{
  "generated_at": "2026-10-18T12:57:50.266Z",
  "passed": true,
  "total_catalogue": 1110,
  "expansion_default_output_checks": 1847,
  "available_tools": 1010,
  "review_hold": 100,
  "deterministic_reference_examples": 1108,
  "extra_independent_scenarios": 23,
  "invalid_input_checks": 7581,
  "core_boundary_groups": [
    "Expression precedence, parentheses, unary and exponent",
    "Leap dates, end of month, working days and ISO year",
    "Loan non-amortisation and exact zero-rate payoff",
    "Pack boundary floating noise and genuine fractional pack",
    "Triangle and polygon invalid geometry",
    "Critical path cycle, unknown references and parallelism",
    "Integer exactness and number conversion",
    "Network boundaries",
    "Contrast invariance, chemistry grammar and safe colour parsing",
    "Geometry scale and savings monotonicity"
  ],
  "random_tools": "Browser action/state checks; a short sequence does not prove fairness",
  "limitations": [
    "Not an external expert review",
    "Original 500: independent default references cover primary outputs, with selected secondary checks. Expansion: every default output has a reference; see additional expansion scenarios.",
    "Specialist tools held from interactive release",
    "Sources and narrower first-version scopes require editorial review before public launch"
  ],
  "errors": [],
  "interactive_experiments_worker": {
    "passed": true,
    "checks": 27,
    "production_bundle": "Exact built Worker and assets-first router matching the generated deployment config in local workerd; no mocked routes",
    "model_calls": 0,
    "source_routes": [
      "/try/",
      "/try/eleven/?from=video",
      "/try/percent-swap/?from=video",
      "/try/discounts/?from=video",
      "/index.html",
      "/find/",
      "/calculators/percentage-of-a-number/"
    ],
    "checked_at": "2026-10-18T12:58:25.198Z"
  },
  "interactive_experiments": {
    "passed": true,
    "reference_scenarios": 702,
    "independent_output_checks": 1206,
    "invalid_input_checks": 99,
    "sharing_checks": 726,
    "challenge_reference_checks": 15,
    "property_checks": 2524,
    "model_calls": 0,
    "method": "Independent Python Decimal arithmetic; final prices rounded half up to pennies; no intermediate rounding.",
    "checked_at": "2026-10-18T12:57:50.678Z"
  },
  "interactive_experiments_browser": {
    "passed": true,
    "checks": 173,
    "full_video_example_journeys": 9,
    "viewports": [
      320,
      390,
      1440
    ],
    "reduced_motion": true,
    "page_errors": [],
    "failed_requests": [],
    "api_requests": [],
    "model_calls": 0,
    "coverage": [
      "Hub to exact video example",
      "Live changed-input results and no stale invalid output",
      "Replay, pause, manual stepping, restart and keyboard",
      "Correct, incorrect, blank, hinted and revealed predictions",
      "Unique session score and reload reset",
      "Shared scenarios and hidden-answer challenge round trips",
      "Existing calculators preserve inputs",
      "Malformed, wrong-version and out-of-range link rejection",
      "99 × 11 carry boundary; sub-penny discount rounding",
      "Private page-memory events contain no input values"
    ],
    "checked_at": "2026-10-18T12:57:45.674Z"
  },
  "ai_usage": {
    "passed": true,
    "independent_scenarios": 12,
    "model_calls": 0,
    "database": "Real workerd D1, two independent bindings sharing one persistent database; process restart verified",
    "limits": {
      "siteDaily": 200,
      "userDaily": 25,
      "userMinute": 6,
      "concurrent": 3,
      "leaseMs": 90000,
      "maxInputBytes": 200000,
      "maxOutputTokens": 2600,
      "retentionDays": 7
    },
    "coverage": [
      "Atomic global concurrency",
      "User minute and daily races",
      "Global day race and persistence across restart",
      "UTC reset and crash lease expiry without refund",
      "Every retry reserved separately",
      "Fail closed without DB or during outage/kill switch",
      "Byte/output limits before network",
      "No stored questions, email or raw user identity",
      "Token/outcome metadata and lazy retention",
      "Anonymous status and sign-in"
    ],
    "checked_at": "2026-10-18T12:57:56.569Z"
  },
  "finder_chat_browser": {
    "passed": true,
    "checks": 42,
    "viewports": [
      320,
      390,
      1440
    ],
    "model_calls": 0,
    "mocked_provider_calls": 27,
    "page_errors": [],
    "coverage": [
      "Conversation and correction context",
      "No automatic calculations",
      "Independent 80 at 25% = 60, 100 at 25% = 75, 120 at 25% = 90",
      "Result cards and restore without rerunning",
      "Full-page link preserves inputs",
      "Six-turn cap prevents invalid extra API request",
      "Reload/new conversation clears memory",
      "401 sign-in, 429 retry, local fallback",
      "Cancelled stale response ignored"
    ],
    "checked_at": "2026-10-18T09:13:11.434Z"
  },
  "full_finder_numerical": {
    "passed": true,
    "supported": 1010,
    "strict_unit_contracts": 50,
    "labelled_deterministic_adapters": 958,
    "interactive_tools": 2,
    "routing_name_cases": 1010,
    "bounded_schema_cases": 1010,
    "reference_scenarios": 2178,
    "independent_output_checks": 5812,
    "invalid_request_checks": 10229,
    "model_calls": 0,
    "limitations": [
      "Reference fixtures are existing developer-worked examples, not new independent expert review.",
      "Name retrieval is not a broad paraphrase accuracy score.",
      "Labelled adapters require the displayed input units and retain original calculator guards; they add no unit conversions.",
      "Variable outputs retain runtime labels and types; no fixed default shape is assumed."
    ],
    "checked_at": "2026-10-18T12:57:54.146Z"
  },
  "full_finder_sources": {
    "passed": true,
    "checks": 984,
    "coverage": [
      "Explicit label binding across the new catalogue",
      "Never override a newer user correction with an earlier labelled value",
      "Literal record blocks retain empty cells and exact rows",
      "Unknown and alternative numeric values remain unresolved",
      "Empty list positions cannot become zeros or disappear"
    ],
    "model_calls": 0,
    "checked_at": "2026-10-18T12:57:54.573Z"
  },
  "full_finder_http": {
    "passed": true,
    "checks": 13,
    "provider_calls": 0,
    "mocked_provider_requests": 1,
    "coverage": [
      "Actual authenticated HTTP handler uses full finder registry",
      "Strict response schema and store:false preserved",
      "No answer fixtures in the model context",
      "Authentication, same-origin and content-type gates precede inference",
      "Held tools unavailable",
      "Unsupported live-data requests make no provider request"
    ],
    "checked_at": "2026-10-18T12:57:54.690Z"
  },
  "full_finder_browser": {
    "passed": true,
    "route_viewport_checks": 1058,
    "active_tools_at_390px": 1010,
    "additional_viewports": [
      320,
      1440
    ],
    "independent_display_output_checks": 2311,
    "invalid_input_result_clearing_checks": 1055,
    "page_errors": [],
    "errors": [],
    "model_calls": 0,
    "limitations": [
      "Local matching, form rendering and mathematical execution are exercised here. Live model interpretation is evaluated separately.",
      "All 1010 active tools checked at 390px; representative format/engine tools additionally checked at 320/1440px."
