Mister-Meeseeks · Oct 14, 2026 · Oct 14, 2026
diff --git a/research/p112-reader-memo-seed92/BENCHMARK.json b/research/p112-reader-memo-seed92/BENCHMARK.json
@@ -0,0 +1,158 @@
+{
+  "base_commit": "9ef085e6c07c8590329ad00b7eac382aef06124c",
+  "certificate_sha256": "d1140c3af508b420ab63b1de5fb29af9ed47631492ab1eb7d5b19b06a7d83646",
+  "conflicting_kernels": 128,
+  "environment": {
+    "implementation": "CPython",
+    "machine": "arm64",
+    "python": "3.12.13",
+    "system": "Darwin"
+  },
+  "excluded_runs": "Scalar-cache prototype and earlier table-preparation pilots are not benchmark evidence. The preceding five-pair run overlapped an untimed counting replay and is excluded.",
+  "fixture": "Entire unchanged PR845 run() and its committed RESULTS.json; no sampled sign cases.",
+  "inherited_controls": 9,
+  "measurement_protocol": "Five pairs alternate reference/memo and memo/reference. Each engine gets a fresh interpreter and cache; no other agent-launched CPU workload ran concurrently. Filesystem caches were not flushed. No process affinity or resource limits beyond the 120-second worker timeout. Ratios are descriptive, never a mathematical pass gate.",
+  "median_process_seconds": {
+    "memo": 1.4025932499789633,
+    "ratio": 1.401297976445339,
+    "reference": 1.9654510829714127
+  },
+  "median_run_seconds": {
+    "memo": 1.3471562499762513,
+    "ratio": 1.4191348442594394,
+    "reference": 1.9117963750031777
+  },
+  "package_controls": [
+    "ordered basis permutation",
+    "in-place basis mutation",
+    "in-place label mutation",
+    "uncached X coordinate mutation",
+    "uncached coefficient mutation",
+    "integer-equal float basis still rejects",
+    "integer-equal float label still rejects",
+    "bounded eviction and recomputation",
+    "omitted constraint row",
+    "exchanged sign-row order",
+    "erased reader values",
+    "type-sensitive rank receipt",
+    "omitted live source pin"
+  ],
+  "pairs": [
+    {
+      "memo": {
+        "cache": {
+          "currsize": 1,
+          "hits": 3272,
+          "maxsize": 4,
+          "misses": 1
+        },
+        "process_seconds": 1.3969656249973923,
+        "run_seconds": 1.3417054579476826
+      },
+      "order": [
+        "reference",
+        "memo"
+      ],
+      "reference": {
+        "cache": null,
+        "process_seconds": 1.9654510829714127,
+        "run_seconds": 1.9117963750031777
+      }
+    },
+    {
+      "memo": {
+        "cache": {
+          "currsize": 1,
+          "hits": 3272,
+          "maxsize": 4,
+          "misses": 1
+        },
+        "process_seconds": 1.4423786249826662,
+        "run_seconds": 1.3871181670110673
+      },
+      "order": [
+        "memo",
+        "reference"
+      ],
+      "reference": {
+        "cache": null,
+        "process_seconds": 1.9578862090129405,
+        "run_seconds": 1.9031278339680284
+      }
+    },
+    {
+      "memo": {
+        "cache": {
+          "currsize": 1,
+          "hits": 3272,
+          "maxsize": 4,
+          "misses": 1
+        },
+        "process_seconds": 1.4025932499789633,
+        "run_seconds": 1.3471562499762513
+      },
+      "order": [
+        "reference",
+        "memo"
+      ],
+      "reference": {
+        "cache": null,
+        "process_seconds": 1.9603819170151837,
+        "run_seconds": 1.90456820797408
+      }
+    },
+    {
+      "memo": {
+        "cache": {
+          "currsize": 1,
+          "hits": 3272,
+          "maxsize": 4,
+          "misses": 1
+        },
+        "process_seconds": 1.430742874974385,
+        "run_seconds": 1.376137999992352
+      },
+      "order": [
+        "memo",
