Research: audit PR #2809's NODE stock, peel and owned return against PR #2753's logical node

[commdata2338]

Change and mathematical scope

Adds research/selected-node-stock-refinement/, an independent conditional source refinement and finite audit of PR #2809's NODE/peel/family-split and MERGE/root-return boundary against PR #2753's logical node. It takes PR #3041's released enclosing-node handoff. Complete family coordinates, separate fresh stock, borrowed-input preservation and a distinct owned output are proved under the explicit completed-primitive, capacity and successful-execution premises stated in the package's PROOF.md.

The actual quotient is U=R/(W*N*H), while leading=U*B. The inherited K=269241600*N*H reserve therefore yields U=t*9475*K^(d-1) from R=t*K^d. A literal guard control shows why local acceptance is insufficient: R=K passes at n=128, but its rank-42 child has width 84 and receives only 9475 rows, failing the child's 28416-family divisor. PRs #2753 and #3041 already state this distinction; the control is corroboration, not a newly discovered source defect.

No new numerical stage coefficient, complete physical reserve, supplier admission or kappa is claimed. The selected record and all predecessor bytes remain unchanged.

Task and overlap

Task: selected-stock-envelope, track adversarial-audit, scope key f5c32b23bcd01d2e356d, priority revision 6. Base and final refreshed canonical main: 463113d675af7410ac91d53a4fed51ec6494eb9a. Contribution head: b3e192fa4e11555cad1cf9914f6632ad66531ccd. The claim is in discussion #2319, following PR #3041's released handoff.

Dependency	Exact author head
PR #2809, Jose Manuel / jmbarrancoml	e7d93dd0de0bb81308bc3356f65eddeeb22fd712
PR #2753, David Arias	7dba5452a4071a856a0b11cebdcd2d3b6bcd7659
PR #3041, Nick Petalas	b89c9eef729eb1e5889e2095d59a6fbf4738447d
PR #2989, Nick Petalas	133cd8a7c5ce16a3d66e3372c5fdd9e88067d75b

The executable baseline is PR #2809's package as retained on main, pinned through 135 live inputs. Its loader's documented adaptation accepts the current selected record after main's HPE verifier correction. All 121 pinned files from that package equal PR #2809's head by blob, and every retained inputs/pr2753/ file equals PR #2753's head.

Task card:

• Object/question: actual Machine.node, fresh-row quotient, original-offset peel, complete family interleaving and Machine.returned; do these preserve the logical node's full coordinates, fresh/spectator distinction and ownership contract?
• Scope status: changed. The numerical card is narrowed to its released implementation-refinement obligation. The conditional coefficients of PR Prove a conditional GL64 parent-stock envelope for the selected BIT bootstrap #2084 (Adam Gall) and PR Tighten the conditional selected BIT parent-stock bound #2120 (Nick Petalas), the cutoff arithmetic of PR Audit: stock-margin chain closes for the remaining cutoff pairs (no new kappa) #2972 (Hiroki Kumazaki) and the m=73 source audit of PR Selected stock envelope at m=73: bound quantities and the exact missing shape lemma #3113 (Aaron Wang) are already covered.
• Audited implication: PR Independently audit selected compiler ordering and owner guards #2989 assumes completed residuals and checks middle-word ordering; PR Independently prove conditional selected residual/CALL refinement #3041 checks residual/CALL but leaves the enclosing node boundary explicit. This supplies that separate conditional boundary.
• Method/domain: all-width source-level algebra for selected m=64; five complete q=2/3, m=2 Boolean packets; a separate q=5 peel distinguishing 2 from its inverse 3; actual m=64 guard-only calls. New Cartesian and provenance oracles, actual source execution and rejecting controls.
• Milestone/budget: one bounded audit, no saving target or construction-family expansion. Complete physical storage, movement, preparation, H2 and outer-wrapper costs remain separate.
• Success/stop rule: establish one exact source implication and discriminating finite evidence, or report the first failed premise. Stop before GL64 payload construction or a dependent search. This audit reaches the conditional implication.
• Artifact/reproduction: package README and PROOF, source pins, immutable typed receipt, verifier and controls; commands below. The full structured card and remaining questions are in TASK.json.

Overlap used the repository's offline helper for the card and for 3041, NODE/peel, enclosing node, fresh-row quotient/split and nearby compiler terms, plus current-main sources. Entry coverage: 245 open PRs at 2026-10-17T17:04:11.636955+00:00 and 1,985 Discussion records at 17:04:42.538402+00:00. Pre-commit coverage: 251 open PRs at 17:37:12.534077+00:00 and 2,015 Discussion records at 17:39:04.580509+00:00. The final recheck used 253 open PRs at 17:49:11.608413+00:00 and 2,017 Discussion records at 17:52:04.322463+00:00, with complete pagination and no additional matching handoff. A closing check of open-PR changed paths (through #3216, 18:10:48Z) and of PRs opened through #3219 found no other package on this boundary, and canonical main was unchanged at 18:21:51Z. Offline metadata excludes arbitrary diffs and unpublished work and is not a novelty certificate.