    ],
    "checked_at": "2026-10-18T00:12:22.960Z"
  },
  "full_finder_model_browser": {
    "passed": true,
    "route_viewport_checks": 30,
    "prepared_field_checks": 75,
    "independent_display_output_checks": 36,
    "model_calls": 0,
    "mocked_provider_calls": 30,
    "method": "Saved real model proposals pass through the authenticated HTTP handler and actual browser finder. No new inference requests.",
    "viewports": [
      320,
      390,
      1440
    ],
    "page_errors": [],
    "checked_at": "2026-10-18T09:12:22.284Z"
  },
  "full_finder_model_evaluation": {
    "date": "2026-10-18",
    "model": "gpt-6-luna",
    "reasoning": "low",
    "finder_supported": 1010,
    "strict_unit_contracts": 50,
    "labelled_deterministic_adapters": 958,
    "interactive_tools": 2,
    "held_excluded": 100,
    "drafts_excluded": 1890,
    "live_sets": {
      "named_labelled_new_tools": {
        "completed": 960,
        "passed": 960,
        "incorrect_ready": 0,
        "operational_errors": 0,
        "input_tokens": 6335450,
        "output_tokens": 166000,
        "latency_ms": {
          "median": 1918,
          "p95": 2529
        },
        "completed_at": "2026-10-18T00:25:41.483Z"
      },
      "previous_core_regression": {
        "completed": 100,
        "passed": 100,
        "incorrect_ready": 0,
        "operational_errors": 0,
        "input_tokens": 1606603,
        "output_tokens": 11255,
        "latency_ms": {
          "median": 1620,
          "p95": 2095
        },
        "completed_at": "2026-10-18T00:20:38.962Z"
      },
      "fresh_prose_holdout_before_final_sign_fix": {
        "completed": 46,
        "passed": 45,
        "incorrect_ready": 0,
        "operational_errors": 0,
        "input_tokens": 525470,
        "output_tokens": 5128,
        "latency_ms": {
          "median": 1647,
          "p95": 2429
        },
        "completed_at": "2026-10-18T00:19:46.596Z"
      },
      "fresh_signed_date_questions_on_release_code": {
        "completed": 5,
        "passed": 5,
        "incorrect_ready": 0,
        "operational_errors": 0,
        "input_tokens": 48328,
        "output_tokens": 721,
        "latency_ms": {
          "median": 1759,
          "p95": 1800
        },
        "completed_at": "2026-10-18T00:21:11.769Z"
      }
    },
    "release_replays": {
      "all-complete-qualified": {
        "source": "all-complete-qualified",
        "cases": 960,
        "passed": 960,
        "incorrect_ready": 0,
        "changed_results": 0,
        "failures": [],
        "kind": "Saved model responses revalidated on release code. Not new live model requests.",
        "model_calls": 0,
        "release_source_hashes": {
          "server/finder-core.mjs": "711d05815194fc5adbbec4e52b27029d633599611f73b81203f4b2bf86581f82",
          "server/labelled-source.mjs": "c65a883c454e02eb722bf418508776b5e6fc3bd121687cf65ed5ff322d020f91",
          "src/finder-search.mjs": "ffa652dbc745e211d2898af4d580e49384e0978596a18c870374721d556bda3b",
          "server/finder-registry.json": "d67c1465c5c2a5f243e3f1f8b1e89867925a3cd270dee42335f66d86acc80dae"
        },
        "checked_at": "2026-10-18T00:25:53.948Z"
      },
      "core-regression-v2": {
        "source": "core-regression-v2",
        "cases": 100,
        "passed": 100,
        "incorrect_ready": 0,
        "changed_results": 0,
        "failures": [],
        "kind": "Saved model responses revalidated on release code. Not new live model requests.",
        "model_calls": 0,
        "release_source_hashes": {
          "server/finder-core.mjs": "711d05815194fc5adbbec4e52b27029d633599611f73b81203f4b2bf86581f82",
          "server/labelled-source.mjs": "c65a883c454e02eb722bf418508776b5e6fc3bd121687cf65ed5ff322d020f91",
          "src/finder-search.mjs": "ffa652dbc745e211d2898af4d580e49384e0978596a18c870374721d556bda3b",
          "server/finder-registry.json": "d67c1465c5c2a5f243e3f1f8b1e89867925a3cd270dee42335f66d86acc80dae"
        },
        "checked_at": "2026-10-18T00:25:54.155Z"
      },
      "fresh-holdout-final": {
        "source": "fresh-holdout-final",
        "cases": 46,
        "passed": 46,
        "incorrect_ready": 0,
        "changed_results": 1,
        "failures": [],
        "kind": "Saved model responses revalidated on release code. Not new live model requests.",
        "model_calls": 0,
        "release_source_hashes": {
          "server/finder-core.mjs": "711d05815194fc5adbbec4e52b27029d633599611f73b81203f4b2bf86581f82",
          "server/labelled-source.mjs": "c65a883c454e02eb722bf418508776b5e6fc3bd121687cf65ed5ff322d020f91",
          "src/finder-search.mjs": "ffa652dbc745e211d2898af4d580e49384e0978596a18c870374721d556bda3b",
          "server/finder-registry.json": "d67c1465c5c2a5f243e3f1f8b1e89867925a3cd270dee42335f66d86acc80dae"
        },
        "checked_at": "2026-10-18T00:25:54.019Z"
      },
      "signed-date-fresh-release": {
        "source": "signed-date-fresh-release",
        "cases": 5,
        "passed": 5,
        "incorrect_ready": 0,
        "changed_results": 0,
        "failures": [],
        "kind": "Saved model responses revalidated on release code. Not new live model requests.",
        "model_calls": 0,
        "release_source_hashes": {
          "server/finder-core.mjs": "711d05815194fc5adbbec4e52b27029d633599611f73b81203f4b2bf86581f82",
          "server/labelled-source.mjs": "c65a883c454e02eb722bf418508776b5e6fc3bd121687cf65ed5ff322d020f91",
          "src/finder-search.mjs": "ffa652dbc745e211d2898af4d580e49384e0978596a18c870374721d556bda3b",
          "server/finder-registry.json": "d67c1465c5c2a5f243e3f1f8b1e89867925a3cd270dee42335f66d86acc80dae"
        },
        "checked_at": "2026-10-18T00:25:54.166Z"
      }
    },
    "development_results": {
      "initial_name_retrieval_misses": 65,
      "initial_full_named_run": {
        "completed": 960,
        "passed": 940,
        "incorrect_ready": 0,
        "operational_errors": 0,
        "input_tokens": 6288299,
        "output_tokens": 166921,
        "latency_ms": {
          "median": 1918,
          "p95": 2563
        },
        "completed_at": "2026-10-18T00:20:02.686Z"
      },
      "first_prose_set": {
        "completed": 42,
        "passed": 39,
        "incorrect_ready": 0,
        "operational_errors": 0,
        "input_tokens": 481349,
        "output_tokens": 4999,
        "latency_ms": {
          "median": 1577,
          "p95": 2169
        },
        "completed_at": "2026-10-18T00:14:23.886Z"
      },
      "repaired_prose_set": {
        "completed": 42,
        "passed": 42,
        "incorrect_ready": 0,
        "operational_errors": 0,
        "input_tokens": 466581,
        "output_tokens": 4811,
        "latency_ms": {
          "median": 1643,
          "p95": 2083
        },
        "completed_at": "2026-10-18T00:17:21.581Z"
      }
    },
    "policy": {
      "answers_private": true,
      "evaluator_retries": 0,
      "output_computation": "Original browser engines; model never computes a result",
      "user_action_required": true,
      "live_public_traffic": false
    },
    "limitations": [
      "Named/labelled cases test selection and supplied-field preparation, not 960 independent natural-language paraphrase families.",
      "The original 50 strict unit contracts remain unchanged. Other tools require their printed units/formats; no additional conversions or chaining are enabled.",
      "One of 46 new prose questions initially asked unnecessarily about subtracting days. Its saved response passes final host validation; five further signed-date questions passed live on that release code.",
      "All final named-tool and core results remain unchanged in release-code replay. Replays are not new live requests.",
      "Numerical references and questions are developer-authored; no external expert certification or real-user load qualification is claimed."