+        "reference"
+      ],
+      "reference": {
+        "cache": null,
+        "process_seconds": 1.9686416669865139,
+        "run_seconds": 1.9120537909911945
+      }
+    },
+    {
+      "memo": {
+        "cache": {
+          "currsize": 1,
+          "hits": 3272,
+          "maxsize": 4,
+          "misses": 1
+        },
+        "process_seconds": 1.3938636250095442,
+        "run_seconds": 1.3381605410249904
+      },
+      "order": [
+        "reference",
+        "memo"
+      ],
+      "reference": {
+        "cache": null,
+        "process_seconds": 2.002653875038959,
+        "run_seconds": 1.9471341249882244
+      }
+    }
+  ],
+  "sign_cases": 3272,
+  "source_pr845_head": "125ce25bdae421627b56077adc14940c2d0aa52f",
+  "status": "PASS_P112_READER_MEMO",
+  "support_certificates": 511,
+  "targeted_exact_evaluations": {
+    "memo": 896,
+    "method": "Untimed instrumented full source-bound replays, both compared against the complete committed certificate.",
+    "reference": 979776
+  },
+  "timing_scope": "process_seconds: fresh admitted worker from process launch through exit, including source admission, imports, complete run, complete-certificate comparison and output writing. Excludes parent control suite and parent child-output decoding/comparison. run_seconds: complete source run() boundary.",
+  "verify_py_sha256": "9fb00f9b71243cafbf35c7cc78f9de8213a15a99ac0dadc23e4227a7385bd3f0"
+}
diff --git a/research/p112-reader-memo-seed92/CLAIM.md b/research/p112-reader-memo-seed92/CLAIM.md
@@ -0,0 +1,7 @@
+CLAIM — verification engineering, OS assignment seed **92**. Base `9ef085e6c07c8590329ad00b7eac382aef06124c`.
+
+Exact object: Mihir Srivastava's incorporated PR845 `research/p112-single-lift/verify.py`, specifically repeated binary reader-form evaluations in its complete 3,272-sign-case constraint stream. Method: source-pinned opt-in memoization keyed by the complete ordered reader basis and label, with bounded in-process storage; alternate fresh-process baseline/candidate pairs, compare every field of the original certificate and retain all original controls. Add changed-label/basis and collision controls. No new mathematics or kappa; frozen sources and selected frontier unchanged.
+
+Three candidates screened: exact moment kernels (overlaps retained exact-exp tooling), registered source-pin batching (overlaps #1452/#1595), and this PR845 reader-form kernel. Seed `92 mod 3 = 2` selects the third, after checking live bodies/Discussion claims. #1144 accelerates a different free-sink determinant audit; #1584 accelerates the finite-bit edge scorer; the active selected phase-replay claim is distinct.
+
+Expected deliverable: one opt-in verifier package with a full-certificate comparison, paired timings, meaningful fault controls, exact pins and a bounded replay hook. Expected finish **2026-10-14 05:53 UTC**; expires **06:10 UTC**. No pinned sprint issue exists; this claim will also be retained locally. Prepared for Charan (@theon199) with substantial OpenAI Codex assistance.
diff --git a/research/p112-reader-memo-seed92/README.md b/research/p112-reader-memo-seed92/README.md
@@ -0,0 +1,130 @@
+# Exact compiled reader tables for the PR845 replay
+
+**Verification tooling; no new mathematics or κ.** The opt-in compiled-table
+engine runs Mihir Srivastava's complete P112 single-lift verifier about **1.40×
+faster** in five alternating fresh-process pairs on macOS arm64 / CPython
+3.12.13. Both engines reproduce every field of its committed certificate:
+511 support certificates, 3,272 sign cases, 128 conflicting kernels and all
+nine inherited controls. The original package is unchanged.
