Independently validate joint relations and conditional κ = 0.003026674765

research/joint-relations-independent-validation/CLAIMS.json

@@ -0,0 +1,58 @@
+{
+  "schema_version": 1,
+  "review_status": "READY_FOR_COLD_REPLAY",
+  "canonical_main": "1ec1119972c9aefe42525172d07b332fcff81cb4",
+  "selected_baseline": ".0030217758",
+  "selected_metadata_role": "Frozen historical provenance; future selected-result updates do not invalidate this review.",
+  "candidates": [
+    {
+      "pr": 5383,
+      "head": "4bd6efa51a495a7c7b7754915cf0244e7e60969e",
+      "author": "Ali Romman",
+      "kappa": "151333/50000000",
+      "decimal": ".00302666",
+      "package": "research/kappa-joint-relations-20261023",
+      "word_payload_sha256": "2aee5e997df99726fc763e510d9cb69fbe4f86114dc8d2cd33ad26c6c462a21d",
+      "last_pivot": 288
+    },
+    {
+      "pr": 5387,
+      "head": "b1ab650d36f59e5be682a813145b7c67e219e341",
+      "author": "Dugongue",
+      "kappa": "605334953/200000000000",
+      "decimal": ".003026674765",
+      "package": "research/eight-owned-relations",
+      "word_payload_sha256": "9aa0d7112b4e13f1aa13a93dab0af6f4435ffe99b842aad841685becdafee492",
+      "last_pivot": 261
+    }
+  ],
+  "assumptions": [
+    {
+      "id": "native_all_columns",
+      "statement": "Retained native operators, phase conventions, spectators and whole-job adapters work at every positive column multiplicity."
+    },
+    {
+      "id": "owned_erasure",
+      "statement": "Owned streams and suspended centers obey the retained allocation, erasure and ownership contracts."
+    },
+    {
+      "id": "compiler",
+      "statement": "The finite mixed-target compiler and chronological product, completion and splitting satisfy the retained uniform closure contract."
+    },
+    {
+      "id": "precision_tape",
+      "statement": "Retained fixed-tape movement and precision/storage prescriptions apply to the new fixed finite template."
+    },
+    {
+      "id": "analytic_recovery",
+      "statement": "Analytic decomposition, fine-prime avoidance and exact integer recovery follow the retained upstream/outer framework."
+    }
+  ],
+  "selection_authorized": false,
+  "construction_claim": "No new native construction is claimed by this independent review.",
+  "selected_kappa": "15108879/5000000000",
+  "same_selected_assumptions_required": true,
+  "new_unconditional_or_Lean_theorem": false,
+  "uniform_word_materialized": false,
+  "selected_record_changed": false
+}

research/joint-relations-independent-validation/COMMANDS.json

@@ -0,0 +1,95 @@
+{
+  "schema_version": 1,
+  "commands": [
+    {
+      "name": "algebra-5383",
+      "argv": [
+        "agents/algebra/check_joint.py"
+      ],
+      "timeout_seconds": 120
+    },
+    {
+      "name": "algebra-5387",
+      "argv": [
+        "agents/algebra/check_joint.py",
+        "--candidate",
+        "agents/sources/pr5387/research/eight-owned-relations/WORD.json.gz",
+        "--candidate-pr",
+        "5387",
+        "--candidate-head",
+        "b1ab650d36f59e5be682a813145b7c67e219e341",
+        "--out",
+        "agents/algebra/RESULT-5387.json"
+      ],
+      "timeout_seconds": 120
+    },
+    {
+      "name": "ownership-5383",
+      "argv": [
+        "agents/ownership/check.py",
+        "--exact",
+        "--controls",
+        "--output",
+        "agents/ownership/RESULT.json"
+      ],
+      "timeout_seconds": 120
+    },
+    {
+      "name": "ownership-5387",
+      "argv": [
+        "agents/ownership/check.py",
+        "--exact",
+        "--controls",
+        "--candidate",
+        "agents/sources/pr5387/research/eight-owned-relations/WORD.json.gz",
+        "--candidate-head",
+        "b1ab650d36f59e5be682a813145b7c67e219e341",
+        "--output",
+        "agents/ownership/RESULT-5387.json"
+      ],
+      "timeout_seconds": 120
+    },
+    {
+      "name": "event-invoice-and-prices-both",
+      "argv": [
+        "agents/prices/check_prices.py"
+      ],
+      "timeout_seconds": 120
+    },
+    {
+      "name": "outer-BIT-compiler-both",
+      "argv": [
+        "agents/outer/check_outer.py",
+        "--candidate",
+        "PR5383",
+        "repository/research/kappa-joint-relations-20261023/RESULT.json",
+        "--candidate",
+        "PR5387",
+        "agents/sources/pr5387/research/eight-owned-relations/RESULT.json",
+        "--bit-source",
+        "repository/research/query-thinned-controllers/RESULT.json",