    ]
  },
  "development_batch2": {
    "passed": true,
    "checked_at": "2026-10-18T12:57:51.336Z",
    "new_tools": 30,
    "scenarios": 60,
    "independent_output_checks": 218,
    "invalid_checks": 68,
    "property_checks": 820,
    "ids": [
      851,
      852,
      853,
      854,
      856,
      858,
      859,
      860,
      861,
      862,
      863,
      864,
      869,
      870,
      875,
      876,
      877,
      879,
      880,
      881,
      883,
      884,
      885,
      887,
      890,
      892,
      894,
      896,
      897,
      899
    ],
    "limitations": [
      "Developer-authored fixtures, not independent human or print-shop certification.",
      "Workload and physical production figures depend on the stated measured inputs and simplified models.",
      "No new AI finder contracts or model calls."
    ],
    "errors": []
  },
  "development_batch2_browser": {
    "passed": true,
    "checked_at": "2026-10-15T19:48:17.990Z",
    "routes": 30,
    "viewports": [
      320,
      390,
      1440
    ],
    "responsive_route_checks": 90,
    "changed_input_output_checks": 327,
    "invalid_stale_result_checks": 30,
    "keyboard_submit": true,
    "errors": []
  },
  "development_batch1": {
    "passed": true,
    "checked_at": "2026-10-18T12:57:51.265Z",
    "new_tools": 20,
    "scenarios": 40,
    "independent_output_checks": 146,
    "invalid_checks": 43,
    "property_checks": 406,
    "ids": [
      501,
      502,
      503,
      505,
      509,
      510,
      524,
      537,
      538,
      541,
      545,
      550,
      601,
      602,
      613,
      615,
      618,
      621,
      632,
      636
    ],
    "limitations": [
      "Developer-authored fixtures, not independent human certification.",
      "AI finder contracts are unchanged; these tools run in the ordinary browser calculator library."
    ],
    "errors": []
  },
  "development_batch1_browser": {
    "passed": true,
    "checked_at": "2026-10-15T19:39:35.046Z",
    "routes": 20,
    "viewports": [
      320,
      390,
      1440
    ],
    "responsive_route_checks": 60,
    "changed_input_output_checks": 219,
    "invalid_stale_result_checks": 20,
    "keyboard_submit": true,
    "errors": []
  },
  "expansion": {
    "passed": true,
    "checked_at": "2026-10-18T12:57:51.200Z",
    "new_calculators": 50,
    "changed_input_cases": 50,
    "changed_input_output_checks": 127,
    "invalid_boundary_cases": 60,
    "identity_and_extended_visual_checks": 389,
    "limitations": [
      "Independent software examples and mathematical identities, not external expert certification",
      "Physical calibration and planning models depend on measured inputs and their stated assumptions"
    ]
  },
  "expansion_browser": {
    "passed": true,
    "checked_at": "2026-10-15T12:33:25.954Z",
    "changed_input_mobile_routes": 50,
    "changed_input_output_checks": 127,
    "additional_responsive_checks": 18,
    "viewports": [
      320,
      390,
      1440
    ],
    "checks": [
      "All new default and changed-input screens render",
      "All changed-input outputs match independently worked examples",
      "New methods appear on every route",
      "No page overflow at tested phone/desktop widths",
      "Representative expanded visuals inspected"
    ],
    "errors": []
  },
  "methods": {
    "passed": true,
    "checked_at": "2026-10-18T12:57:51.096Z",
    "pages": 1110,
    "source_formula_pages": 445,
    "explained_method_pages": 665,
    "documented_deterministic_outputs": 2584,
    "example_substitutions": 544,
    "random_method_pages": 2,
    "held_pages": 100,
    "checks": [
      "Every implemented output has a method",
      "Every deterministic output appears in the worked example",
      "All input symbols retain their labels and units",
      "Unsupported notation and missing methods fail the build",
      "Specialist review gates preserved"
    ],
    "limitations": [
      "Documentation coverage and agreement with the implementation, not independent expert certification",
      "Numeric reference and edge checks are reported separately; not all secondary outputs have independent fixtures"
    ]
  },
  "methods_browser": {
    "passed": true,
    "checked_at": "2026-10-15T10:03:06.805Z",
    "responsive_routes": [
      {
        "tool": 2,
        "width": 320
      },
      {
        "tool": 18,
        "width": 320
      },
      {
        "tool": 21,
        "width": 320
      },
      {
        "tool": 26,
        "width": 320
      },
      {
        "tool": 52,
        "width": 320
      },
      {
        "tool": 145,
        "width": 320
      },
      {
        "tool": 243,
        "width": 320
      },
      {
        "tool": 269,
        "width": 320
      },
      {
        "tool": 473,
        "width": 320
      },
      {
        "tool": 476,
        "width": 320
      },
      {
        "tool": 2,
        "width": 390
      },
      {
        "tool": 18,
        "width": 390
      },
      {
        "tool": 21,
        "width": 390
      },
      {
        "tool": 26,
        "width": 390
      },
      {
        "tool": 52,
        "width": 390
      },
      {
        "tool": 145,
        "width": 390
      },
      {
        "tool": 243,
        "width": 390
      },
      {
        "tool": 269,
        "width": 390
      },
      {
        "tool": 473,
        "width": 390
      },
      {
        "tool": 476,
        "width": 390
      },
      {
        "tool": 2,
        "width": 1440
      },
      {
        "tool": 18,
        "width": 1440
      },
      {
        "tool": 21,
        "width": 1440
      },
      {
        "tool": 26,
        "width": 1440
      },
      {
        "tool": 52,
        "width": 1440
      },
      {
        "tool": 145,
        "width": 1440
      },
      {
        "tool": 243,
        "width": 1440
      },
      {
        "tool": 269,
        "width": 1440
      },
      {
        "tool": 473,
        "width": 1440
      },
      {
        "tool": 476,
        "width": 1440
      }
    ],
    "checks": [
      "No page overflow at 320, 390 and 1440 pixels",
      "Method symbol disclosure opens and closes",
      "Worked example remains labelled and fixed while live inputs change",
      "Load example restores the documented answers"
    ],
    "errors": []
  },
  "browser": {
    "passed": true,
    "routes": 550,
    "checks": [
      "Search and locally saved tools",
      "Live recalculation, invalid state, reset, share round-trip and CSV",
      "Mobile layout, random actions, reset, name output escaping",
      "All 550 routes, including 100 gated specialist pages",
      "Representative numeric, text, matrix and date layouts at 390px"
    ],
    "errors": [],
    "checked_at": "2026-10-15T14:01:12.709Z"
  },
  "ai": {
    "passed": true,
    "checks": 126,
    "model_calls_in_this_suite": 0,
    "coverage": [
      "Explicit numeric evidence",
      "No invented zeros or defaults",
      "Held tools blocked",
      "Supported conversions only",
      "Required inputs and calculation guards",
      "Authentication and same-origin requirements",
      "Body size and conversation limits",
      "Provider error sanitization",
      "Incomplete model output",
      "Durable per-user attempt throttling and fast isolate backstop",
      "Strict schema and store:false"
    ],
    "checked_at": "2026-10-18T12:57:54.792Z"
  },
  "ai_recovery": {
    "passed": true,
    "checks": 26,
    "policy": {
      "attemptMs": 12000,
      "totalMs": 25000,
      "maxAttempts": 2,
      "baseDelayMs": 250,
      "jitterMs": 250,
      "maxRetryAfterMs": 1000
    },
    "model_calls": 0,
    "coverage": [
      "Stalled headers and body",
      "One recovery attempt",
      "Total deadline",
      "Client cancellation before/during/backoff",
      "Transient HTTP errors and network resets",
      "Quota/auth/invalid requests never retried",
      "Long Retry-After deferred",
      "Model validation failures never retried",
      "Attempt telemetry and real HTTP handler integration"
    ],
    "checked_at": "2026-10-18T12:57:56.855Z"
  },
  "ai_field_mapping": {
    "passed": true,
    "checks": 48,
    "model_calls": 0,
    "coverage": [
      "Descriptive calculator-scoped keys",
      "Separate scalar/list schema branches",
      "One-item and multiple-item lists",
      "Missing middle field and arbitrary field order",
      "Evidence, currency and unit guards preserved",
      "Duplicate aliases and cross-tool keys rejected",
      "No positional reassignment",
      "Per-item price and cost wording",
      "Reject denominator borrowed from other amounts",
      "Require explicit dimension-unit evidence",
      "Honour explicit choice requests with negation",
      "Reject superseded bag weights"
    ],
    "checked_at": "2026-10-18T12:57:56.915Z"
  },
  "ai_retention": {
    "passed": true,
    "checks": 47,
    "model_calls": 0,
    "checked_at": "2026-10-18T12:57:56.975Z"
  },
  "chemistry_boundaries": {
    "passed": true,
    "checks": 16,
    "repaired_tools": [
