Tooling: 4.78x faster exact bounded quadratic-code replay

.github/workflows/cube-xor-replay.yml

@@ -0,0 +1,27 @@
+name: Exact cached cube XOR replay
+on:
+  workflow_dispatch:
+  pull_request:
+    paths:
+      - 'research/cube-xor-replay/**'
+      - '.github/workflows/cube-xor-replay.yml'
+  push:
+    branches: [main]
+    paths:
+      - 'research/cube-xor-replay/**'
+      - '.github/workflows/cube-xor-replay.yml'
+permissions:
+  contents: read
+jobs:
+  exact:
+    runs-on: ubuntu-24.04
+    timeout-minutes: 15
+    steps:
+      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1
+        with:
+          persist-credentials: false
+      - uses: actions/setup-python@5fda3b95a4ea91299a34e894583c3862153e4b97
+        with:
+          python-version: '3.12'
+      - name: Every scheduled row and complete receipts
+        run: python -I -O -B research/cube-xor-replay/verify.py

research/cube-xor-replay/BENCHMARK.json

@@ -0,0 +1,51 @@
+{
+  "case": "t6",
+  "fresh_process_each_sample": true,
+  "measured_utc": "2026-10-13T17:38:00.253806+00:00",
+  "median_seconds": {
+    "cached": 18.35716083389707,
+    "original": 87.66262483294122
+  },
+  "platform": "macOS-26.5.2-arm64-arm-64bit-Mach-O",
+  "python": "3.14.6 (main, Jun 10 2026, 10:03:53) [Clang 21.0.0 (clang-2100.0.123.102)]",
+  "samples": [
+    {
+      "backend": "original",
+      "pair": 0,
+      "receipt_sha256": "1c790023631dc5390ed599f499b0d660d5099ee20427aadb894087163dba9186",
+      "seconds": 87.68134720891248
+    },
+    {
+      "backend": "cached",
+      "pair": 0,
+      "receipt_sha256": "1c790023631dc5390ed599f499b0d660d5099ee20427aadb894087163dba9186",
+      "seconds": 18.35716083389707
+    },
+    {
+      "backend": "cached",
+      "pair": 1,
+      "receipt_sha256": "1c790023631dc5390ed599f499b0d660d5099ee20427aadb894087163dba9186",
+      "seconds": 18.350820249994285
+    },
+    {
+      "backend": "original",
+      "pair": 1,
+      "receipt_sha256": "1c790023631dc5390ed599f499b0d660d5099ee20427aadb894087163dba9186",
+      "seconds": 87.66262483294122
+    },
+    {
+      "backend": "original",
+      "pair": 2,
+      "receipt_sha256": "1c790023631dc5390ed599f499b0d660d5099ee20427aadb894087163dba9186",
+      "seconds": 87.64076658303384
+    },
+    {
+      "backend": "cached",
+      "pair": 2,
+      "receipt_sha256": "1c790023631dc5390ed599f499b0d660d5099ee20427aadb894087163dba9186",
+      "seconds": 18.381115874974057
+    }
+  ],
+  "scope": "Wall time includes imports, all source checks, base generation, every scheduled row, ordered elimination, complete receipt check and write; mask cache starts empty each sample. No concurrent replay launched by this harness.",
+  "speedup": 4.775391228858738
+}