+
+## What changes
+
+PR845's `constraint_rows` repeatedly evaluates the same reader bases against
+the same live labels. `ConstraintMemo` in [verify.py](verify.py) computes the
+eight-by-112 table with the **original exact evaluator**, then builds every
+original ordered constraint row. An instrumented complete replay counts
+979,776 targeted evaluations in the reference and 896 in the candidate.
+
+The cache key contains all ordered basis entries and all labels. Each call
+snapshots their current values; it caches neither coefficients nor X. The LRU
+holds at most four tables of 896 integers. Only the fixed-size built-in integer
+list/tuple domain uses this path. Other inputs retain the original builder,
+including its failure behavior. There is no cross-process cache or persisted
+reuse receipt.
+
+Every other predecessor check runs unchanged. Acceptance compares the whole
+certificate using canonical, type-sensitive JSON, including ordered row hashes,
+original-row support minors, exact kernels, conflict ports and control outputs.
+Rank or histogram agreement alone is insufficient. Live inputs are admitted
+by exact SHA256 before import; an omitted source pin fails.
+
+## Reproduce
+
+From the repository root, using Python 3.12 and the standard library:
+
+```sh
+python3.12 -B scripts/research_package.py --package research/p112-reader-memo-seed92
+```
+
+Expected final marker:
+`PASS_REGISTERED_RESEARCH_PACKAGE research/p112-reader-memo-seed92 outcome=completed`.
+The registered replay compares both full engines, runs five focused test groups
+under optimization, and invokes PR845's unchanged independent checker.
+
+For a single faster, source-bound complete replay, create a fresh output:
+
+```sh
+reader_work=$(mktemp -d)
+python3.12 -I -B research/p112-reader-memo-seed92/verify.py \
+  --engine memo --output "$reader_work/REPLAY.json"
+```
+
+Expected: `PASS_P112_READER_MEMO_ENGINE engine=memo`. The output contains the
+entire predecessor certificate and cache statistics. Existing files are never
+overwritten. The default engine is `compare`, which deliberately runs both
+full replays and therefore takes longer than either single engine.
+
+To repeat the paired benchmark:
+
+```sh
+reader_work=$(mktemp -d)
+python3.12 -I -B research/p112-reader-memo-seed92/verify.py \
+  --pairs 5 --report "$reader_work/BENCHMARK.json"
+```
+
+Expected: `PASS_P112_READER_MEMO` with `supports=511 signs=3272
+conflicting_kernels=128 inherited_controls=9 controls=13 pairs=5`, and complete
+certificate digest
+`d1140c3af508b420ab63b1de5fb29af9ed47631492ab1eb7d5b19b06a7d83646`.
+Timing ratios are reported and never used as a mathematical pass threshold.
+
+## Evidence and limits
+
+[BENCHMARK.json](BENCHMARK.json) records all five pairs and their environment.
+Median complete worker-process time is **1.965451 s → 1.402593 s** (1.4013×).
+This includes source admission, imports, the complete verifier, full-certificate
+comparison and output writing. It excludes the parent driver's controls and
+comparison of child outputs. Median time inside the unchanged `run()` boundary
+is **1.911796 s → 1.347156 s** (1.4191×). These are local workload measurements,
+not a hardware-independent, selected cold-replay or multiplication speed claim.
+
+Thirteen additional controls cover ordered and mutated bases, mutated labels,
+uncached X/coefficients, integer-equal float failures, eviction/recomputation,
+erased reader values, omitted/reordered constraints, a type-altered rank receipt,
+and a missing source pin. Five tests additionally cover every proof section,
+lazy/fallback inputs, source drift, duplicate pins/JSON, actual optimized
+workers, read-only relocation in a path with spaces and preserving output after
+failure. PR845's separate integer-minor/set-elimination checker also passes;
+that implementation and its independent-method credit belong to its predecessor.