      485,
      488,
      499
    ],
    "holds_removed": 0,
    "model_calls": 0,
    "scope": "Boundary repairs and independently worked examples; not specialist signoff.",
    "checked_at": "2026-10-18T12:57:51.624Z"
  },
  "ai_batch50": {
    "passed": true,
    "added_tools": [
      36,
      63,
      76,
      77,
      83,
      101,
      102,
      103,
      104,
      105,
      106,
      107
    ],
    "changed_input_cases": 24,
    "independent_output_checks": 30,
    "invalid_cases": 12,
    "engine_changed": false,
    "checked_at": "2026-10-15T18:34:16.386Z"
  },
  "ai_batch50_browser": {
    "passed": true,
    "new_tools_exercised": 12,
    "mocked_provider": true,
    "changed_input_outputs_checked": 15,
    "viewports": [
      320,
      390,
      1440
    ],
    "checks": [
      "Prepared scalar and list inputs",
      "No calculation before review",
      "Changed-input numerical outputs",
      "No overflow or browser errors"
    ],
    "checked_at": "2026-10-15T18:25:36.424Z"
  },
  "ai_expand38": {
    "passed": true,
    "added_tools": [
      "CC-041",
      "CC-062",
      "CC-080"
    ],
    "changed_input_cases": 8,
    "independent_output_checks": 12,
    "rejected_invalid_cases": 6,
    "math_engine_changed": false,
    "checked_at": "2026-10-15T17:24:38.844Z"
  },
  "ai_expand38_browser": {
    "passed": true,
    "new_tools_exercised": 3,
    "mocked_provider": true,
    "changed_input_outputs_checked": 5,
    "viewports": [
      320,
      390,
      1440
    ],
    "checks": [
      "Prepared scalar and list inputs",
      "No calculation before review",
      "Changed-input numerical outputs",
      "No overflow or browser errors"
    ],
    "checked_at": "2026-10-15T17:22:04.462Z"
  },
  "ai_expansion": {
    "passed": true,
    "added_tools": [
      "CC-003",
      "CC-007",
      "CC-014",
      "CC-015",
      "CC-016",
      "CC-017",
      "CC-028",
      "CC-031",
      "CC-032",
      "CC-040"
    ],
    "changed_input_cases": 20,
    "independent_output_checks": 24,
    "rejected_invalid_cases": 21,
    "written_list_and_scalar_disambiguation": true,
    "math_engine_changed": false,
    "checked_at": "2026-10-15T17:21:57.076Z"
  },
  "ai_expansion_browser": {
    "passed": true,
    "new_tools_exercised": 10,
    "mocked_provider": true,
    "changed_input_outputs_checked": 12,
    "viewports": [
      320,
      390,
      1440
    ],
    "checks": [
      "Prepared scalar and list inputs",
      "Negative values in supported lists",
      "No calculation before review",
      "Changed-input numerical outputs",
      "No overflow or browser errors"
    ],
    "checked_at": "2026-10-15T16:22:26.287Z"
  },
  "ai_evaluation": {
    "checked_at": "2026-10-15T18:34:16.500Z",
    "model": "gpt-6-luna",
    "reasoning": "low",
    "executable_tools": 50,
    "added_tools": [
      "CC-036",
      "CC-063",
      "CC-076",
      "CC-077",
      "CC-083",
      "CC-101",
      "CC-102",
      "CC-103",
      "CC-104",
      "CC-105",
      "CC-106",
      "CC-107"
    ],
    "previous_release_evaluation": {
      "checked_at": "2026-10-15T17:29:09.587853+00:00",
      "model": "gpt-6-luna",
      "reasoning": "low",
      "executable_tools": 38,
      "added_tools": [
        "CC-041",
        "CC-062",
        "CC-080"
      ],
      "deferred_tools": [
        "CC-043",
        "CC-081"
      ],
      "previous_release_evaluation": {
        "checked_at": "2026-10-15T16:55:42.714381+00:00",
        "model": "gpt-6-luna",
        "reasoning": "low",
        "previous_release_evaluation": {
          "checked_at": "2026-10-15T16:26:04.666441+00:00",
          "model": "gpt-6-luna",
          "reasoning": "low",
          "previous_release_evaluation": {
            "checked_at": "2026-10-15T15:47:25.823630+00:00",
            "model": "gpt-6-luna",
            "reasoning": "low",
            "previous_release_evaluation": {
              "checked_at": "2026-10-15T15:05:50.437378+00:00",
              "model": "gpt-6-luna",
              "reasoning": "low",
              "baseline": {
                "cases": 300,
                "passed": 289,
                "incorrect_ready": 0,
                "operational_errors": 3,
                "latency_ms": {
                  "median": 2326,
                  "p95": 6937,
                  "max": 10986
                },
                "dataset_sha256": "fdb38234b38cebb956b2f92180f5555efbfc53307d77ad55b606f0e56fe67d45",
                "acceptance_passed": false
              },
              "live_regression": {
                "cases": 300,
                "passed": 298,
                "incorrect_ready": 0,
                "operational_errors": 0,
                "latency_ms": {
                  "median": 2118,
                  "p95": 6993,
                  "max": 10116
                },
                "dataset_sha256": "fdb38234b38cebb956b2f92180f5555efbfc53307d77ad55b606f0e56fe67d45",
                "acceptance_passed": true
              },
              "first_fresh_evaluation": {
                "cases": 100,
                "passed": 92,
                "incorrect_ready": 0,
                "operational_errors": 0,
                "latency_ms": {
                  "median": 2178,
                  "p95": 5746,
                  "max": 11099
                },
                "dataset_sha256": "cf5d26a4266854cd51da71febb346ae2daf23c4e1efec1795ce609e63e7efdd8",
                "acceptance_passed": false
              },
              "post_guard_replay": {
                "calculator-ai-repair-release-20261015": {
                  "cases": 300,
                  "passed": 298,
                  "improvements": [],
                  "new_regressions": 0
                },
                "calculator-ai-fresh-20261015": {
                  "cases": 100,
                  "passed": 93,
                  "improvements": [
                    "F082"
                  ],
                  "new_regressions": 0
                }
              },
              "post_guard_replay_is_live_model_evaluation": false,
              "runtime_sha256": "1294633871dcf3d720a7f80cae84062a67d9f3f3e774eaaab609d0424a1cd962",
              "expansion_qualified": false,
              "enabled_ai_tools": 25,
              "limitations": [
                "Synthetic cases authored by coding assistant, not an independent human audit",
                "One live attempt per final case; no reliability guarantee",
                "First fresh evaluation remains 92/100; replay does not overwrite it",
                "No changes to numerical engines",
                "No public launch; preserve private audience"
              ]
            },
            "first_natural_language_regression": {
              "cases": 400,
              "passed": 399,
              "incorrect_ready": 0,
              "operational_errors": 0,
              "by_group": {
                "complete": {
                  "cases": 200,
                  "passed": 200,
                  "pass_rate": 1,
                  "incorrect_ready": 0,
                  "operational_errors": 0
                },
                "missing": {
                  "cases": 60,
                  "passed": 60,
                  "pass_rate": 1,
                  "incorrect_ready": 0,
                  "operational_errors": 0
                },
                "correction": {
                  "cases": 35,
                  "passed": 35,
                  "pass_rate": 1,
                  "incorrect_ready": 0,
                  "operational_errors": 0
                },
                "conversion": {
                  "cases": 25,
                  "passed": 25,
                  "pass_rate": 1,
                  "incorrect_ready": 0,
                  "operational_errors": 0
                },
                "ambiguous": {
                  "cases": 30,
                  "passed": 30,
                  "pass_rate": 1,
                  "incorrect_ready": 0,
                  "operational_errors": 0
                },
                "invalid": {
                  "cases": 20,
                  "passed": 20,
                  "pass_rate": 1,
                  "incorrect_ready": 0,
                  "operational_errors": 0
                },
                "unsupported": {
                  "cases": 30,
                  "passed": 29,
                  "pass_rate": 0.9666666666666667,
                  "incorrect_ready": 0,
                  "operational_errors": 0
                }
              },
              "latency_ms": {
                "median": 2223,
                "p95": 7924,
                "max": 14696
              },
              "dataset_sha256": "3ad41676b6c57ba7814667bbace3f0246518f662b49362b0c405355029571bef",
              "acceptance_passed": true
            },
            "first_natural_language_fresh": {
              "cases": 100,
              "passed": 97,
              "incorrect_ready": 1,
              "operational_errors": 0,
              "by_group": {
                "complete": {
                  "cases": 50,
