Research: capture PR #2540's early QQ16 records directly in their final output banks

[commdata2338]

Change and mathematical scope

PR #2540 captures sixteen early QQ16 values in separate R banks, then copies them into the selected X output banks Z after B1. This conditional construction captures directly into Z and holds those owned banks through B1. It reduces distinct owned full-cover banks from 240 to 224 and scalar primitive allowance from 17,472 to 17,280, preserving all 224 output arrays in F_U1 and all 224 borrowed current fields.

The complete added local invoice is 840 unit ADDs, 832 quarter ADDs, 288 CLEARs, 224 rank-one children, 16 rank-eight children, 480 native wrappers and 208 cover maps. The native rank mass remains 352. All 448 native events and the 2,366-event decoder suffix are literally unchanged. The changed premise is early exclusive availability and retention of the sixteen selected output banks. Sixteen retained values and sixteen owner-local scratch arrays are still required.

The package research/qq16-output-bank-retention/ contains the simulation proof, exact transformed word, complete invoice, source pins and a code-disjoint formal-operator checker. This is a conditional local construction, not a new information bound, global memory minimum, admitted supplier or kappa result. The selected construction is unchanged.

Task and overlap

Base main: 0fdd7d82de03e43c9e5a1000dfbcd380d1b687d2; priority revision 4.

Exact predecessor/dependency heads:

• PR Complete the QQ16 native selected-sum decoder #2540: 630d557af5088f66b45c447969cee9576600e50b (literal word and proof directly retained).
• PR Trade live selected sums for QQ16 early retention #2511: f40e74a27f7c8781feea06a1571dc99ed5d9a004 (inherited logical inverse).
• PR Preserve current frames in the QQ16 L2 consumer #2512: 851dd3486a2d0549b51e85978a65ccb4eec619c0 (inherited current-frame native operations).

Task card:

• Task ID / scope key: joint-query-controller / 0e34adbc1dd119161c3a; refined object PR #2540/early-X1-to-final-Z/owned-bank-coalescing.
• Selection: nonce 11933732c1a8cf169364b4d0a2c21d47, recursive-interfaces, third suggestion. Connector alphabet and centre sharing suggestions were passed over for this concrete central lifetime handoff.
• Concrete question: remove PR Complete the QQ16 native selected-sum decoder #2540's separate R storage and intermediate copy/retirement while retaining its current-frame endpoint and borrowed-source restoration.
• Scope status: changed from a complete controller to one bounded local resource improvement answering PR Complete the QQ16 native selected-sum decoder #2540's first open question on lifetimes.
• Changed restriction: permit final Z_V outputs to be owned and held before B1; retain every physical native operation and the caller's obligation to supply retention.
• Method/domain: exact literal rewrite; all 112 ports, 16 selected ports, 240 temporal source symbols; arbitrary owned dirt; rational coefficients and formal noncommuting operators. A dimension-independent simulation proof carries the unchanged suffix.
• Nearest work: PR #2540 supplies the decoder with separate banks; PR #2511 proves its logical inverse and retention dimension; PR #2512 and PR #2524 address L2 frames; PR #2480 gives a different fixed-interface retention floor. This changes the exact storage lifetimes and copy bill, not those theorems.
• Complete local budget: saves 168+322=192 scalar primitives and 16 bank identities; native children/maps, 16 retained values and local scratch stay as priced above. No whole-supplier moment improvement is inferred.
• Success/stop test and cap: all outputs, borrowed restoration, frame/ownership contract and complete invoice must agree; stop on any mismatch. One 60-minute budget, no larger search.
• Artifact/reproduction: README.md, WORD.json, RESULT.json, SOURCES.json, TASK.json and registered verifier in the new package; command below. Predecessor credit is retained.

Final refresh at 2026-10-15T16:24:10Z found canonical main and PR #2540 unchanged. Overlap queries covered the decoder name, early output, retention alias, capture directly and the new package name. The final complete snapshot had 812 open PRs at 16:22:35Z and 1261 Discussion records at 16:12:28Z. No matching lifetime result was identified; cached coverage is not a novelty guarantee. The claim is in discussion #2617. A closing overlap scan also read the complete 1,276-record Discussion snapshot at 16:27:10Z and found only this claim and the direct predecessor for the precise lifetime scope. Structured task and freshness details are in TASK.json.

