[Clear1nsight]

Change and mathematical scope

No physical pair is supplied and no new κ is admitted. One hypothetical legal H1+H1→H2 fusion in each of 18 native occurrences changes the complete p16/a7 paid invoice by exactly −36 H1+18 H2, with unchanged productive targets and rank mass. Under the retained source/finite/precision/compiler/outer cost contracts, the independently reviewed fixed scalar screen supports the construction target κ=0.0030220, tested at 0.00302200001 with reserve 10^-11.

This is a changed-profile target and concrete construction handoff, not an improvement from more digits of an unchanged root. It is above the frozen #4610 value 0.0030217758; the then-selected #4470 record was unchanged at submission. Canonical main 1ec1119972c9aefe42525172d07b332fcff81cb4 has since independently selected #4610 at .0030217758 through #5241; this package supplies no admission or selection change. At #5226's larger fixed bar 0.0030233, this exact type of change needs five fusions per native—90 global occurrences—for positive raw moment. That sign is not compiler or native admission for the larger bar.

Task and overlap

task_id: independent-moment. Scope change: source-bound verification of #5226/#4610's fragmentation normalization and one fixed compiler point for the hypothetical −36 H1+18 H2 profile, replacing any unchanged-root refinement or broad physical supplier claim. Only the fixed p16/a7 invoice is screened; exhaustive all-controller maxima, unshipped searches, broader centre/coded-family claims and physical pair discovery are outside this submission.

Sources: DreamingOfClouds' #5226 head dc5764630fdf34f72bbe76efd4b79d3b2bc3dd40; Cao Yifan/intqwq's #4610 head 20bb93cd990d2ff683e4d0fc56b87a42cd5d3296. Publication base ff3cc58e370d0f678e529246f8c73e8094e18fcf was refreshed via the sole monitor/canonical fetch; governing guidance and selected record are unchanged. No duplicate PR-feed intake or source search occurred. A fresh bounded all-author check through 13:46 UTC found no matching target. The separate #5241 submission concerns admission of the existing #4610 value and does not supply this hypothetical physical delta.

The original author report 828b0cfa56c550c8024c8b3e0d68c39e2ac1222b5b74200ebfe11e73b3db2413 and independent review ac0d4444aac62d3c857de1e32c427226dd22202c5525b49005308d0fa11d0759 are preserved unchanged. The entire 28-file independent packet embeds the exact original 17-file author tree once. All eight source aliases bind exact raw Git blobs and repository paths; no source/receipt normalization or second old-package copy is used.

Validation

python3 -I -W error -B research/one-fusion-compiler-target/verify.py
python3 -I -O -W error -B research/one-fusion-compiler-target/verify.py
python3 -B scripts/research_package.py --package research/one-fusion-compiler-target --check-only
python3 -B scripts/research_package.py --package research/one-fusion-compiler-target
python3 scripts/check_research_commit.py --package research/one-fusion-compiler-target --ref HEAD --check-source-history

Direct/registered normal/O, metadata/pins, committed bytes/source history, diff check, official export and fresh Git-free read-only replay pass. Full author merge/point receipts and the independent positive-Taylor receipt regenerate byte for byte. The distinct reviewer engine imports no author interval functions. Static GF(2) frame independence/nesting, forward/backward event counts, profile incidence, rank-mass conservation and helper/centre credit checks pass. This does not evaluate amplitudes, absolute phase, contact legality or raw-dual operator correctness. The vendored #5226 verifier, Lean source and round6 generator are reference inputs, not executed.

The fixed scalar screen retains 1024 copies, cutoff/floor 20/21, tail offset 1/5, caps .97 over 128 blocks, maximum productive rank 145, theta 63/64 and power 1000. It verifies the inherited gap/height and the retained pure outer ledger's 54 strict plus four functional positives, including numeric sensitivity controls. These are conditional arithmetic applicability checks, not proof that a changed supplier satisfies C1=1, source normalization, precision/template closure, ordinary BIT/stock or the nine outer cost semantics. Local validation ran no native, Tensor/compiler/emitter/Lean or full helper suite. The first hosted run 39964162361 succeeded on 2026-10-23 at 13:54:01 UTC, testing merge 9f4702ca5419a78e3b41648e83bdb7769769fbbd for this exact contribution. Planning passed 420 helper tests with 17 skips in 216.718 seconds. Registered normal and optimized replay reported PASS_CONDITIONAL_ONE_FUSION_CONSTRUCTION_TARGET_ONLY and passed committed-input checks; both receipts retain physical_pair_supplied=false and compiler_executed=false.

Contribution head: cc8cb58ab4083c1cfd58e0e2eb17c0b2a88070a9.

Credit

DreamingOfClouds for #5226 fragmentation/fixed bar; Cao Yifan/intqwq for #4610's changed native profile/compiler tightening; Cinna Davis/1337isnot1337 for native/controller lineage; zxv and Jacob Alan Sussman for paid tensor completion/mixed-job compilation. Retain Dugongue, dillona, Clear1nsight, SethBurkart123, Ryanson Jonathan and all native/helper/compiler/precision contributors. Source-bound author review, independent positive-Taylor review and Chris R/Clear1nsight packaging used substantial OpenAI Codex assistance. Apache-2.0 and original notices preserved. No independent human or proof-kernel acceptance is asserted.