                  "passed": 50,
                  "pass_rate": 1,
                  "incorrect_ready": 0,
                  "operational_errors": 0
                },
                "missing": {
                  "cases": 10,
                  "passed": 9,
                  "pass_rate": 0.9,
                  "incorrect_ready": 0,
                  "operational_errors": 0
                },
                "conversion": {
                  "cases": 10,
                  "passed": 10,
                  "pass_rate": 1,
                  "incorrect_ready": 0,
                  "operational_errors": 0
                },
                "correction": {
                  "cases": 10,
                  "passed": 9,
                  "pass_rate": 0.9,
                  "incorrect_ready": 0,
                  "operational_errors": 0
                },
                "ambiguous": {
                  "cases": 10,
                  "passed": 9,
                  "pass_rate": 0.9,
                  "incorrect_ready": 1,
                  "operational_errors": 0
                },
                "invalid": {
                  "cases": 5,
                  "passed": 5,
                  "pass_rate": 1,
                  "incorrect_ready": 0,
                  "operational_errors": 0
                },
                "unsupported": {
                  "cases": 5,
                  "passed": 5,
                  "pass_rate": 1,
                  "incorrect_ready": 0,
                  "operational_errors": 0
                }
              },
              "latency_ms": {
                "median": 2176,
                "p95": 4020,
                "max": 5365
              },
              "dataset_sha256": "b6c38a2e3a43dbfea727c7064c85f079c73721b8021c402240901e5620b28491",
              "acceptance_passed": false
            },
            "final_regression": {
              "cases": 500,
              "passed": 499,
              "incorrect_ready": 0,
              "operational_errors": 1,
              "by_group": {
                "complete": {
                  "cases": 250,
                  "passed": 249,
                  "pass_rate": 0.996,
                  "incorrect_ready": 0,
                  "operational_errors": 1
                },
                "missing": {
                  "cases": 70,
                  "passed": 70,
                  "pass_rate": 1,
                  "incorrect_ready": 0,
                  "operational_errors": 0
                },
                "correction": {
                  "cases": 45,
                  "passed": 45,
                  "pass_rate": 1,
                  "incorrect_ready": 0,
                  "operational_errors": 0
                },
                "conversion": {
                  "cases": 35,
                  "passed": 35,
                  "pass_rate": 1,
                  "incorrect_ready": 0,
                  "operational_errors": 0
                },
                "ambiguous": {
                  "cases": 40,
                  "passed": 40,
                  "pass_rate": 1,
                  "incorrect_ready": 0,
                  "operational_errors": 0
                },
                "invalid": {
                  "cases": 25,
                  "passed": 25,
                  "pass_rate": 1,
                  "incorrect_ready": 0,
                  "operational_errors": 0
                },
                "unsupported": {
                  "cases": 35,
                  "passed": 35,
                  "pass_rate": 1,
                  "incorrect_ready": 0,
                  "operational_errors": 0
                }
              },
              "latency_ms": {
                "median": 2506,
                "p95": 7892,
                "max": 25002
              },
              "dataset_sha256": "e13adf6807d30c0b18c771f7271571739f703d453e359b01986125a83d9625f7",
              "acceptance_passed": false
            },
            "final_fresh_first_attempt": {
              "cases": 50,
              "passed": 50,
              "incorrect_ready": 0,
              "operational_errors": 0,
              "by_group": {
                "complete": {
                  "cases": 25,
                  "passed": 25,
                  "pass_rate": 1,
                  "incorrect_ready": 0,
                  "operational_errors": 0
                },
                "missing": {
                  "cases": 5,
                  "passed": 5,
                  "pass_rate": 1,
                  "incorrect_ready": 0,
                  "operational_errors": 0
                },
                "conversion": {
                  "cases": 5,
                  "passed": 5,
                  "pass_rate": 1,
                  "incorrect_ready": 0,
                  "operational_errors": 0
                },
                "correction": {
                  "cases": 5,
                  "passed": 5,
                  "pass_rate": 1,
                  "incorrect_ready": 0,
                  "operational_errors": 0
                },
                "ambiguous": {
                  "cases": 5,
                  "passed": 5,
                  "pass_rate": 1,
                  "incorrect_ready": 0,
                  "operational_errors": 0
                },
                "invalid": {
                  "cases": 3,
                  "passed": 3,
                  "pass_rate": 1,
                  "incorrect_ready": 0,
                  "operational_errors": 0
                },
                "unsupported": {
                  "cases": 2,
                  "passed": 2,
                  "pass_rate": 1,
                  "incorrect_ready": 0,
                  "operational_errors": 0
                }
              },
              "latency_ms": {
                "median": 2315,
                "p95": 6744,
                "max": 10923
              },
              "dataset_sha256": "0575be96ddaa99ec046a8c25f737425cbe86596a8fc7fbc40895134e67789dd4",
              "acceptance_passed": true
            },
            "runtime_hashes": {
              "server/number-words.mjs": "335a2e3b649bf64b06f6403dd513c1c08e61099957b03c8aa5761c58e5e6174a",
              "server/ai-core.mjs": "391ebbedd7289a35692fd13e60ff64d199e4aa71cd5af93dc0a4243a91e06ae8",
              "server/pilot-registry.json": "28b9affc6a7d4793a3c0080cd56a250d37e0043456e6e65bee69346a6094413e",
              "src/orchestrator.mjs": "374db3af1161b891a78de6ac464b1873f5c2e7433206095de7875c23b09f6349",
              "src/engine.mjs": "46ae7dcac171c33ee97e06e124a9af73bca0877800e4a4d35e341d020b5eec59",
              "src/catalog.mjs": "18dc5fe430df9b20d20c510cfe63fd64da16940a13411a3c24e229e1ba13ef25",
              "src/math.mjs": "6e8caa526f94621f0ee7455b890e0987645976be5758b6b79f2f57df398d9a19",
              "src/expansion-math.mjs": "0674e5b6a4567e010d52cb213b0c232aa88da609a52e57a4523cee5ba74ec378"
            },
            "expansion_qualified": false,
            "enabled_ai_tools": 25,
            "remaining_failures": [
              {
                "id": "F015",
                "group": "complete",
                "errors": [
                  {
                    "kind": "operational",
                    "message": "The operation was aborted due to timeout"
                  }
                ],
                "incorrect_ready": false
              }
            ],
            "limitations": [
              "Synthetic cases authored and graded by the coding assistant, not independent human certification",
              "No retries replace final failures; fresh answers were frozen and never included in requests",
              "One pass is not a reliability guarantee or production load test",
              "No numerical engine changes; existing reference tests retained",
              "Model or host clarification is not universally overridden",
              "GBP only, bounded English cardinals and enumerated unit conversions",
              "Private pilot; public launch still requires durable quotas and wider review"
            ]
          },
          "timeout_qualification": {
            "cases": 550,
            "passed": 549,
            "incorrect_ready": 0,
            "operational_errors": 0,
            "by_group": {
              "complete": {
                "cases": 275,
                "passed": 275,
                "pass_rate": 1,
                "incorrect_ready": 0,
                "operational_errors": 0
              },
              "missing": {
                "cases": 75,
                "passed": 75,
                "pass_rate": 1,
                "incorrect_ready": 0,
                "operational_errors": 0
              },
              "correction": {
                "cases": 50,
                "passed": 49,
                "pass_rate": 0.98,
                "incorrect_ready": 0,
                "operational_errors": 0
              },
              "conversion": {
                "cases": 40,
                "passed": 40,
                "pass_rate": 1,
                "incorrect_ready": 0,
