Independent verification (no κ): #3015's R1 restore edit on #2979's words is exact and correctly priced; R1 unique in its rewrite class; #3018 journal and #3021 composition reproduced; no defect found

research/t74-3015-restore-edit-audit/TASK.json

@@ -0,0 +1,91 @@
+{
+ "schema_version": 1,
+ "task_id": "independent-moment",
+ "priority_revision": 6,
+ "base_commit": "d8e0f83f00191df62174b641e9fff055c357f641",
+ "scope_key": "pr3015-r1-restore-edit-independent-audit",
+ "original_card": {
+  "id": "independent-moment",
+  "question": "Does one changed supplier or assembly claim lack a source-bound independent check of a specific new implication?",
+  "starting_points": [
+   "certificates/selected-result.json",
+   "research/rank7-joint-endframes/README.md",
+   "docs/research/research-family-status.md"
+  ],
+  "methods": [
+   "Independently reconstruct a coordinated frame/circuit change and its single-change control, then derive the complete rank and physical ledger.",
+   "Implement certified moment inequalities through a different exact enclosure method."
+  ],
+  "acceptance_gate": "Demonstrate independence, source binding, positive cases and semantic failure controls. Name the existing checks and the missing coverage. More digits or another equivalent evaluation of an unchanged root is not by itself new verification coverage.",
+  "first_step": {
+   "command": [
+    "python3",
+    "-B",
+    "scripts/research_baseline.py"
+   ],
+   "expected": "PASS_RESEARCH_BASELINE_BINDINGS_ONLY",
+   "inspect": "Read #2979 joint rank-7 end-frame proposal and the revision-6 evidence status. Its claimed kappa is not selected. Compare coordinated changes with the losing single-change controls and locate the complete physical ledger; an aggregate moment alone does not validate the mechanism.",
+   "scope": "Positive control / source-bound orientation only; this is not a new construction or full cold replay."
+  },
+  "first_experiment": "Prioritize a bounded independent audit of #2979: reconstruct one joint group and its single-lowering control, check the declared frames/chronology, then trace changed children, idle widths and cleanup buffers into the complete invoice and both supplier caps. Choose one still-missing implication after checking existing audits. Stop with exact unresolved dependencies if full admission exceeds the session; do not launch a new lowering census on an unvalidated premise or substitute extra root digits for mechanism checking.",
+  "bottleneck_link": "Independent pricing prevents unsupported frontier claims and can reveal genuine accounting slack.",
+  "allowed_variations": [
+   "Reconstruct either a source ledger or a numeric enclosure with a different method."
+  ],
+  "scope_change_triggers": [
+   "Importing the primary function and calling its output independent verification.",
+   "Promoting #2979 from merged source archive to validated supplier solely from author PASS receipts or aggregate pricing."
+  ],
+  "overlap_terms": [
+   "independent moment",
+   "rank histogram",
+   "exact enclosure",
+   "2979",
+   "rank7 joint endframes",
+   "paired extraction",
+   "joint lowering"
+  ]
+ },
+ "scope_status": "refined",
+ "actual_scope": "Independent verification of open PR #3015 (Jeremy Kothe, head e244296debb45fe6f4da84d29e4d64fbab55fda0): the declared restore edit R1 on PR #2979's words (merged, head be191346391661f4d35fc0d1508d0a6029572576) with PR #2979's literal T74 allocation. Checked with code independent of both PRs: the hunk (own diff), the net operator, both edited words exact forward and dual through cleanup and endpoint (632 inputs, 36 initial corrections), the frame recount in two implementations with the paid delta -2H1+H2 attributed to register 63's merged climbs, the literal cleanup against an own re-derivation, the allocation priced in two implementations (271,728 roles) under the new local profiles (complete profile, delta, idle, physical stock 66813), the endpoint C and kappa with own and vendored rational enclosures, the exact gain over PR #2979, a census of R1's rewrite class on the parent and on the edited words, and PR #3018's journal applied to the same fill (its profile and endpoint reproduced), with the PR3015+PR3018 composition computed here (without having read PR #3021, which published it 19 minutes earlier) and equal to PR #3021's published finite records (profile, idle, kappa, gain).",
