Independent verification (no new κ): record candidate #2269 T74 borrow43 — words, 437-helper fill and moment root reproduce; stale evidence, unpriced buffers, X2 open

[thakrarsagar]

Status: independent verification. The construction claims of record candidate #2269 reproduce, and no counterexample was found. Evidence, metadata and accounting defects are listed below. There is no new κ, no frontier change, and no endorsement of κ beyond what is checked here.

Object

#2269 (jeremykothe2, head b479fc5567bc09ab9185bffe89b62594f36d74e5, research/gmx-t74-borrow43/) claims conditional κ = 0.0016962471390584742…, which is +1.4% over the selected record. It adds 43 borrowed-helper aliases from #2226 (Cinna Davis' anticipatory dirty borrowing) to the T74 pivot-566+554 words and re-fills the 666-call hosting.

When this check started, nothing committed independently checked the new 437-helper fill. The shipped evidence/INDEPENDENT.json describes a different, pre-borrow 480-helper fill.

What reproduces, using fresh stdlib code

The new code imports, executes and copies no code from #2269, #2226 or #2260. The longest shared token run against 19 PR scripts is 27 tokens. All 40 inputs are pinned by git blob id and sha256.

Words. For both BORROW-WORD-43{base,extra}:

• Warps: all 43 have G invertible over GF(2), dim U = dim E, G·span U = span E, and K_G mapping L_U to L_E, so the Fourier rank is 0.
• No anticipatory terms: every alias is a "high" alias.
• Frames: frame climbs are nested.
• Exact replay over ℚ⟨F₄₃⟩, with arbitrary dirty registers:
  • Yᵢ ← Yᵢ + Xᵢ;
  • the other 549 live registers and the 76 zero buffers are restored;
  • the dual orientation also replays.
• Recounted profiles:
  • 43base {1:2351,2:501,3:414,4:129,5:42,8:8};
  • 43extra {1:2339,2:509,3:418,4:125,5:42,8:8};
  • mass 5385, 437 helpers, widths {4:36,5:9,7:36,8:116,9:240}.

Fill. For OWNERS.json.gz, 291,042 rows:

• exactly one role per (call, live helper), and the removed helpers are absent;
• widths match the words;
• 1,332 windows are rebuilt from the event stream, they equal PLAN, and every role lies strictly inside its window;
• families and namespaces are correct, with no role on the caller's active pair;
• slices are disjoint, within their child, and prefix-filled across 149,184 bins;
• the idle recount is {1:23976,3:8673,5:813,7:150,48:24479}.

Price.

• The global profile composed from our recounts equals PROFILE.json.
• A fresh certified atanh/exp evaluator brackets the moment root to width 1e-130. The bracket contains the claimed b, and complex_lower + 1e-90 is rejected.
• The κ arithmetic from b is checked.

Positive controls.

• The canonical words reproduce the old {5:82} profiles.
• The T37 parent fill reproduces its own spec.
• Every script has targeted negative controls.

The gain attributable to borrowing is Δb = 1.64922·10⁻⁵ over #2044's b.

Defects and open obligations (none changes b as checked)

1. No admission receipt for the borrowed supplier.
   • 8 evidence and metadata files are byte-identical to the pre-borrow Conditional κ = 0.00167975495612554452133567099392070257699831143850258029546032752016286454826032563882630565613284: recertified 1e-30000 supplier endpoint and 1e-30010 slacks (kappa +2.00336e-500 over PR1897): fine source-prime padding (… #2044 package, including evidence/INDEPENDENT.json, evidence/ASSEMBLY.json, evidence/FULL-ADMISSION.json, PRICE.json and the top-level SPEC.json.
   • The top-level PLAN and SPEC disagree with each other.
2. The 76 cleanup zero buffers per call are unpriced.
   • Conditional κ = 0.001696247139058474: capped phase-cell fusion outer (PR1673) on the T74 supplier with anticipatory dirty borrowing (PR2226, 43 high-helper aliases) on the T74 pivot-566+554 supplier: conditional kappa 0.0016962471390584742… #2269's emitter allocates them eagerly, giving a peak of 513 fields against 480 unborrowed.
   • physical_stock_identity has no buffer term. Improve conditional κ to 0.0016805 with selective dirty borrowing and physical owner repacking #2260 charged its buffers, and Research: 80-helper anticipatory borrowing component and exact price rejection (no new kappa) #2226 PROOF §4 lists this as open.
   • Charging them as dedicated families leaves the finite constants unchanged. Hosting them in idle children would move b; that case was not evaluated.
3. X2 is not checked. That is pending-axis disjointness across children sharing a family, and it applies to 37,408 of 54,145 hosted families. The 73-column charts are not checked either. A failure would change the idle profile and therefore b.
4. Metadata errors:
   • CLAIM.json sets construction_improvement: false;
   • the gain field mixes the outer and strict κ;
   • the README's lineage and width wording is wrong;
   • stale numbers remain in the outer notes.