                "operational_errors": 0
              },
              "ambiguous": {
                "cases": 45,
                "passed": 45,
                "pass_rate": 1,
                "incorrect_ready": 0,
                "operational_errors": 0
              },
              "invalid": {
                "cases": 28,
                "passed": 28,
                "pass_rate": 1,
                "incorrect_ready": 0,
                "operational_errors": 0
              },
              "unsupported": {
                "cases": 37,
                "passed": 37,
                "pass_rate": 1,
                "incorrect_ready": 0,
                "operational_errors": 0
              }
            },
            "latency_ms": {
              "median": 2010,
              "p95": 3693,
              "max": 15021
            },
            "dataset_sha256": "e13950c9de0fc966ecf32d056eccd3a6a1f22c82a9dd7182723b2b393ecb1dae",
            "acceptance_passed": true,
            "transport": {
              "attempts": 551,
              "retried_requests": 1,
              "recovered_requests": 1,
              "attempt_timeouts": 1
            }
          },
          "first_expansion_fresh": {
            "cases": 100,
            "passed": 92,
            "incorrect_ready": 1,
            "operational_errors": 0,
            "by_group": {
              "complete": {
                "cases": 40,
                "passed": 34,
                "pass_rate": 0.85,
                "incorrect_ready": 0,
                "operational_errors": 0
              },
              "missing": {
                "cases": 20,
                "passed": 19,
                "pass_rate": 0.95,
                "incorrect_ready": 0,
                "operational_errors": 0
              },
              "correction": {
                "cases": 10,
                "passed": 10,
                "pass_rate": 1,
                "incorrect_ready": 0,
                "operational_errors": 0
              },
              "invalid": {
                "cases": 20,
                "passed": 20,
                "pass_rate": 1,
                "incorrect_ready": 0,
                "operational_errors": 0
              },
              "ambiguous": {
                "cases": 10,
                "passed": 9,
                "pass_rate": 0.9,
                "incorrect_ready": 1,
                "operational_errors": 0
              }
            },
            "latency_ms": {
              "median": 1833,
              "p95": 3796,
              "max": 8684
            },
            "dataset_sha256": "9146045b9f3fbf0b4a58966ffcb2085c44a8a3d5f3924565e86777947184de69",
            "acceptance_passed": false,
            "transport": {
              "attempts": 100,
              "retried_requests": 0,
              "recovered_requests": 0,
              "attempt_timeouts": 0
            }
          },
          "stopped_candidate_regression": {
            "stopped_at": "2026-10-15T16:15:42.653954+00:00",
            "reason": "New-tool first-attempt evaluation exposed parser/annual-rate bugs. Candidate withdrawn for repair; preserve partial run, do not count as a complete regression.",
            "completed": 312,
            "passed": 311,
            "failures": [
              "H027"
            ],
            "in_flight_requests_may_have_unreported_usage": true
          },
          "targeted_development": {
            "cases": 8,
            "passed": 7,
            "incorrect_ready": 0,
            "operational_errors": 0,
            "by_group": {
              "complete": {
                "cases": 6,
                "passed": 6,
                "pass_rate": 1,
                "incorrect_ready": 0,
                "operational_errors": 0
              },
              "missing": {
                "cases": 1,
                "passed": 0,
                "pass_rate": 0,
                "incorrect_ready": 0,
                "operational_errors": 0
              },
              "ambiguous": {
                "cases": 1,
                "passed": 1,
                "pass_rate": 1,
                "incorrect_ready": 0,
                "operational_errors": 0
              }
            },
            "latency_ms": {
              "median": 1632,
              "p95": 2445,
              "max": 2445
            },
            "dataset_sha256": "83cdd21382cd65dad7381a2c0610519f825f6798d95c60a7f8b1a123d744f975",
            "acceptance_passed": false,
            "transport": {
              "attempts": 8,
              "retried_requests": 0,
              "recovered_requests": 0,
              "attempt_timeouts": 0
            }
          },
          "final_regression": {
            "cases": 650,
            "passed": 646,
            "incorrect_ready": 0,
            "operational_errors": 0,
            "by_group": {
              "complete": {
                "cases": 315,
                "passed": 314,
                "pass_rate": 0.9968253968253968,
                "incorrect_ready": 0,
                "operational_errors": 0
              },
              "missing": {
                "cases": 95,
                "passed": 93,
                "pass_rate": 0.9789473684210527,
                "incorrect_ready": 0,
                "operational_errors": 0
              },
              "correction": {
                "cases": 60,
                "passed": 59,
                "pass_rate": 0.9833333333333333,
                "incorrect_ready": 0,
                "operational_errors": 0
              },
              "conversion": {
                "cases": 40,
                "passed": 40,
                "pass_rate": 1,
                "incorrect_ready": 0,
                "operational_errors": 0
              },
              "ambiguous": {
                "cases": 55,
                "passed": 55,
                "pass_rate": 1,
                "incorrect_ready": 0,
                "operational_errors": 0
              },
              "invalid": {
                "cases": 48,
                "passed": 48,
                "pass_rate": 1,
                "incorrect_ready": 0,
                "operational_errors": 0
              },
              "unsupported": {
                "cases": 37,
                "passed": 37,
                "pass_rate": 1,
                "incorrect_ready": 0,
                "operational_errors": 0
              }
            },
            "latency_ms": {
              "median": 1871,
              "p95": 3234,
              "max": 10033
            },
            "dataset_sha256": "1cbcef1c356e028a36083195caaf0a8238519cc706cbebca3b38f7133de8b87e",
            "acceptance_passed": true,
            "transport": {
              "attempts": 650,
              "retried_requests": 0,
              "recovered_requests": 0,
              "attempt_timeouts": 0
            }
          },
          "final_fresh_first_attempt": {
            "cases": 40,
            "passed": 40,
            "incorrect_ready": 0,
            "operational_errors": 0,
            "by_group": {
              "complete": {
                "cases": 20,
                "passed": 20,
                "pass_rate": 1,
                "incorrect_ready": 0,
                "operational_errors": 0
              },
              "missing": {
                "cases": 10,
                "passed": 10,
                "pass_rate": 1,
                "incorrect_ready": 0,
                "operational_errors": 0
              },
              "invalid": {
                "cases": 5,
                "passed": 5,
                "pass_rate": 1,
                "incorrect_ready": 0,
                "operational_errors": 0
              },
              "ambiguous": {
                "cases": 5,
                "passed": 5,
                "pass_rate": 1,
                "incorrect_ready": 0,
                "operational_errors": 0
              }
            },
            "latency_ms": {
              "median": 1666,
              "p95": 2282,
              "max": 10531
            },
            "dataset_sha256": "8f0c8d03e63ea1417a3ff373947fff539622e797b6058d1b40f79d8142779d4c",
            "acceptance_passed": true,
            "transport": {
              "attempts": 40,
              "retried_requests": 0,
              "recovered_requests": 0,
              "attempt_timeouts": 0
            }
          },
          "runtime_hashes": {
            "src/pilot-contracts.mjs": "c0ac3f8592e8769f744d590e6a7923a9dfbfc8e0ad2f1e89b202cbf889bcb881",
            "server/provider-request.mjs": "c993a2a57d4a6d6dac0b3d1d89acfcb1ef2b563389f61a1a56b9023818cd0f79",
            "server/ai-http.mjs": "02713a469671ac6ddba2df33636dbedb8a96c48ff971bed8a9d1743a51091c22",
            "server/number-words.mjs": "6db9cafec5a461923497c1061d96489eda7c3a5f96d8a1e465582b367fa81b92",
            "server/ai-core.mjs": "999a31ab5ac913c1a358a2f45b0829befe442c3b6bc048606abe70badf52b4e8",
            "server/pilot-registry.json": "cefb40c7307583a1bf72e45875da25c03580c727e562070d4c66f04a348e8403",
            "src/orchestrator.mjs": "2ecf87b958e8e65a020e1eec60006f13bf7b7ada2e7930f1850ac567fc0bbbae",