+        "--out",
+        "agents/outer/RESULT.json"
+      ],
+      "timeout_seconds": 120
+    },
+    {
+      "name": "applicability-both",
+      "argv": [
+        "agents/applicability/check_applicability.py"
+      ],
+      "timeout_seconds": 120
+    },
+    {
+      "name": "actual-word-semantic-mutants-5383",
+      "argv": [
+        "agents/falsifier/check_mutants.py",
+        "--mode",
+        "full",
+        "--out",
+        "agents/falsifier/CONTROLS-FULL.json"
+      ],
+      "timeout_seconds": 300
+    }
+  ]
+}

research/joint-relations-independent-validation/PROOF.md

@@ -0,0 +1,217 @@
+# Joint validation and exact dependency argument
+
+## Claim and five retained premises
+
+For each of the two exact literal words in CLAIMS.json, the new independent
+engines establish the algebra, ownership, complete invoice and numerical
+compiler/outer specialization described below. The implication remains
+conditional on the **same five literal premises** in CLAIMS.json, which the
+source guard compares with the authenticated selected PR5241 claim:
+
+1. Native operators, phase conventions, spectators and whole-job adapters work
+   at every positive column multiplicity.
+2. Owned streams and suspended centers obey the retained allocation, erasure
+   and ownership contracts.
+3. The finite mixed-target compiler and chronological product, completion and
+   splitting satisfy the retained uniform closure contract.
+4. The retained fixed-tape movement and precision/storage prescriptions apply
+   to the new fixed finite template.
+5. Analytic decomposition, fine-prime avoidance and exact integer recovery
+   follow the retained upstream/outer framework.
+
+The frozen selected record is 15108879/5000000000 = .0030217758 at canonical main
+1ec1119972c9aefe42525172d07b332fcff81cb4. It is historical provenance, with an exact
+vendored copy; no future live selected-file equality is a premise. No source
+object is fetched or substituted for missing working input during replay.
+
+## Distinct literal words and algebraic domains
+
+| Item | Ali Romman PR5383 | Dugongue PR5387 |
+| --- | --- | --- |
+| Source head | 4bd6efa51a495a7c7b7754915cf0244e7e60969e | b1ab650d36f59e5be682a813145b7c67e219e341 |
+| Conditional κ | 151333/50000000 = .00302666 | 605334953/200000000000 = .003026674765 |
+| Last pivot | L288 | L261 |
+| Eighth snapshot source event/role | 11306 / L288 | 11296 / L256 |
+| Eighth zero source event/role | 17580 / L288 | 17233 / L256 |
+| Backward scalar allowance | 158744 | 158776 |
+
+Both eighth substitutions use frame 281, but the unrestricted source relation
+and the retained arbitrary input domain differ. Equal invoices cannot establish
+their semantic equivalence. Each algebra invocation independently replays all
+21,449 original backward events, derives eight source-center-minus-snapshot rows
+and their eight matching later zero rows, reduces the simultaneous rank-eight
+system, and checks the actual preparation. It then replays the entire respective
+candidate on 624 independent retained inputs plus 17 arbitrary owned entry
+inputs. The pivots are 225,227,229,230,241,245,257 and the candidate-specific last
+pivot above. No relation basis, WITNESS, author evaluator or PASS receipt is the
+source of those identities. See reports/algebra/REPORT.md.
+
+The independent dyadic row engine keeps integer numerators and a common power
+of two denominator. Terms are independent inputs acted on by reduced free words.
+Only adjacent generator/inverse pairs cancel; different generators do not
+commute. For every positive width f, substitute the inherited actual f-column
+operator for each generator and its inverse for the inverse symbol. Evaluation
+is an algebra homomorphism, so the unrestricted formal identities hold on all
+payload vectors, including nonseparable inputs. Rational coefficients stay
+outside evaluation. No tensor power is distributed over a sum.
+
+All new preparation coefficients use the checked same-frame ADD/CLEAR grammar.
+The emitted preparation equals the independent joint RREF assignments and has
+no pivot on any pivot right-hand side. All eight snapshots coexist before the
+first center substitution, each in a distinct cleared owned bank. Same-frame
+COPY substitutes the exact source output with its physical gauge; identity at
+a proved complete zero vector removes the later job. Identity does not transport
+an arbitrary nonzero value between frames or erase approximate error.