PRs #2743 and #2749 (Nick Petalas) inventory other storage categories; this does not repeat those bounds. This account's PR #2160/#2188 descriptor work and PR #2399 substitution work remain credited predecessors. The concurrent claim on PR #2982's ordered high-field ADD handoff targets another primitive's scratch and traffic, not this NODE/return interface.

Validation

python3 -I -B scripts/research_package.py --package research/selected-node-stock-refinement --check-only
python3 -I -B scripts/research_package.py --package research/selected-node-stock-refinement
python3 -B scripts/check_research_commit.py --package research/selected-node-stock-refinement

Observed locally (CPython 3.12.3, Linux x86_64, standard library; the registered replay takes about 95 seconds):

PASS_PACKAGE_METADATA_AND_PINS (not a mathematical acceptance check)
PASS_SELECTED_NODE_STOCK_REFINEMENT cases=5 coordinates=198336 no_new_kappa
Ran 10 tests in 32.302s
OK
PASS_REGISTERED_RESEARCH_PACKAGE research/selected-node-stock-refinement outcome=partial
PASS_RESEARCH_COMMIT_BYTES {"commit": "b3e192fa4e11555cad1cf9914f6632ad66531ccd", "files": 143, "package": "research/selected-node-stock-refinement"}

Evidence includes 198,336 full NODE coordinates, 15,625 separate q=5 peel coordinates, six selected-size stock-guard decisions and 243 exact quotient identities. Controls reject omitted peel, changed family order, missing high fibers or stock, spectators used as fresh stock, wrong old-prefix weights, borrowed output alias, omitted inverse, three source mutations and Boolean/integer receipt drift. The intermediate family comparison prevents split/merge cancellation from hiding an ordering fault.

The exact commit was exported with scripts/export_research_package.py, checked and extracted into a read-only Git-free tree. The optimized verifier and all ten behavioral groups passed there from a separate working directory, and all 143 exported files stayed byte-identical. A copy with a changed exported macro_machine.py was rejected at the source pin before loading. AST parsing, relative links, whitespace and patch integrity were checked. Replay does not regenerate RESULT.json.

Finite small-prime fixtures are not selected large-prime or GL64 payload execution. Complete primitive proofs, universal capacity/backing/movement costs, H2, outer-stage reserve, interruption recovery, selected cold replay, formal-kernel proof, human review and hosted CI are not claimed.

Credit

Jose Manuel (jmbarrancoml) supplied PR #2809's physical compiler, PR #2697's node and PR #2234's compatible stock; David Arias supplied PR #2753's logical caller and its coordinate and stock contracts. Nick Petalas supplied the PR #2989 and PR #3041 ordering and residual/CALL audits and the specific handoff. Credit is also retained for ywz65's parking work (PR #2635); starship-s's earlier parent-identity audit of PR #2809; the PR #589/#555/source527 lineage of DreamingOfClouds and eumemic, with the credited geometry and hosting contributors; this account's earlier PR #2160/#2188 descriptors and PR #2399 substitution; Adam Gall's conditional parent-stock envelope (PR #2084); Josie Elliston's and ToikY's bootstrap audits (PRs #1956/#1965 and #1990); and Aakash Thakkar's and Sagar Thakrar's stock-fit work (PRs #2324 and #2343). The selected assembly retains Rohan Arun's and Jeremy Kothe's credit. Existing notices are unchanged. New package author: commdata2338; Apache-2.0.

AI assistance: submitted by @commdata2338 with substantial AI assistance. OpenAI Codex (GPT-6 Astra) did the derivation, the implementation, the checking and the writing; it was packaged and submitted with Claude Code. No independent human review is claimed.

Open next questions

The package README and TASK.json list two follow-ups. First, independently discharge one remaining completed-primitive or fixed-capacity premise at a selected source call, starting with one reached weighted leaf or route and preserving the whole packet and dirty state; until then the combined conditional node/word/residual implication is not caller admission. Second, join the actual live/backing lifetimes, movements, preparation, H2 and wrapper costs into a complete physical reserve and movement invoice before assigning a numerical selected-stage coefficient. Another cutoff calculation cannot fill that gap, and the present conditional composition does not make those costs zero.

Workflow feedback (optional)

The released PR #3041 handoff made this bounded audit possible. Adding the current physical-caller map and its residual/CALL successor to the stock card's existing_checks would help; its older cutoff-centered entry otherwise requires a long navigation chain to reach the current missing implementation boundary.

🤖 Generated with Claude Code

https://claude.ai/code/session_01X86sCSXdoDd61SDFfC1PKG