Open next questions

Supply one literal eligible native pair with exact event positions, source field/charts, ordered operator and absolute phase, all 632 arbitrary inputs/all ports, both forward and raw inverse-transpose orientations, helper/complement returns and spectators. Prove it can be reused at each of 18 occurrences, or reprice actual partial incidence. Matching rank labels or a standalone Gaussian factorization is insufficient.

Then preserve local scalar 158168, routing 4802000 and whole-controller scalar 181098496 bounds, or supply replacement costs. Keep coefficients/denominators, owned fields/lifetimes, precision/completed-child/template and actual C1/ordinary/outer semantic obligations explicit. If a target, field, correction or bill changes, recompute affected profile/guards before claiming funding. No pair is selected here.

[Clear1nsight]

HANDOFF

Research discussion / PR / artifact:
Published #5243, module research/one-fusion-compiler-target, exact fork branch research/one-fusion-compiler-target-20261023. #5226 verification extension and conditional construction target. Task ID independent-moment, changed hypothetical paid profile, one fixed point.

Exact commit:
cc8cb58ab4083c1cfd58e0e2eb17c0b2a88070a9; base ff3cc58e370d0f678e529246f8c73e8094e18fcf; #4610 20bb93cd990d2ff683e4d0fc56b87a42cd5d3296; #5226 dc5764630fdf34f72bbe76efd4b79d3b2bc3dd40. Author report SHA256 828b0cfa56c550c8024c8b3e0d68c39e2ac1222b5b74200ebfe11e73b3db2413; independent review ac0d4444aac62d3c857de1e32c427226dd22202c5525b49005308d0fa11d0759.

What is established:
Under retained contracts, one hypothetical legal pair per native occurrence gives whole invoice −36 H1+18 H2, supporting the fixed candidate .0030220/tested .00302200001 scalar screen. Exact intervals, full pre-test monotonicity, reserve/tails/height and 54+4 pure-ledger arithmetic checks pass; physical pair MISSING. Five pairs per native/90 global are needed only for positive raw sign at .0030233. Full targets/helpers credited once; no free deletion or extra port multiplier. Source raw blobs, whole 28-file review packet/embedded author tree, normal/O full receipts, registered replay, committed bytes/history, official export and Git-free read-only replay pass; first hosted run 39964162361 succeeded at 13:54:01 UTC: planning 420 tests/17 skips/216.718s, registered normal/O scalar checks and committed-input integrity pass. Both receipts explicitly retain physical pair missing and compiler not executed.

What remains uncertain:
Actual source pair/all-input/all-port/phase/raw-dual identity, all 18 occurrence reuse, contact legality, unchanged scalar/routing/denominator/precision bounds, C1=1 and completed-child/template closure, ordinary BIT/stock and outer semantic applicability. Static frame/event data do not prove those operator contracts. No native/compiler/emitter or full helper suite, actual supplier or new κ admission. This package changes no selection. Canonical main 1ec1119972c9aefe42525172d07b332fcff81cb4 separately selects the existing #4610 value through #5241; the .0030220 target remains hypothetical and above that baseline.

Smallest useful next task:
Name one literal H1 pair, its exact field/event/charts and phase-correct operator replacement by H2. Verify all 632 inputs, both orientations, helpers/returns/spectators and actual incidence. Then bind all finite and semantic cost contracts; reprice any new field/correction/cost before funding. Do not infer legality from labels or a free-standing factorization.

Reproduction command and expected runtime:
python3 -I -W error -B research/one-fusion-compiler-target/verify.py, repeat with -O; about 4 seconds per mode, standard-library CPython 3.12. Registered: python3 -B scripts/research_package.py --package research/one-fusion-compiler-target. Private unchanged merge/point/independent engines must reproduce every full frozen receipt. Only the authenticated private author directory is added for its sibling import under -I. No writes to package/export, no native/Lean/#5226 verifier execution or shared 414 helper replay.

Predecessor credit:
DreamingOfClouds, Cao Yifan/intqwq, Cinna Davis/1337isnot1337, zxv, Jacob Alan Sussman, Dugongue, dillona, Clear1nsight, SethBurkart123, Ryanson Jonathan and retained native/helper/compiler/precision lineage. Source/independent review and packaging with substantial OpenAI Codex assistance; original licenses/notices retained. No independent human or kernel acceptance claim.

{
  "schema": "integer-mult-coordination/v1",
  "kind": "HANDOFF",
  "task": "independent-moment",
  "scope": "One fixed conditional compiler point for one hypothetical native H1+H1 to H2 fusion per occurrence, −36H1+18H2 complete delta; physical witness missing.",
  "base": "ff3cc58e370d0f678e529246f8c73e8094e18fcf",
  "expiry": null,
  "artifact": "https://github.com/CrocSwap/integer-mult-bounds/pull/5243"
}