+
+## Chronology, physical factors and owners
+
+The independent owner parser separately checks both whole backward words and
+both unchanged forward words, all frame transitions, exact retained endpoints,
+zero retags and owned returns. Each backward has 17 distinct complete owners
+(eight pivots and C0..C8); each forward has one C0 owner. Arbitrary dirt is cleared
+only in owned fields. Every removed pivot loses its productive target. No
+snapshot, suspended center, query scratch or parked bank receives productive
+credit. The raw p16 controller needs at least 16*17+2=274 separately tagged LOCAL
+banks before its other controller/compiler storage.
+
+Every retained warp, phase marker and adapter has exact original chronology;
+canonical operators and frames are unchanged. Source signed native and paid
+phase facts are **authenticated historical reuse** from the selected PR5241
+foundation, whose full 31-file proof/checker/receipt closure is retained under
+capsule/agents/sources/pr5241. The receipt binds the original source payload
+cfcd36b40326d7a5815e1db5313444379b3fb756b27d5201c561574c085cc0dd,
+180 native classes, all 408 helpers in both orientations, and 5710 oriented paid
+classes/6596 paid event instances. The historical first-cold wrapper differs
+from canonical verify.py; its mathematical phase/native/helper checkers and
+regeneration code retain the recorded hashes. No new phase lowering is claimed.
+The canonical maintainer PROOF-REVIEW authorizes this unchanged foundation reuse.
+New scalar chronology, joint identities and owner domains are checked afresh.
+PR5384's single-relation review supplies generic COPY/zero/all-f interpretation,
+not the simultaneous joint proof or new inventory.
+
+## Whole-controller invoice, directed prices and outer point
+
+The event census computes paid ranks from actual endpoint frame dimensions,
+checks physical scalar tariffs and reads the literal p16 controller. It obtains
+64,148 paid jobs and 10,800 productive fields for each word. Relative to the
+source controller it removes 128 H1 and 128 H8 paid calls and surrenders all
+128 H9 productive targets; the rank surplus remains 752. This is a complete
+controller invoice, not eight added local savings. It includes main, idle,
+query and completion terms. See reports/prices/REPORT.md.
+
+A fresh independent fixed-point rational interval engine uses directed rounding
+at 10^-65, an atanh logarithm series and exponential Taylor remainder. It checks
+the actual signed moments, unique fixed-profile crossing, both completion tails,
+strict alpha reserve, eta/complement capacity and splitting. It independently
+recomputes both candidate parameter choices; source-identical profile reuse is
+justified by the separate literal event censuses. The raw crossing is strictly
+between .00302667476807 and .00302667476808. This is a ceiling for this exact
+invoice, not an optimum over native constructions or relation systems.
+
+| Quantity | PR5383 | PR5387 |
+| --- | --- | --- |
+| Tested complex saving | .00302667 | .003026674766 |
+| Reserved alpha slack | 1e-9 | 1e-12 |
+| Tensor exponent / cutoff / floor | 1024 / 20 / 21 | 1024 / 20 / 21 |
+| Outer gap | 1/700000000 | 1/7000000000000 |
+| Least winning j on stated sufficient grid | 44836 | 58507 |
+| Compiler H, N=2^H | 31385200000000 | 409549000000000000 |
+
+The new rational outer engine derives all 54 strict and four functional rows
+from its own exact parameter family, checks all positive, and independently
+finds the winning compiler integer and failure at the preceding sufficient-grid
+index. It compares author parameter records only after fresh computation.
+Ordinary BIT is source-bound to 17/5000=.0034 with the exact unchanged ABI;
+its native/interface admission is retained, not replayed. The compiler threshold
+is an algorithmic finite specification. The enormous expanded word is not
+materialized. See reports/outer/REPORT.md and reports/applicability/REPORT.md.
+
+The author engines' separate normal/optimized cold replay evidence is an
+external execution gate. This review's own replay independently recomputes the
+invoice, directed moments and outer predicates; it does not promote author
+RESULT status labels into a premise.