+
+This contributes finite replay equivalence and tooling. It does not prove the
+all-real-parameter theorem anew, admit a physical supplier or audit the full
+multiplication theorem. Selected cold replay, native/Lean builds and hosted CI
+are not claimed. All-size/compiler, realization, hosting, precision/tape,
+analytic transfer and bit H1–H5 obligations remain separate.
+
+## Sources, assignment and attribution
+
+Exact research base: **`9ef085e6c07c8590329ad00b7eac382aef06124c`**.
+[SOURCES.json](SOURCES.json) pins all eight live inputs with Git blob IDs and
+SHA256, plus historical provenance. The reviewed PR845 head is
+`125ce25bdae421627b56077adc14940c2d0aa52f`. The selected record at this base is
+PR1094, conditional κ = `836330098538459/500000000000000000`, reviewed in
+`docs/research/community-batch1099-review.md`; its complete command is
+`make frontier1094-verify`. That selected replay was not run for this isolated
+tooling contribution.
+
+OS seed **92**, verification engineering, selected this object after screening
+three narrow candidates. [TASK.json](TASK.json) and [CLAIM.md](CLAIM.md) record
+the selection, bounds and [sprint claim](https://github.com/CrocSwap/integer-mult-bounds/discussions/1166#discussioncomment-18915419).
+The final overlap coverage and limitations are recorded in TASK.json.
+#1144 accelerates a different P112 free-sink determinant checker; #1584 the
+finite-bit edge scorer; #1452/#1595 source-pin reads. The active cold phase
+replay claim concerns different inputs and kernels. No pending code is imported.
+
+Credit **Mihir Srivastava / @mihir-s-05 (PR845)** for the theorem, geometry,
+certificate and both original checkers; **DreamingOfClouds (PR717/PR446)** for
+geometry and the inherited physical interpretation; **Joel Pulikkan /
+GamingPuzzled (PR620)** for the coordinate cost 64 and retained lineage.
+Prepared for **Charan (@theon199)** with substantial **OpenAI Codex assistance**
+in profiling, implementation, verification and writing. No subagents or
+independent human review. New material is Apache-2.0; predecessor sources and
+notices, `upstream/` and the selected frontier remain unchanged.
+
+## Open next questions
+
+Maintainers can review whether to adopt this table preparation in PR845's
+frozen package. Start with the registered replay and paired benchmark, then
+migrate source pins deliberately if approved. That review and pin migration
+block adoption; this package remains opt-in. No additional mathematical search
+is justified by the verifier speedup.
diff --git a/research/p112-reader-memo-seed92/SOURCES.json b/research/p112-reader-memo-seed92/SOURCES.json
@@ -0,0 +1,92 @@
+{
+  "schema_version": 1,
+  "base_commit": "9ef085e6c07c8590329ad00b7eac382aef06124c",
+  "files": [
+    {
+      "path": "certificates/selected-result.json",
+      "usage": "provenance",
+      "git_blob": "ae6a99f28aa468649eb66a361e764d769363bf11",
+      "sha256": "10b9e340fd8e354a37a0c4813c90e5de1dd42044b5f6aca4a3fc5a42b685fa32"
+    },
+    {
+      "path": "docs/research/agent-task-catalog.json",
+      "usage": "provenance",
+      "git_blob": "06c96f5579bb67ed1b86d334e78fbed7da6945f1",
+      "sha256": "6686ad58f1fbdbc87134a78901b60fd53a20d3984355e91defbbebc9e9fa8075"
+    },
