Dugongue · Oct 19, 2026
diff --git a/research/xor-endpoint-tapes/INPUTS.json b/research/xor-endpoint-tapes/INPUTS.json
@@ -0,0 +1,8 @@
+{
+  "schema_version": 1,
+  "sha256": {
+    "inputs/PR3480-verify.py": "4ce24c8e835b1defae20fa8a2630c44536523ba08bce6226ab39c1045754de6b",
+    "inputs/PR3480-README.md": "d537f1342e8de682826e98e7a1ec9bb30f9d4152d3d7ba270443eb91877f2c9f",
+    "inputs/PR2297-PROOF.md": "9020e4b7ea5c5bd6e49148f940a81c5c95e8f91ce870389a2f73cac4ff276696"
+  }
+}
diff --git a/research/xor-endpoint-tapes/LIBRARY-BRIDGES.json b/research/xor-endpoint-tapes/LIBRARY-BRIDGES.json
@@ -0,0 +1,54 @@
+{
+  "schema_version": 1,
+  "sources": [
+    {
+      "module_path": "Mathproof/External/IntegerMultBounds20261008/Matrix/CommutingPhases.lean",
+      "sha256": "95d64270a9973e00b75d2b1f8a29ed95d15f0c3013e604f2ef14249a2a7c9d48",
+      "declarations": [
+        "translate",
+        "xor_translation_commutes",
+        "raw_phase_translation_commutes",
+        "gauge_phase_noncommuting_witness"
+      ],
+      "evidence": "Inspected canonical source and direct mathematical adaptation; no new Lean compile, typed-port composition or kernel audit."
+    },
+    {
+      "module_path": "Mathproof/Computability/MultitapeLocality20261014.lean",
+      "sha256": "4e5df1bdd9764d335424acb2a3aad8364f647892e632d8b09aec85e45cfb5b31",
+      "declarations": [
+        "act_preserves_unscanned",
+        "step_locality",
+        "moveOffset_unit"
+      ],
+      "evidence": "Inspected canonical source and direct mathematical adaptation; no new Lean compile, typed-port composition or kernel audit."
+    },
+    {
+      "module_path": "Mathproof/Computability/MultitapeFrameRule20261014.lean",
+      "sha256": "e5aa9ec056340c67bc6e382eee0a859e6903d9944d3a0e021f2a24a12459cb66",
+      "declarations": [
+        "FootprintedRuns.frame",
+        "FootprintedRuns.compose"
+      ],
+      "evidence": "Inspected canonical source and direct mathematical adaptation; no new Lean compile, typed-port composition or kernel audit."
+    }
+  ],
+  "search_coverage": "Four source-catalog queries capped at 12 hits each; two DAG name lookups capped at 3 hits each. Not exhaustive.",
+  "dag_results": [
+    {
+      "query": "xor_translation_commutes",
+      "results": []
+    },
+    {
+      "query": "FootprintedRuns.frame",
+      "results": [
+        {
+          "id": "source:7977930",
+          "snapshot": "source-catalog",
+          "name": "Mathproof.Multitape.Absolute.FrameRule.FootprintedRuns.frame",
+          "type_node": null,
+          "axioms": null
+        }
+      ]
+    }
+  ]
+}
diff --git a/research/xor-endpoint-tapes/LIBRARY-BRIDGES.md b/research/xor-endpoint-tapes/LIBRARY-BRIDGES.md
@@ -0,0 +1,29 @@
+# Linked-library and predecessor bridges
+
+The shared source catalogue and authoritative DAG were queried with bounded
+packets. Retrieval is not a fresh kernel check or an exhaustive novelty
+search. Exact inspected source hashes and declaration names are stored in
+LIBRARY-BRIDGES.json; no absolute local user path is needed by the package.
+
+| Inspected source | Reused concept and exact limit |
+|---|---|
+| `Mathproof.External.IntegerMultBounds20261008.Matrix.CommutingPhases` | `translate`, `xor_translation_commutes` and `raw_phase_translation_commutes` supply the exact XOR operator/commutation precedent. The new proof specializes to bounded h-bit records and implements the resulting permutation on tapes. |
+| Same module, `gauge_phase_noncommuting_witness` | Prevents assuming a deferred translation passes every native gauge. The new correction is applied at an explicit endpoint and does not use free global relabeling. |
+| `Mathproof.Computability.MultitapeLocality20261014` | `act_preserves_unscanned`, `step_locality` and `moveOffset_unit` guide the literal observation/write/unit-move semantics. The new tables are proved directly; they are not represented as Lean-compiled runs. |
+| `Mathproof.Computability.MultitapeFrameRule20261014` | `FootprintedRuns.frame` and `FootprintedRuns.compose` guide exact window preservation and additive subroutine composition. The explicit template scans give the new footprint and restored head endpoints. No existing Lean footprint is silently reused for a new machine. |
+| PR2297, Cinna Davis | Defines `C_q=((1+i)I+(1-i)X_q)/2` and the five-plane local correction. The generic C_E² identity is written out in PROOF.md and checked on small exact complex streams. No full cover search or 16-term decoder is rerun. |
+| PR3480, Gonzalo Mier | Supplies the conditional shared-register cancellation handoff and the source-pinned 23 masks. Here each is corrected separately; summing isolated tape bills does not establish native owner placement. |
+| `research/high-route-permutation-frames/` | Tekalign's existing paid permutation-frame scheduling prevents a first-use claim for carrying permutations. The new standalone XOR record machines use an explicit different ABI and full bit-tape invoice. |
+| Coordinate-CNOT spectator and integer-address signed-SWAP packages | Their native-alphabet prices are not prices for the present literal tape scans. This comparison prevents treating an address map as free or conflating binary payload operations with address operations. |
+
+The bounded source packet used 4 queries with at most 12 hits per query.