+
+## Effective expanded bounds and exact physical domain
+
+The detailed construction and strong-induction proof are retained in
+reports/applicability/REPORT.md, with exact canonical FINITE-PRECOMPILATION and
+FINITE-TEMPLATE-CLOSURE notes under capsule/repository. For the explicitly chosen
+integer batch clearing, sequential base kernels, full routes and separately
+tagged owned banks, define
+
+```
+D = 2^262144
+F(0) = D
+X_n = (n+1)*(D+1)*(F(n-1)+1)                 (n>=1)
+F(n) = X_n^(16*(n+1)^2)                    (n>=1)
+N = 2^H,  M = F(N)
+Wbar = Pbar = Obar = Rbar = Gbar = M
+Zbar = 2*M
+Ebar = 64*(Zbar+N+Pbar+Gbar+1)^3
+Bbar = Pbar+Ebar
+C0bar = 32*N*Bbar^2
+```
+
+These are terminating integer recursions and fixed finite constants, with
+candidate-specific H. D pays the squared complete role inventory and rank-router
+factors of the strictified 1024-fold unit. For top rank n, there are at most
+(D+1)^(2n+2) mixed-unit instances; product clearing over distinct smaller widths
+is at most F(n-1)^(n-1). Expanded incidence fits X_n^(3n+5); routing and coefficient
+height fit X_n^(9n+20). For every n>=1,
+16(n+1)^2-(9n+20)=16n^2+23n-4>=35, proving the strong induction. The finite probe
+of this inequality corroborates the algebraic proof; it does not replace it.
+The raw 181098496 bound is never reused as the expanded cap.
+
+The actual W/P still determines efficiency; enlarged envelopes create no
+productive outputs. The fresh semantic recurrence is
+A(e)<=A(floor(e/N))+Pbar*floor(e/N)+Ebar, with elementary leaves <=8e.
+It gives A(e)<=2*Bbar*e and complete-layer height below C0bar*d. The actual
+identity at zero retags propagates residual error through this changed program;
+there is no artificial numerical reset at a symbolic zero.
+
+For d<=c*p^(a/b), choose
+Lprec=bit_length(4*C0bar*ceil(c)),
+tprec=max(4,ceil(b*Lprec/(b-a))) and p_prec=2^tprec.
+For p>=p_prec the retained fine-grid prescription uses at most 2p bits per signed
+component. Fixed-tape class restrictions remain f<=p, A=O(p),
+K/log(p)->infinity, complete compact/control fields, restored dirty temporaries
+and the source superpolynomial record fiber. Their asymptotic domain-entry
+threshold remains the retained premise where no numerical growth rate is given.
+
+With Dord(e)=0 below N and 1+Dord(floor(e/N)) otherwise, reserve complex stock
+[Wbar*(Obar+1)]^Dord(d), **multiply** it by simultaneous BIT stock, and pad once
+to the full product. For retained BIT coefficient L_BIT, define
+Lstock=ceil(log2 Wbar)+ceil(log2(Obar+1))+L_BIT+2.
+Above d>=2^(2*(Lstock+2)), the combined reserve uses fewer than d/4 complete
+chunks; padding costs a factor below 2. Include stopping, precision, movement
+and descriptor cutoffs in the final maximum. This pays all suspended banks and
+preserves the retained movement exponent and spectator restoration.
+
+## Independence, controls and invalidation
+
+No new engine imports an author native, price, interval or endpoint evaluator.
+The only cross-engine import is the independent algebra implementation used by
+the actual-word semantic falsifier. The wrapper runs both exact words separately,
+then checks fresh coverage fields and source/K/payload identities. Fresh result
+files do not enter the frozen input inventory and cannot form a hash cycle.
+The normal/-O schedule is identical, with explicit optimization propagated to
+every scientific subprocess and fresh bytecode caches before scientific imports.
+
+Independent controls cover preparation sign/coefficient/pivot, wrong saved row,
+wrong COPY gauge, omitted relation, arbitrary owned dirt, missing owner returns,
+source chronology, borrowed erasure, storage alias, productive credit, extra H1,
+reserve, compiler grid and outer boundary failures. Six actual-word mutants
+recompute their payload digests and reach the mathematical parser directly;
+rejection is not an old-digest comparison. Source and headline mutation controls
+separately exercise the working-input guard and offline Git membership binding.
+
+An early algebra parser rejected legal signed-denominator SCALE [2,-1]; sign
+normalization corrected that checker defect before the frozen successful runs.
+Early applicability/falsifier resource failures and the wrong-marker parser
+attempt remain reported as implementation failures, never candidate evidence.
+Source phase, ordinary BIT native construction, physical precision/tape execution,
+analytic decomposition, prime avoidance and integer recovery are reused in
+explicit quantified domains. This is not an unconditional theorem, Lean/kernel
+verification, independent human acceptance, selected promotion or practical
+speedup. Any relevant word, native factor, ownership, invoice, parameter,
+scientific engine, source pin or premise edit invalidates its dependent replay.