research/cube-xor-replay/CLAIM.json

@@ -0,0 +1,38 @@
+{
+  "status": "completed; one open tooling-improvement PR submitted for review",
+  "track": "verification engineering",
+  "assignment_seed": 94,
+  "randomness": "Python secrets.randbelow(100), operating-system randomness",
+  "base_commit": "2921874911956718fcaaaf9396fe8267b3cfa9f0",
+  "started_utc": "2026-10-13T17:24:36Z",
+  "claimed_utc": "2026-10-13T17:28:07.922857Z",
+  "expected_finish_utc": "2026-10-13T17:56:36Z",
+  "verification_reserved_from_utc": "2026-10-13T17:48:36Z",
+  "candidates": [
+    {
+      "object": "historical block-label atanh interval evaluation",
+      "decision": "Not chosen: PR667 already supplies the same integer-Horner mechanism."
+    },
+    {
+      "object": "quadratic-code XOR row permutations in scripts/experiments/cube_core_family.py",
+      "decision": "Chosen: eliminate repeated mask construction while preserving every generated row and ordered elimination. Largest admitted t=6, default 4096-row cap; t=3,4,5 full historical certificate as regressions."
+    },
+    {
+      "object": "historical additive-label modular elimination",
+      "decision": "Not chosen: small workload and adjacent modular-rank tooling PR976; prefer a distinct kernel."
+    }
+  ],
+  "method": "Precompute immutable exact adjacent-block swap masks per replay; compare complete original and accelerated receipts and every generated row. Alternating same-input fresh-process benchmarks. No result caching or row omission.",
+  "expected_deliverable": "One opt-in source-pinned replay adapter, exact-equivalence argument, reproducible before/after benchmark and corruption controls; no kappa or new impossibility claim.",
+  "overlap_checked": "757 open PRs through #1060; exact xor_permute search empty; cube_core search only broad encoding PR946. PR752 quadratic-affine dot-product packing, PR898 Leech minors, PR667 logarithms and PR976 joint-frame rank concern different kernels.",
+  "negative_results": "Historical cube note excludes specified full-support quadratic cores t=3,4,5, not all supported matrices/subcodes. This does not rerun discovery or claim t=6 exclusion.",
+  "attribution": "Ross Richardson with substantial OpenAI Codex assistance; original cube audit by Douglas R. Colkitt. Apache-2.0.",
+  "result_utc": "2026-10-13T17:43:20.653820+00:00",
+  "pull_request": "https://github.com/CrocSwap/integer-mult-bounds/pull/1075",
+  "submission_base": "abdc6a7002abdbcd9e21bfc5e1a795784deacfdf",
+  "verification_started_utc": "2026-10-13T17:38:42+00:00",
+  "result": "Exact cached mask replay: median 87.6626 -> 18.3572 seconds (4.7754x), three alternating fresh-process pairs with identical complete receipts. Full differential comparison covers 4324 rows / 8596927616 coefficients. 17 package tests, 9 inherited tests, Python 3.12 replay, 38 source pins and selected integrity pass. t=6 rank lower bound 3088; no exclusion and no new kappa.",
+  "final_overlap": "Refreshed open queue through PR1072 immediately before submission; exact cached XOR/xor_permute searches empty; cube_core_family search only PR946. No pinned coordination issue still exists.",
+  "labels": "Neither codex nor codex-automation is available in the repository; no new labels created.",
+  "remaining": "Independent review and hosted Linux CI. No full repository suite, native cold replay, Lean or all-size/compiler/analytic proof claimed."
+}

research/cube-xor-replay/COMPATIBILITY.json

@@ -0,0 +1,37 @@
+{
+  "files": {
+    "scripts/audit_fused_block.py": {
+      "current": {
+        "git_blob": "41221dadc99d80952a000ab166ed3471df610405",
+        "sha256": "6fe64f886309d9154e33a3d67a16e8714cebc424a7a248b3080f8cab7d94179b"
+      },
+      "original": {
+        "git_blob": "c63869d22c6454cf6989bb6e70d41f0571eef985",
+        "sha256": "2afb83ca8e5f193ca8a4c242becf267ad1d9dc2578ba238b820e98d5185f6288"
+      }
+    },
+    "scripts/exclusion_circuit.py": {
+      "current": {
+        "git_blob": "b0e2f3e03d7de791ca936dfd92cee2134c7dd4f5",
+        "sha256": "f9fbcd6e85e442cbd0926ac0887414e20735b2dc923d2d5822c803ceb46726d3"
+      },
+      "original": {
+        "git_blob": "aaa99986c4b81c5d4fbe016fd216a4818940bb95",
+        "sha256": "b253f9d7318becda1c1f3b24e2287ed44a9c27fcea00cf9d531afb950d0457a0"
+      }
+    },
+    "scripts/shared_point_circuit.py": {
+      "current": {
+        "git_blob": "d0864271884398d44ee1fc70f0d99bb982418156",
+        "sha256": "360d9bffc0f55315fe294789b72c71c79e0991524555b3a5f42a2ca504d7da95"
+      },
+      "original": {
+        "git_blob": "8347a0bda63fbfb51b86c6c6d10a7bf541816308",
+        "sha256": "c9bb7a5df92d260a4e3ad532b53e26936730d7c93f7facfb0784957c1e81c8ba"
+      }
+    }
+  },
+  "original_manifest_sha256": "e2be1e7c88e7a6814ec726bd5cd397af261514edac960d1d8ae6590c39c9e608",
+  "scope": "Three transitive import dependencies changed on canonical main. Their changed methods are not called by the cube replay. Core, row schedule, elimination, certificates, fixture and original source manifest remain frozen.",
+  "upstream_commit": "ffbd90dd90dbe0055464f3a1411b25a3a4fdc6b8"
+}