+    {
+      "path": "research/p112-single-lift/RESULTS.json",
+      "usage": "input",
+      "git_blob": "be204ad0bb27dda715acc3e01fbe7ad6ddc1e010",
+      "sha256": "0947e7380f931dc8e447751b9b33d531a1ced92a6b94225be7ea29ea57497c05"
+    },
+    {
+      "path": "research/p112-single-lift/SOURCES.json",
+      "usage": "input",
+      "git_blob": "a36ad2fa75d5f2bbce60997a4caa21b0f12786c3",
+      "sha256": "3895e1e5d97987d760fca7ef124cccabcfe5a675ce2928c1c9b2de1f48bb7667"
+    },
+    {
+      "path": "research/p112-single-lift/geometry.py",
+      "usage": "input",
+      "git_blob": "34c1abf059d7be5fe5e5044cf01a09b843071c68",
+      "sha256": "3857c11ec731a602b8059d85d58fbb3b641bab687bb33016a0e54f7567752c8a"
+    },
+    {
+      "path": "research/p112-single-lift/independent_check.py",
+      "usage": "input",
+      "git_blob": "dec55ff6c048c8aa2e4c9bc274345c56d8c016d6",
+      "sha256": "9d74b6fde539f82c31fda88074c7c2dd0465f9892ed61b8d13b5dae0fbf5d29b"
+    },
+    {
+      "path": "research/p112-single-lift/reference-geometry.json",
+      "usage": "input",
+      "git_blob": "d862663d3c0ffe3c72df510e89997beb85cef640",
+      "sha256": "4f44af368e046002553d632ee90c6470990827cea3fae41f0f54018522d7a080"
+    },
+    {
+      "path": "research/p112-single-lift/reference/PR717-NOTICE.md",
+      "usage": "input",
+      "git_blob": "481f6bf99bb19c7f2f73b0991a751b6b9187b16c",
+      "sha256": "7809c20cd037dcc29edb5207cbc333bed7b2bd1026cb4f1137f150de90d6866d"
+    },
+    {
+      "path": "research/p112-single-lift/reference/pr717_e8model.py",
+      "usage": "input",
+      "git_blob": "489a3e2c12e58c0e824eb6c49f5168ddc6756085",
+      "sha256": "7255d13ef349e7a549c0f26addf35ef72d28c72343f3abfcbab2f5eba74834a5"
+    },
+    {
+      "path": "research/p112-single-lift/verify.py",
+      "usage": "input",
+      "git_blob": "b7a81ad2f5beb21cd5595b93c0c07b467a3a9715",
+      "sha256": "8da963c33c5d09f1f23437cd89cb9c05f397ac1cddae49d3dc47b2d719555546"
+    },
+    {
+      "path": "research/p112-single-lift/PROOF.md",
+      "usage": "provenance",
+      "git_blob": "609a2b1f83f17523a570a1a2e1731817c87a587a",
+      "sha256": "02147aa2a93a449dcbd15886b42d993bd0dbb554d7c15d347a9567fbe2a442e1"
+    },
+    {
+      "path": "research/p112-single-lift/README.md",
+      "usage": "provenance",
+      "git_blob": "41ee5713cd31f82b3a2882a178b03ed8fbede436",
+      "sha256": "b0aa432b38e6317661d4edfe5fa37c377cff937a527fe83a55261e7202ce2442"
+    },
+    {
+      "path": "research/p112-single-lift/NOTICE.md",
+      "usage": "provenance",
+      "git_blob": "eb811ef7e601441f07c996fa3a265639b354eb8f",
+      "sha256": "54fcd00460fed3afecad295c0e8310cc1f44d00a21c9ac2872be8c4e3ca26dbe"
+    },
+    {
+      "path": "docs/research/community-batch952-review.md",
+      "usage": "provenance",
+      "git_blob": "512ce9f18b32007bea71452eb5f3eb85c4d577ad",
+      "sha256": "04db59dfc2c31453229b6061bbf59547e71ce3abdc0ce0f6779144bb0dde127b"
+    }
+  ],
+  "predecessor_head": "125ce25bdae421627b56077adc14940c2d0aa52f",
+  "attribution": "Mihir Srivastava PR845; DreamingOfClouds PR717 geometry and PR446 physical interpretation; Joel Pulikkan/GamingPuzzled PR620 coordinate cost. Original sources remain in place and unchanged."
+}