            "src/engine.mjs": "46ae7dcac171c33ee97e06e124a9af73bca0877800e4a4d35e341d020b5eec59",
            "src/catalog.mjs": "18dc5fe430df9b20d20c510cfe63fd64da16940a13411a3c24e229e1ba13ef25",
            "src/math.mjs": "6e8caa526f94621f0ee7455b890e0987645976be5758b6b79f2f57df398d9a19",
            "src/expansion-math.mjs": "0674e5b6a4567e010d52cb213b0c232aa88da609a52e57a4523cee5ba74ec378"
          },
          "expansion_qualified": true,
          "enabled_ai_tools": 35,
          "remaining_failures": [
            {
              "id": "H027",
              "group": "missing",
              "errors": [
                {
                  "kind": "value",
                  "field": "b",
                  "expected": [
                    150
                  ],
                  "actual": null
                }
              ],
              "incorrect_ready": false
            },
            {
              "id": "H038",
              "group": "correction",
              "errors": [
                {
                  "kind": "status",
                  "expected": [
                    "ready"
                  ],
                  "actual": "clarify"
                },
                {
                  "kind": "value",
                  "field": "a",
                  "expected": [
                    7,
                    15
                  ],
                  "actual": null
                },
                {
                  "kind": "value",
                  "field": "b",
                  "expected": [
                    120
                  ],
                  "actual": null
                }
              ],
              "incorrect_ready": false
            },
            {
              "id": "X060",
              "group": "missing",
              "errors": [
                {
                  "kind": "value",
                  "field": "c",
                  "expected": 12,
                  "actual": null
                }
              ],
              "incorrect_ready": false
            },
            {
              "id": "F034",
              "group": "complete",
              "errors": [
                {
                  "kind": "status",
                  "expected": [
                    "ready"
                  ],
                  "actual": "clarify"
                },
                {
                  "kind": "value",
                  "field": "b",
                  "expected": 16,
                  "actual": null
                }
              ],
              "incorrect_ready": false
            }
          ],
          "limitations": [
            "40 new questions for the ten added tools, frozen before final repair; first live attempts. Authored by the coding assistant, not independently blinded.",
            "One application request per question. Application may retry transient transport failures once within 25 seconds; evaluator never retries or replaces failed answers.",
            "Runtime, grading and answer fixtures frozen throughout each run.",
            "Routing/form preparation checked independently against answer fixtures; model never sees expected answers.",
            "Gate >=95% overall, >=98% complete, zero incorrect-ready forms and final operational errors.",
            "Not a real-user traffic estimate, independent human audit or load test.",
            "Private pilot; no public load test or durable global billing cap",
            "Retry may incur additional usage; unreported timeout usage is unknown",
            "Numerical engines unchanged; user reviews prepared inputs before calculation"
          ]
        },
        "targeted_development": {
          "cases": 20,
          "passed": 20,
          "incorrect_ready": 0,
          "operational_errors": 0,
          "by_group": {
            "complete": {
              "cases": 5,
              "passed": 5,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "missing": {
              "cases": 10,
              "passed": 10,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "correction": {
              "cases": 5,
              "passed": 5,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            }
          },
          "latency_ms": {
            "median": 1988,
            "p95": 3785,
            "max": 8900
          },
          "dataset_sha256": "7dff6a17f932e2f34dde8e654009bd849094f1495f00cf8c669d94cdc0cbc7cb",
          "acceptance_passed": true,
          "transport": {
            "attempts": 20,
            "retried_requests": 0,
            "recovered_requests": 0,
            "attempt_timeouts": 0
          }
        },
        "live_regression_before_choice_guard": {
          "cases": 750,
          "passed": 747,
          "incorrect_ready": 2,
          "operational_errors": 0,
          "by_group": {
            "complete": {
              "cases": 363,
              "passed": 363,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "missing": {
              "cases": 119,
              "passed": 118,
              "pass_rate": 0.9915966386554622,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "correction": {
              "cases": 67,
              "passed": 66,
              "pass_rate": 0.9850746268656716,
              "incorrect_ready": 1,
              "operational_errors": 0
            },
            "conversion": {
              "cases": 40,
              "passed": 40,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "ambiguous": {
              "cases": 66,
              "passed": 65,
              "pass_rate": 0.9848484848484849,
              "incorrect_ready": 1,
              "operational_errors": 0
            },
            "invalid": {
              "cases": 58,
              "passed": 58,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "unsupported": {
              "cases": 37,
              "passed": 37,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            }
          },
          "latency_ms": {
            "median": 2109,
            "p95": 4187,
            "max": 14634
          },
          "dataset_sha256": "51a3c938933453e9cbb5fb9dbfbfb83063596f79388ac20a9c32a28dea381796",
          "acceptance_passed": false,
          "transport": {
            "attempts": 751,
            "retried_requests": 1,
            "recovered_requests": 1,
            "attempt_timeouts": 1
          }
        },
        "final_saved_response_replay": {
          "cases": 750,
          "passed": 748,
          "model_calls": 0,
          "method": "Revalidate all saved live proposals after host-only guards; not a new live model evaluation.",
          "incorrect_ready": 0,
          "operational_errors": 0,
          "new_regressions": 0,
          "improved": [
            "E259"
          ],
          "by_group": {
            "missing": {
              "cases": 119,
              "passed": 118
            },
            "complete": {
              "cases": 363,
              "passed": 363
            },
            "conversion": {
              "cases": 40,
              "passed": 40
            },
            "invalid": {
              "cases": 58,
              "passed": 58
            },
            "correction": {
              "cases": 67,
              "passed": 66
            },
            "unsupported": {
              "cases": 37,
              "passed": 37
            },
            "ambiguous": {
              "cases": 66,
              "passed": 66
            }
          },
          "failures": [
            {
              "id": "E40-Y025",
              "errors": [
                {
                  "kind": "value",
                  "field": "a",
                  "expected": 34.8765,
                  "actual": null
                }
              ],
              "incorrect_ready": false
            },
            {
              "id": "N075",
              "errors": [
                {
                  "kind": "status",
                  "expected": [
                    "ready"
                  ],
                  "actual": "clarify"
                },
                {
                  "kind": "value",
                  "field": "c",
                  "expected": 25,
                  "actual": null
                }
              ],
              "incorrect_ready": false
            }
          ],
          "runtime_hashes": {
            "src/pilot-contracts.mjs": "c0ac3f8592e8769f744d590e6a7923a9dfbfc8e0ad2f1e89b202cbf889bcb881",
            "server/provider-request.mjs": "c993a2a57d4a6d6dac0b3d1d89acfcb1ef2b563389f61a1a56b9023818cd0f79",
            "server/ai-http.mjs": "02713a469671ac6ddba2df33636dbedb8a96c48ff971bed8a9d1743a51091c22",
            "server/number-words.mjs": "6db9cafec5a461923497c1061d96489eda7c3a5f96d8a1e465582b367fa81b92",
            "server/ai-core.mjs": "c037697f3178442345348232d29c5e09e18c6ee38a00a6c936c65e8a570b3073",
            "server/pilot-registry.json": "cefb40c7307583a1bf72e45875da25c03580c727e562070d4c66f04a348e8403",
