Audit: PR442 hosting windows make the partial width-seven bit rows unreachable for replica-local hosting on the frozen recipe (no new kappa)

research/bit589-width7-windows-seed72/SOURCES.json

@@ -0,0 +1,96 @@
+{
+  "schema_version": 1,
+  "base_commit": "0b1cccf86e2f5132ed286a26ef0a681a495740ea",
+  "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/bit-partial-width7-optimum/EXPECTED.json",
+      "usage": "input",
+      "git_blob": "b118ea87a6e9ec6805b27f847b9ae7d12ecbdc8a",
+      "sha256": "a6ec64866b949626d6508ee7172b144c3fd29f1065822b523fafd7d2a59e2e91"
+    },
+    {
+      "path": "research/bit589-width-packing/RESULT.json",
+      "usage": "input",
+      "git_blob": "f11f6e569cd6d780946548c9bd3bc10d65ed6ab9",
+      "sha256": "200b91e2d801c789426835b640381b7d0a4bae5e67c3859cbd97b52dc7053b09"
+    },
+    {
+      "path": "research/bit589-width-packing/SOURCES.json",
+      "usage": "input",
+      "git_blob": "f6a97de32691a1d426ac4819aad24c0c8414d97e",
+      "sha256": "53fe98c8f01de7e32e3110e591d62e60292abb38cc11ecde4839f2eebd3b5fbc"
+    },
+    {
+      "path": "research/bit589-width-packing/packing.py",
+      "usage": "input",
+      "git_blob": "a6079ee0dd4d8037be477a3c4497f472348bde07",
+      "sha256": "2bb5515807714807961289e4ee6958c3ad7e0c27bbec4d9212b138c662a6cc7b"
+    },
+    {
+      "path": "research/community-frontier1094-audit/ASSEMBLY.json",
+      "usage": "input",
+      "git_blob": "5fd676df060f3b450782fdf5ff5f6beb63a5eeb0",
+      "sha256": "1122cb3a575f8c788345fc1bf4c108ac7c3250fb81a5988671f5eb04b3ab3045"
+    },
+    {
+      "path": "research/e8-bit-p111-tuned/data/selection9-p111.json",
+      "usage": "input",
+      "git_blob": "e69ea66a3add357f570fc5552ab6368b3c8f9bb2",
+      "sha256": "4ab1ed4529683cd184987f8bc3031fb7408b9cd22c25ab9cf541a8aaf896eaf1"
+    },
+    {
+      "path": "research/e8-bit-p111-tuned/tune/EXPECTED.json",
+      "usage": "input",
+      "git_blob": "e293ddc557742637f9f679f083a1a282d87b1b13",
+      "sha256": "b9f0e5ca038abe837b0007c8fe006f366a6f02009244ac57705a8a200e2301f4"
+    },
+    {
+      "path": "research/e8-bit-p111-tuned/vendor/pr510/e8-r480-nine-stage/vendor/e8-r715-eleven-stage-whole/code/audit_hosted_k.py",
+      "usage": "input",
+      "git_blob": "8864ca9b942a3507171edca0cf6cbaaf10fcf434",
+      "sha256": "57e558c80a2fea6e291587b04afd76dfaee3550729997798b61f33daf90b8ec9"
+    },
+    {
+      "path": "research/e8-bit-p111-tuned/vendor/pr510/e8-r480-nine-stage/vendor/e8-r715-eleven-stage-whole/code/global_k.py",
+      "usage": "input",
+      "git_blob": "3df12bd632f1fdba825ced57110a1fee82eb2857",
+      "sha256": "6368196a393ac25bb2918880e4c4bd31ed733b762c27b6873ab391cbc0c5c3c8"
+    },
+    {
+      "path": "research/e8-bit-p111-tuned/vendor/pr510/e8-r480-nine-stage/vendor/idle-axis-hosting/hosting_core.py",
+      "usage": "input",
+      "git_blob": "68b536daa9e6546c0aa046934bf840aaa3c3fde3",
+      "sha256": "d8b90cd5d152df3c7c351230863d9e61816ad5e1f21d69bea1a86a5daad559a3"
+    },
+    {
+      "path": "research/native-complex-moment552/engines/base_two_moment.py",
+      "usage": "input",
+      "git_blob": "14a90fa0e788217856f070382c9e61bd7c603c61",
+      "sha256": "aee70c242a2f1d225528fabe2002e1564d1dc884797c07d28a831c7878ee9098"
+    },
+    {
+      "path": "research/native-complex-moment552/engines/moment.py",
+      "usage": "input",
+      "git_blob": "7d697cb45203fe72d97037403d6da41f71f3b7fd",
+      "sha256": "81cbc864e106418fc7854322746f72180cb8110ff39f74450a3b8d55a82c1482"
+    },
+    {
+      "path": "research/native-complex-moment552/engines/outer.py",
+      "usage": "input",
+      "git_blob": "43007488ad2c43187c77cc6798beff8b1cd76ad2",
+      "sha256": "c1ebb52163a7219504c728a5c2b098bf1e50bf20148a2879c5a74c3cadc7c03c"
+    }
+  ]
+}