+ "scope_change_reason": "The card's lead object (#2979's coordinated lowerings) already had six independent audits (#2993, #2987, #3006, #3008, #3010, #3014) and one improving journal (#3018) and two live claims at entry; PR #3015, the newest conditional candidate on the same words, had none. The audit follows the card's method (independent recheck of a changed cost through the complete invoice) on the changed object; the census extends the check to the uniqueness of R1 within its rewrite class, which the author's unshipped search claims.",
+ "declared_milestone": "audit",
+ "budget_minutes": 240,
+ "stop_rule": "Stop after the independent checks and controls pass or a defect is found; no frame re-solve, no reproduction of the author's 2-4-gate block search, no native/admission/outer replay.",
+ "outcome": "completed",
+ "started_utc": "2026-10-16T18:12:00Z",
+ "author": "Sagar Thakrar (thakrarsagar)",
+ "ai_assistance": "Substantial AI assistance (Claude Code, Anthropic) for implementation, checks and writing; no independent human review is claimed.",
+ "existing_checks_consulted": [
+  "#2993 (petalas): frame paths and V45 subsets of #2979, not word semantics or the R1 edit",
+  "#2987/#3010 (RossRichardson): chronological packing of #2979's fill; the fill is reused unchanged by #3015",
+  "#3006/#3008/#3014 (petalas): V45 native endpoint/events and frozen-owner invoice of #2979",
+  "#3018 (petalas): an integrality exchange on #2979's fill; its journal is applied here and its profile and endpoint reproduced",
+  "#3021 (petalas, published 2026-10-16 18:21 UTC): composition of R1 with #3018's journal, author chain run; its finite records equal the composition computed here; its verify.py check_word verifies R1's preimage, 3x3 scalar identity, changed sections and re-derived cleanup, reading the profile delta from the report",
+  "PR #3015's own edits_review.py (author-run, uses the inherited #2833 auditor); not imported or executed here",
+  "petalas, Discussion 2612 comment 18981612 (19:40 UTC, local-only claim): a second restore rewrite (restore[162:178], shared covector X40 - 2 X54 across two non-adjacent consumers), outside the class censused here"
+ ],
+ "open_questions": [
+  {
+   "question": "Does a frame re-solve of the whole edited word (all gate frames free, monotone, same gates) give a paid profile below {1:2443, 2:707, 3:232, 4:108, 5:18, 8:8}?",
+   "why_it_matters": "R1's frames were re-solved by the author's certified frame pricer, which is not shipped; this audit checks the given frames for legality and prices them, not their optimality.",
+   "first_step": "Formulate the frame assignment as a shortest-path problem per register on the lattice of subspaces with the gate constraints (a gate's registers share its frame) and solve it exactly on the parent and edited words.",
+   "blocked_by": "A monotone, gate-coupled frame optimizer over GF(2)^9 subspaces has to be written; none of this contributor's checkers contains one."
+  },
+  {
+   "question": "Beyond R1's class, which 2-4-gate restore blocks of the edited words can be re-synthesised with the same net operator and a lower paid profile?",
+   "why_it_matters": "PR #3015 reports R1 as the first edit of an exhaustive search over all 2-4-gate restore-phase blocks; the census here covers one class (adjacent same-target pairs, three-gate rewrite) and finds no further gain in it.",
+   "first_step": "Enumerate contiguous blocks of 2-4 restore gates, the finite set of equivalent gate sequences (commutations, carrier changes), and price each with the own chain.",
+   "blocked_by": "The equivalence class enumeration needs a specification of the allowed re-synthesis; the author's harness is not shipped."
+  },
+  {
+   "question": "Does PR #3021's release at 0.001729151203983018442998 survive an independent admission check?",
+   "why_it_matters": "The finite side (profile, idle, endpoint, kappa, gain) is reproduced here and equals PR #3021's published records; the native/prefix/outer side is author-run.",
+   "first_step": "Read PR #3021's admission receipts against the composed fill, as #2804's stock and prefix checks did for #2530.",
+   "blocked_by": "Nothing but time."
