 7:120, dim 8:1.
2. Per-register net frame climbs from the recorded start/final frame ids: every X slot climbs 7, every Y slot climbs 7, every helper climbs 8.

Class	Registers	Net climb each	Mass
X slots	111	7	777
Y slots	111	7	777
Helpers	439	8	3512
Copy block (nine retained totals)	9	7	63
Total	661		5129

3. Total net climb is 5066 = 8R + 14v for R = 439, v = 111. With the recorded nine copy ranks (7 each, 63 = 9 x 7, equal to the recorded decoder constant 63) this gives 5129 = 8R + 63 + 14v, matching word.legality.rank_mass and the identity asserted by the supplier's own verify.sh. The class decomposition of the mass aggregate is word-geometric and had no existence outside the checker chain.
4. Aggregate cross-consistency: word.local_histogram has weighted sum sum(r * n_r) = 5129 equal to the re-derived mass; nine.profile has weighted sum 1808352 = 64W - (W - 63N) with W = 32 * 8 * 111 = 28416 and N = 32 * 9 = 288, i.e. the recorded bit deficit W - 63N = 10272.
5. Five semantic mutation controls are rejected: collapsing the full frame to one label; setting one register's start frame to its final frame; mutating one copy rank to 6; a one-count mutation of local_histogram; a one-count mutation of nine.profile.

Reproduce

From the repository root (Python 3.10+, standard library only, seconds):

python3 -B research/bit-ledger-frame-ranks-seedfeb2/verify.py
python3 scripts/research_package.py --package research/bit-ledger-frame-ranks-seedfeb2 --check-only

Observed on this host (CPython 3.13): PASS_BIT_LEDGER_FRAME_RANKS with 13 checks and 5 rejected controls, and PASS_PACKAGE_METADATA_AND_PINS (not a mathematical acceptance check). Note the repository's make targets are not runnable on this host (no make installed); the underlying Python commands above are the documented steps.

Limits and remaining obligations

• The 2920-record word.local_histogram is not decomposed into per-record MOVE prices. That decomposition lives in the emitter's register-level event stream (COHORT249-RECORDS.bin via stage_cohort.py, priced by legality_h.cpp, which needs a C++17 compiler and Boost); the histogram is only cross-checked at aggregate level. This is the exact obstruction retained from the predecessor's stated blocker.
• The nine copy ranks (7 each) are the recorded word.copy_ranks block, not re-derived from the gate stream.
• nine.profile is checked at aggregate level only, not rebuilt from the 32-replica template expansion.
• Not checked: moment acceptance, H1-H5 interface obligations, hosting/precision admission, the complex ledger, all cold replays. This is a scoped finite audit, not a multiplication theorem or a formal verification.

Overlap and credit

Pre-submission scripts/find_overlap.py --refresh over 364 open PRs and 750 discussion records (2026-10-14T19:15Z) found no competing remote work on this object; matches were only this package and its predecessor's open question. Predecessor credit: the open question and the pricing model are from Navinor's research/selected-suppliers-de-diagnostic/ (building on Douglas Colkitt's pricing model in autoresearch.md and research/pending-claims-independent-pricing); the pinned bit word, its format and its checker chain are the PR 589 by DreamingOfClouds / PR 555 by eumemic lineage (with their credited predecessors, retained unchanged by the selected record); the bit-side coverage identity and the port-direction analysis are ctoth's discussion-1587 finding with its disclosed reviewing models. This package adds the word-geometric decomposition of the paid-rank mass aggregate that none of those record.