research/bit589-width7-windows-seed72/TASK.json

@@ -0,0 +1,72 @@
+{
+  "schema_version": 1,
+  "task_id": "uniform-transfer-audit",
+  "priority_revision": 3,
+  "base_commit": "0b1cccf86e2f5132ed286a26ef0a681a495740ea",
+  "original_card": {
+    "id": "uniform-transfer-audit",
+    "question": "Can one exact all-size/compiler, precision, hosting or restoration implication be proved or falsified?",
+    "starting_points": [
+      "certificates/selected-result.json",
+      "docs/audit.md"
+    ],
+    "methods": [
+      "Prove one missing implication with explicit quantified hypotheses.",
+      "Find an exact counterexample at the claimed interface boundary."
+    ],
+    "acceptance_gate": "Identify the proposition and what would change downstream; do not equate a finite replay with a universal proof.",
+    "first_step": {
+      "command": [
+        "python3",
+        "-B",
+        "scripts/research_baseline.py"
+      ],
+      "expected": "PASS_RESEARCH_BASELINE_BINDINGS_ONLY",
+      "inspect": "Read the selected review scope and choose one exact all-size or precision implication.",
+      "scope": "Positive control / source-bound orientation only; this is not a new construction or full cold replay."
+    },
+    "first_experiment": "Write its quantified inputs and conclusion; find a missing hypothesis or construct a minimal exact adversarial instance.",
+    "bottleneck_link": "Finite supplier improvements are useful only under the retained transfer obligations.",
+    "allowed_variations": [
+      "Choose one tape, precision, hosting or restoration lemma."
+    ],
+    "scope_change_triggers": [
+      "Replacing a quantified obligation with another finite smoke test."
+    ],
+    "overlap_terms": [
+      "uniform transfer",
+      "all-size",
+      "precision"
+    ]
+  },
+  "scope_status": "refined",
+  "actual_scope": "One hosting implication for the selected PR589 bit supplier: do the window-free partial width-seven rows of merged #1023 and merged #1281 (width-7 residuals placed in capacity-7 idle slots without time order) survive PR442's replica-local hosting windows (audit_hosted_k: s in W[i] for every hosted slot) on the frozen nine-stage recipe, whole residuals, no banks? Exact window-respecting optimum and certificate for every k = 0..74.",
+  "scope_change_reason": "The card asks for one exact hosting implication. This narrows it to the time-order hypothesis that #1023 lists as unchecked, and to the per-replica, frozen-recipe part of #1281's open question 1. That question's PR1179 boundary-contract audit and cross-replica pooling stay open, as do projector compatibility, owner inactivity and a physical width-7 helper.",
+  "bottleneck_link": "Side result on the bit supplier; the selected record is complex-bound, so no kappa changes. Under #1023's one-unit local pricing, the frozen recipe and replica-local hosting, it corrects the bit headroom a partial width-seven helper mechanism can buy: the window-respecting optimum keeps 63.6-67.1% (optimistic) or 56.3-60.7% (pessimistic) of the window-free gain, i.e. the published rows overstate it by about 1.5x or 1.7x, roughly flat in k.",
+  "budget_minutes": 180,
+  "stop_rule": "Stop once every k = 0..74 has an explicit window-feasible placement and a tight grid certificate under both local brackets, k = 0 reproduces the selected bit endpoint, and every #1023 and #1281 row is compared; report any row not decided instead of claiming it.",
+  "outcome": "completed",
+  "evidence": "RESULT.json: the nine-stage idle schedule equals, event by event and field by field, the global_k.py reference (its recipe() and idle-insertion block executed via ast inside a transcribed loop); three window predicates agree on it; the frozen selection lies inside. Capacity-7 children 1/3/5 admit only stages 2/4/6, so at most 3k of 9k width-7 residuals per replica sit in 7-slots. Exact window-respecting optimum, an explicit placement built and checked at run time, and a tight 1e-24 certificate for every k = 0..74, both brackets; k = 0 reproduces 0.001733427039589844234749. k = 37 optimistic 0.001770809187 (+2.157%) vs window-free 0.001789671631 (+3.245%); k = 74 0.001809907119 (+4.412%) vs 0.001849754238 (+6.711%). Retained fraction 0.636-0.671 (optimistic), 0.563-0.607 (pessimistic). All 8 #1023 partial rows unreachable; #1281's closed-form local optimum transcribed and checked against its 75 idle profiles; all 148 local rows priced here from it (8 also priced by #1281, reproduced exactly in local endpoint and histogram) lie above the per-replica bound, as do #1281's 8 priced pooled32 rows with k > 0. Chain reduction confirmed by max-flow on every distribution (462 at k = 1, 6188 at k = 2). #1023's all-width-7 optimum stays window-feasible (104 sextet splits). 14 tests pass.",
+  "author": "Sagar Thakrar (thakrarsagar)",
+  "ai_assistance": "Substantial: Claude (Anthropic) via Claude Code. Parallel read-only scout and judge subagents proposed and ranked audit candidates and found the window obstruction; the DP, placement builder and certificates were written with Claude and checked by the controls in test_package.py. A post-PR adversarial review (three reviewers and a verifier) found overclaims and the missing schedule anchor, fixed in the second commit.",
+  "open_questions": [
+    {
+      "question": "Can cross-replica (T16-style) interleaved owner maps lift the one-stage capacity-7 windows for the bit supplier?",
+      "why_it_matters": "One route (others: a changed recipe, split residuals, banks) to #1281's pooled rows or the window-free local rows; replica-local hosting on the frozen recipe is ruled out here.",
+      "first_step": "Compute windows on an interleaved bit timeline with owner_replica != replica and rerun the DP.",
+      "blocked_by": "An interleaved bit timeline (the selected T16 maps are complex-side) and a cross-replica DP; wopt.solve and the global_k.py anchor are specific to this per-replica schedule."
+    },
+    {
+      "question": "Does another nine-stage stage recipe widen the capacity-7 windows?",
+      "why_it_matters": "The obstruction is recipe-specific; open #1487 enumerates recipes for width-8 hosting only.",
+      "first_step": "Generalise wopt's two-chain DP to arbitrary interval windows (or use a min-cost-flow bound), then apply it to #1487's hostable recipes.",
+      "blocked_by": "The generalised DP: wopt.solve asserts this recipe's chain structure."
+    },
+    {
+      "question": "Do the window-respecting placements pass projector compatibility (#1307) for mixed-width slots?",
+      "why_it_matters": "Windows are necessary, not sufficient; slots that split q-perp lie outside #1307's completion-free argument.",
+      "first_step": "Rebuild the k = 1 placement with verify.build(1, plan, slots) and run #1307's projector check on the returned slot list.",
+      "blocked_by": "None."
+    }
+  ]
+}