+  }
+ ],
+ "family_expansion_decision": "None proposed. This audit admits nothing; it reports no defect in PR #3015's finite claims within the checked scope and leaves the author-run native/admission/outer evidence as stated by the author.",
+ "bottleneck_link": "PR #3015 is the newest conditional candidate of the complex T74 line (kappa 0.00172915 against the selected 0.00167266); an independent check of its only new ingredient (the R1 restore edit and its re-derived paid profile) and of the composed PR3015+PR3018 endpoint is the missing coverage before any of these claims can be treated as more than author-checked. No kappa or supplier change is claimed here.",
+ "evidence": "verify.py (RESULT.json): both edited words exact forward and dual through cleanup and the target endpoint (632 inputs, 36 initial corrections, 72 warps, 408 helpers, 91 buffers); the hunk restore[178:180) -> 3 gates reproduced by an own diff and equal to the old block as an operator; two frame-chain implementations give {1:2445,2:706,3:232,4:108,5:18,8:8} -> {1:2443,2:707,3:232,4:108,5:18,8:8} with the delta -2H1+H2 attributed to register 63's climbs 7->8->9 merging into 7->9; PR #2979's allocation priced in two implementations under the new local profiles gives the complete profile {1:1644348,2:470862,3:159468,4:71928,5:13716,8:5328,48:28190} = PROFILE.json = CLAIM.json, delta 666 x (-2H1+H2), idle unchanged, physical 66813; C = 0.001729150973534224428695 accepted and C + 1e-24 rejected by an own rational enclosure and by the vendored PR2804 enclosure, kappa = C - 1e-46 - 1e-70, gain over PR #2979 exact; census (adjacent same-target restore pairs, every legal third frame, 122 rewrites recounted on the parent words and 108 on the edited words): the R1 pair is the only candidate with an improving option (both carrier orders), none on the edited words; PR #3018's journal reproduces its profile and endpoint; the composed PR3015+PR3018 endpoint 0.001729151203983018442998 equals PR #3021's finite records. 117 negative controls rejected by their named rules (85 distinct), 8 positive controls, 7 vendor-pin controls; checker mutations in --mutations (see README). No defect found in the checked scope."
+}

research/t74-3015-restore-edit-audit/VERIFY.json

@@ -0,0 +1,20 @@
+{
+  "schema_version": 1,
+  "status": "ready",
+  "python": "3.12",
+  "requirements": [],
+  "inputs": [],
+  "commands": [
+    {
+      "argv": [
+        "{python}",
+        "-I",
+        "-B",
+        "{package}/verify.py",
+        "--fetch"
+      ],
+      "timeout_seconds": 600,
+      "checks": "Independent verification of open PR #3015 head e244296d (R1 restore edit on PR #2979 head be191346), in standard-library code that reads PR files only as pinned git data (git blob id and sha256) and never imports or runs PR code; GIT_CONFIG* variables are removed before the first git call; --fetch (git >= 2.29, network) writes only the fetched objects and refs/audit/pr3015, refs/audit/pr3018, refs/audit/pr3021. It checks the declared hunk restore[178:180) -> 3 gates by an own diff, the net operator, both edited words exact forward and dual through cleanup and the target endpoint (632 inputs, 36 initial corrections emulated), the frame recount in two implementations with the paid delta -2H1+H2 attributed to register 63, the literal cleanup, the 91 private buffers, PR #2979's allocation (byte-identical in PR #3015) priced in two implementations under the new local profiles (271,728 roles; complete profile, PROFILE.json, CLAIM.json, rank mass, delta 666 x (-2H1+H2), idle, physical stock 66813), the endpoint C = 0.001729150973534224428695 accepted and C + 1e-24 rejected by an own rational enclosure and by the vendored one, kappa = C - 1e-46 - 1e-70, the exact gain over PR #2979, a census of R1's rewrite class on the parent and edited words (R1 unique, class exhausted after the edit), PR #3018's journal reproducing its profile and endpoint on the same fill, and the endpoint of the PR3015+PR3018 composition computed here and equal to PR #3021's published finite records; it requires every negative control to be rejected by its named rule, the positive controls to pass, and the result to equal RESULT.json. See README.md."
+    }
+  ]
+}