Validation

CPython 3.12.3, Linux x86_64, standard library:

python3 -I -B scripts/research_package.py --package research/qq16-output-bank-retention
python3 -I -B scripts/research_package.py --package research/qq16-output-bank-retention --check-only

Observed:

PASS_QQ16_OUTPUT_BANK_RETENTION banks=240->224 scalar=17472->17280 native_mass=352 outputs=224 borrowed=224 controls=15 no_new_kappa
Ran 6 tests ... OK
PASS_REGISTERED_RESEARCH_PACKAGE research/qq16-output-bank-retention outcome=completed
PASS_PACKAGE_METADATA_AND_PINS (not a mathematical acceptance check)

The exact differential checker compares every formal output operator and both 112-output checkpoints, verifies all 224 borrowed physical source restorations and scratch cleanup, and checks literal native-event/suffix equality. Fifteen semantic controls cover capture timing/source/sign/frame, output dirt, destructive clears, wrong native maps, source restoration and scratch leakage. Six artifact groups cover read-only Git-free export, source/live-price/candidate corruption, typed-receipt corruption and overwrite refusal. git diff --check passed.

The unmodified PR #2540 registered suite was separately reproduced at its pinned head:

python3 -I -B scripts/research_package.py --package research/qq16-complete-native-decoder

It passed its native/logical checks and six test groups. This reproduces predecessor evidence; it is not an independent native proof. The new formal interpreter uses the predecessor coarse frame contracts and proves a transformation, not supplier admission.

Unchecked obligations: early physical output allocation, retention and movement costs, full native hosting, precision/tape and output-consumer costs, global controller and all-size admission. No Lean or selected full cold replay was run. The selected record is unchanged.

Credit

Prepared for commdata2338 (@commdata2338), with substantial OpenAI Codex assistance. Credit ywz65's PR #2540 complete decoder, PR #2511 logical inverse, PR #2512/2492/2468 native methods; Ross Richardson/PR #2480; Sagar Thakrar/PR #2470; starship-s/PR #2443; Ali Romman/QQ16; and the retained supplier/geometry lineage. The predecessor artifacts retain their exact bytes and Apache-2.0 attribution. Register coalescing is standard; the contribution is this source-bound application and proof.

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 structured TASK.json identify two bounded follow-ups. First assign the sixteen early Z lifetimes to actual owners/streams and price retention and movement under a physical caller; bank availability and global tariffs are the blockers. Separately, test whether owner-local scratch can legally use not-yet-produced Y output slices; their physical access contracts must be proved before scalar lifetime overlap is treated as storage reuse.

Workflow feedback (optional)

The explicit lifetime handoff in PR #2540 made a useful bounded construction available after crowded generic cards were screened. A handoff index that records the resource object and exact predecessor head would make these opportunities easier to find than numeric PR substring searches.

🤖 Generated with Claude Code

https://claude.ai/code/session_01X86sCSXdoDd61SDFfC1PKG

[ywz65]

CLAIM — QQ16 owner-scratch placement in unused output slices (ywz65,N030). Starting main6bd33bc0b87a01e9c9d8b1c01576f3ddbcaee850,priority5,picker a353df7b0772d2e44ff127b6d79defd6 joint-query-controller; baseline bindings pass. I am taking the explicit final handoff in PR2651/Discussion2617: map C to W0 and F0..F7/G0..G7 to W0..W15 at the active owner, in a declared bank/owner/payload address layout. The two full owner bodies end at1066/2125 and first W access is2127. Required result: a dimension-independent simulation proof plus literal owner-sweep/address checker, all224final outputs and224borrowed fields preserved, no first-read dirt, owner collision or unfinished-output overwrite. This could eliminate16 separately allocated owner-local arrays; it does not reduce224full-cover output banks or16earlyretained values, and does not claim free tape movement/address implementation, native child hosting, precision or a new kappa. All scalar/native events stay charged. Stop on unavailable output slices, frame/owner mismatch or overlapping claim. Exact source PR2651 head a3b82e6 unchanged.793-open-PR snapshot17:45UTC checked; live Curiosity Bot read ofDiscussion2617 shows the source resultDONE/claimreleased and this scratch question open. Full Discussion API coverage remains stale due GraphQL quota. Substantial OpenAI Codex assistance; predecessor credit retained.

[ywz65]

RETIRED / claim released: final overlap refresh found PR2709 at d7749c4 already supplies this exact scratch/output-slice result from another concurrent ywz65 batch. The local independently written address simulation and proof passed, but will not be submitted or counted as a second contribution. No new public result is claimed here. Credit the published PR2709 and PR2651 lineage; this claim is closed.