research/bit589-width7-windows-seed72/VERIFY.json

@@ -0,0 +1,50 @@
+{
+  "schema_version": 1,
+  "status": "ready",
+  "python": "3.12",
+  "requirements": [],
+  "inputs": [
+    "certificates/selected-result.json",
+    "research/bit-partial-width7-optimum/EXPECTED.json",
+    "research/bit589-width-packing/RESULT.json",
+    "research/bit589-width-packing/SOURCES.json",
+    "research/bit589-width-packing/packing.py",
+    "research/community-frontier1094-audit/ASSEMBLY.json",
+    "research/e8-bit-p111-tuned/data/selection9-p111.json",
+    "research/e8-bit-p111-tuned/tune/EXPECTED.json",
+    "research/e8-bit-p111-tuned/vendor/pr510/e8-r480-nine-stage/vendor/e8-r715-eleven-stage-whole/code/audit_hosted_k.py",
+    "research/e8-bit-p111-tuned/vendor/pr510/e8-r480-nine-stage/vendor/e8-r715-eleven-stage-whole/code/global_k.py",
+    "research/e8-bit-p111-tuned/vendor/pr510/e8-r480-nine-stage/vendor/idle-axis-hosting/hosting_core.py",
+    "research/native-complex-moment552/engines/base_two_moment.py",
+    "research/native-complex-moment552/engines/moment.py",
+    "research/native-complex-moment552/engines/outer.py"
+  ],
+  "commands": [
+    {
+      "argv": [
+        "{python}",
+        "-I",
+        "-B",
+        "{package}/verify.py"
+      ],
+      "timeout_seconds": 600,
+      "checks": "Binds the frontier1094 certificate (named by the selected or a checkpoint record) -> the PR589 record and engines pinned by #1023's packing.py; matches the nine-stage idle schedule event by event with the one the pinned global_k.py produces (recipe and idle-insertion rule extracted and replayed); three window predicates agree on it and the frozen selection lies inside; capacity-7 children admit one stage each; chain reduction by max-flow over every distribution for k = 1, 2; for every k = 0..74 the DP optimum, an explicit residual-level placement built and checked at run time (window, width, offset, disjoint cells, hosted once), both local brackets priced with the pinned engines, a 1e-24 grid certificate tight to 1e-50 against the explicit histogram; #1023's eight partial rows; merged #1281's EXPECTED.json (75 idle profiles; 9 priced cases reproduced exactly in local endpoint and local histogram; local and pooled32 endpoints above the per-replica bound); #1023's all-width-7 optimum window-feasible; RESULT.json byte-identical."
+    },
+    {
+      "argv": [
+        "{python}",
+        "-I",
+        "-B",
+        "-m",
+        "unittest",
+        "discover",
+        "-s",
+        "{package}",
+        "-p",
+        "test_package.py"
+      ],
+      "timeout_seconds": 300,
+      "checks": "Capacity-7 windows; schedule = global_k.py reference, with a dropped bridge and six single-field mutations rejected (and the unanchored three-way path shown to accept the dropped bridge); #1023's k = 1 placement outside its windows; k = 0 endpoint and a recomputed tight certificate; window-free rows above and window-respecting rows below the bound; #1281 local and pooled32 rows, with the k = 1 bound recomputed; all-width-7 feasibility; exhaustive brute-force reduction (6188 distributions at k = 2); dropped chain constraint and perturbed window controls; width-8 Hall vacuity; record binding and the explicit k = 1 slot list."
+    }
+  ]
+}