+The directly relevant translation hit was inspected in the canonical source.
+Queries with no hits do not exclude related declarations under other names.
+Historical/source-catalog DAG entries remain labelled as such; no new typed
+port, compatible combined import environment or kernel/transitive-axiom
+certificate is claimed for the informal machine proof.
+
+The concrete sufficient result consists of the finite transition construction,
+exact travel formulas, dirty-scratch restoration, norm-preserving endpoint
+composition and source-bound bills. XOR permutation, three-swap restoration
+and sequential tape compilation themselves are established techniques.
diff --git a/research/xor-endpoint-tapes/NOTICE.md b/research/xor-endpoint-tapes/NOTICE.md
@@ -0,0 +1,51 @@
+# Attribution
+
+New finite transition compiler, exact movement formulas, dirty-record
+correction primitive, proof/application and checks: Dugongue, with substantial
+OpenAI Codex assistance. Apache-2.0. No independent human or new kernel
+verification is claimed.
+
+The three documents in inputs/ are unmodified public sources. Gonzalo Mier's
+PR3480 at `8e6fcaa557d2bff56c914f167e19929b0b656e1f` supplies the immediate
+shared-register correction question and the explicit 23-mask list. Its
+README and verifier retain the existing OpenAI Codex disclosure. The
+verifier is treated as source data and never imported or executed here.
+Cinna Davis (1337isnot1337)'s PR2297 proof at
+`49f9c1a08bf1fe67a62bd6d2825aa0d03e1b1518` supplies the five-plane endpoint
+definition and local algebra. Its OpenAI disclosure and nested credit remain.
+
+The five-plane lineage credits Philippe Van Velzen/Discussion1587, paired
+planes PR1825, Discussion2024/PR2041 and tetrad-local PR1909. Sagar Thakrar's
+PR3876 at `d359cf4d3e2fc46e989979dcb2be351cbab97451` closes a separate scalar
+decoder contract, building on Arham Tahir/PR2879, Philippe Van Velzen,
+Sagar Thakrar/PR2532, Cinna Davis, Adam Gall/PR1889, Tekalign/PR2782,
+the C5 principal-rank work and other retained contributors. We use that
+scope distinction, not its certificate as a computational input or a
+claimed impossibility of this operator-valued tape primitive.
+
+The merged `research/high-route-permutation-frames/` package, by Tekalign
+(tekuuu) with OpenAI Codex assistance, already carries field permutations
+through later BIT route factors with a paid restoration guard. It credits
+Jose Manuel/PR2915/2899, David Arias/PR2753, Nick Petalas, commdata2338/PR3233,
+Sagar Thakrar/PR3265, Krzysztof Wójt/PR3301, Arham Tahir, ywz65, PR2157/589/555
+and the native geometry/compiler lineage. That work is a different actual
+route compiler and its savings are not attributed to this package.
+
+Gareth Davidson's coordinate-CNOT spectator certificate and ywz65's
+integer-address signed-SWAP theorem use explicit paid native alphabets;
+their address operations are not free operations in our tape bill.
+Their inherited PR1414/1333, PR2544, Aaron Wang/PR2444 and Ross Richardson
+PR1989/2415 credit remains with those packages. They are compared for scope,
+not imported as implementation dependencies.
+
+The linked MathProof `CommutingPhases`, `MultitapeFrameRule` and
+`MultitapeLocality` sources provide translation/phase and exact-frame
+precedents. Their source hashes and declaration references are recorded in
+LIBRARY-BRIDGES.json. No source-only/historical index hit is represented as
+a new kernel check, and no assertion of first use of XOR permutations,
+three-swap scratch restoration or finite tape compilation is made.
+
+The retained Harvey–van der Hoeven, OpenAI manuscript, Douglas Colkitt,
+RaD/hipotures and other upstream/community notices are not replaced.
+No frozen supplier, selected certificate, upstream source or other
+contributor's file is changed.