Not checked

• the outer stack behind κ = b (PR1673, PR1758, PR1786 and the union schedule); only b → κ arithmetic is checked here, and Audit (no new κ): side-condition coverage of the pending T74 outer at #2044 — 554 written conditions, none load-bearing false; or-True and g² review items resolved #2286 inventoried that outer at Conditional κ = 0.00167975495612554452133567099392070257699831143850258029546032752016286454826032563882630565613284: recertified 1e-30000 supplier endpoint and 1e-30010 slacks (kappa +2.00336e-500 over PR1897): fine source-prime padding (… #2044's point;
• the native lowering and the exact Clifford lifts of K_G; the borrowing gain rests on the rank-zero handoffs, and charging warps as plain frame changes would give local mass 5581;
• the 47 strict inequalities and the bootstrap;
• inherited T74 and PR589 premises;
• successor Conditional κ = 0.001696310950415050: exact frame refit of the borrowed gauge44 word (WildIdeas), 1e-3000 certified supplier endpoint, fixed-grid-maximum kappa C - 1/F - B at 1e-3080 slacks (PR2180 / PR2161 retune) and exact-integer coarse… #2288's refit gauge44 word.

Reproduce

python3 -I -B research/t74-borrow43-independent-check/check_all.py --fetch

This takes about 35 s. --fetch fetches refs/pull/2269/head only if b479fc55 is absent; every input is still verified against its pins. Expected last line:

PASS_T74_BORROW43_INDEPENDENT_CHECK words=PASS fill=PASS price=PASS runner_controls=3 commit=b479fc5567bc mode=git

Checks run locally:

• the runner passes from the repo root and from /tmp;
• research_package.py --check-only passes;
• the 65 test_research_* tests pass;
• the Python 3.12 registered replay was not run (3.12 is not installed), and the scripts parse under the 3.10 grammar.

Provenance and credit

• Base: 0fdd7d82 (priority revision 4), unchanged at submission (2026-10-15 04:53 UTC).
• Track and claim: track 75–84, OS seed 76. No sprint coordination issue exists, so the claim was recorded locally. Card: independent-moment, refined.
• Credit:
  • Jeremy Kothe (Conditional κ = 0.001696247139058474: capped phase-cell fusion outer (PR1673) on the T74 supplier with anticipatory dirty borrowing (PR2226, 43 high-helper aliases) on the T74 pivot-566+554 supplier: conditional kappa 0.0016962471390584742… #2269, T74 lineage);
  • Cinna Davis / 1337isnot1337 (Research: 80-helper anticipatory borrowing component and exact price rejection (no new kappa) #2226);
  • Dugongue (Improve conditional κ to 0.0016805 with selective dirty borrowing and physical owner repacking #2260);
  • the T74 lineage (PR1319/1625/1297);
  • icekylinx (checker lineage).
• Related: Independent verification: supplier-admission premises of the #2265 substitution traced into PR2260 (0 FALSE / 0 MISSING-PROOF; no new kappa) #2311, from the same account in a separate session, independently audits the different Improve conditional κ to 0.0016805 with selective dirty borrowing and physical owner repacking #2260/Outer retune of #2260: #1673 capped phase-cell fusion outer on the selective-borrowing supplier, conditional κ = 0.00168341258854… #2265 T8 supplier; it is complementary.

Prepared for Sagar Thakrar with substantial AI assistance (Claude Code, Anthropic). Separate agents wrote the three checks; two adversarial reviewers checked independence and coverage; the main session reran every command. Apache-2.0.

🤖 Generated with Claude Code