            "src/orchestrator.mjs": "2ecf87b958e8e65a020e1eec60006f13bf7b7ada2e7930f1850ac567fc0bbbae",
            "src/engine.mjs": "46ae7dcac171c33ee97e06e124a9af73bca0877800e4a4d35e341d020b5eec59",
            "src/catalog.mjs": "18dc5fe430df9b20d20c510cfe63fd64da16940a13411a3c24e229e1ba13ef25",
            "src/math.mjs": "6e8caa526f94621f0ee7455b890e0987645976be5758b6b79f2f57df398d9a19",
            "src/expansion-math.mjs": "0674e5b6a4567e010d52cb213b0c232aa88da609a52e57a4523cee5ba74ec378"
          },
          "checked_at": "2026-10-15T16:54:43.612Z",
          "acceptance_passed": true
        },
        "final_release_fresh": {
          "cases": 30,
          "passed": 30,
          "incorrect_ready": 0,
          "operational_errors": 0,
          "by_group": {
            "ambiguous": {
              "cases": 10,
              "passed": 10,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "complete": {
              "cases": 16,
              "passed": 16,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "missing": {
              "cases": 4,
              "passed": 4,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            }
          },
          "latency_ms": {
            "median": 1995,
            "p95": 3980,
            "max": 9067
          },
          "dataset_sha256": "38e65947ee19ed9a1ac3feb0920036b7832ab6ca10a0e5ee06957106a5165760",
          "acceptance_passed": true,
          "transport": {
            "attempts": 30,
            "retried_requests": 0,
            "recovered_requests": 0,
            "attempt_timeouts": 0
          }
        },
        "final_correction_fresh": {
          "cases": 12,
          "passed": 12,
          "incorrect_ready": 0,
          "operational_errors": 0,
          "by_group": {
            "complete": {
              "cases": 9,
              "passed": 9,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "missing": {
              "cases": 3,
              "passed": 3,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            }
          },
          "latency_ms": {
            "median": 2026,
            "p95": 10656,
            "max": 10656
          },
          "dataset_sha256": "11b1421718ed05433ac4888c91b2996e9e0f326b8550217e5e28b988a3030839",
          "acceptance_passed": true,
          "transport": {
            "attempts": 12,
            "retried_requests": 0,
            "recovered_requests": 0,
            "attempt_timeouts": 0
          }
        },
        "initial_fresh_original_grading": {
          "cases": 60,
          "passed": 59,
          "incorrect_ready": 0,
          "operational_errors": 0,
          "by_group": {
            "complete": {
              "cases": 28,
              "passed": 28,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "missing": {
              "cases": 14,
              "passed": 14,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "correction": {
              "cases": 7,
              "passed": 7,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "ambiguous": {
              "cases": 6,
              "passed": 6,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "invalid": {
              "cases": 5,
              "passed": 4,
              "pass_rate": 0.8,
              "incorrect_ready": 0,
              "operational_errors": 0
            }
          },
          "latency_ms": {
            "median": 2007,
            "p95": 3407,
            "max": 8471
          },
          "dataset_sha256": "8d8aab81b7bde036b099b59640e3836dba52f906e1b3f23bec30afbd014eadb6",
          "acceptance_passed": true,
          "transport": {
            "attempts": 60,
            "retried_requests": 0,
            "recovered_requests": 0,
            "attempt_timeouts": 0
          }
        },
        "stopped_candidate_regression": {
          "stopped_at": "2026-10-15T16:46:37.136380+00:00",
          "reason": "Withdrew candidate after H044 assumed missing tile dimension units; add explicit unit evidence guard before qualification.",
          "completed": 312,
          "passed": 311,
          "failures": [
            {
              "id": "H044",
              "errors": [
                {
                  "kind": "status",
                  "expected": [
                    "clarify"
                  ],
                  "actual": "ready"
                },
                {
                  "kind": "unresolved_input_filled",
                  "field": "b",
                  "actual": {
                    "value": 30,
                    "unit": "cm"
                  }
                },
                {
                  "kind": "unresolved_input_filled",
                  "field": "c",
                  "actual": {
                    "value": 60,
                    "unit": "cm"
                  }
                },
                {
                  "kind": "missing_question",
                  "field": "b"
                },
                {
                  "kind": "missing_question",
                  "field": "c"
                }
              ]
            }
          ],
          "inflight_usage_may_be_unreported": true
        },
        "final_fresh_first_attempt": {
          "cases": 24,
          "passed": 24,
          "incorrect_ready": 0,
          "operational_errors": 0,
          "by_group": {
            "complete": {
              "cases": 12,
              "passed": 12,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "missing": {
              "cases": 8,
              "passed": 8,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "correction": {
              "cases": 4,
              "passed": 4,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            }
          },
          "latency_ms": {
            "median": 2276,
            "p95": 3312,
            "max": 4612
          },
          "dataset_sha256": "b604736013c2fdf2dbbac9f4870cb0de994aea02b3a5efe0c84c7cfdd78f9348",
          "acceptance_passed": true,
          "transport": {
            "attempts": 24,
            "retried_requests": 0,
            "recovered_requests": 0,
            "attempt_timeouts": 0
          }
        },
        "fresh_grading_correction": {
          "checked_at": "2026-10-15T16:43:50.176247+00:00",
          "case": "Q038",
          "original_score": {
            "passed": 59,
            "cases": 60
          },
          "corrected_score": {
            "passed": 60,
            "cases": 60
          },
          "reason": "Fixture wrongly expected zero deposits to remain in prepared inputs. Existing positive-integer contract rejects zero; host correctly retained target/savings, omitted deposits and asked for a valid count.",
          "existing_contract_field": {
            "key": "c",
            "label": "Remaining deposits",
            "type": "integer",
            "required": true,
            "unit": "deposit",
            "unit_status": "explicit",
            "minimum": 1e-9
          },
          "correct_expectation": {
            "status": [
              "clarify"
            ],
            "tool": "CC-040",
            "inputs": {
              "a": 900,
              "b": 300
            },
            "exact_inputs": true,
            "absent": [
              "c"
            ],
            "missing": [
              "c"
            ]
          },
          "runtime_changed": false,
          "model_rerun": false,
          "original_dataset_and_traces_preserved": true,
          "original_dataset_sha256": "8d8aab81b7bde036b099b59640e3836dba52f906e1b3f23bec30afbd014eadb6"
        },
        "initial_fresh_corrected_result": {
          "passed": 60,
          "cases": 60,
          "incorrect_ready": 0,
          "operational_errors": 0,
          "new_model_calls": 0,
          "basis": "Same first-attempt responses; one fixture corrected against pre-existing positive-integer constraint."
        },
        "original_four_failures": [
          {
            "id": "F034",
            "passed": true,
            "errors": []
          },
          {
            "id": "H038",
            "passed": true,
            "errors": []
          },
          {
            "id": "X060",
            "passed": true,
            "errors": []
          },
          {
            "id": "H027",
            "passed": true,
            "errors": []
          }
        ],
        "runtime_hashes": {
          "src/pilot-contracts.mjs": "c0ac3f8592e8769f744d590e6a7923a9dfbfc8e0ad2f1e89b202cbf889bcb881",
          "server/provider-request.mjs": "c993a2a57d4a6d6dac0b3d1d89acfcb1ef2b563389f61a1a56b9023818cd0f79",
          "server/ai-http.mjs": "02713a469671ac6ddba2df33636dbedb8a96c48ff971bed8a9d1743a51091c22",
          "server/number-words.mjs": "6db9cafec5a461923497c1061d96489eda7c3a5f96d8a1e465582b367fa81b92",
          "server/ai-core.mjs": "c037697f3178442345348232d29c5e09e18c6ee38a00a6c936c65e8a570b3073",
          "server/pilot-registry.json": "cefb40c7307583a1bf72e45875da25c03580c727e562070d4c66f04a348e8403",
          "src/orchestrator.mjs": "2ecf87b958e8e65a020e1eec60006f13bf7b7ada2e7930f1850ac567fc0bbbae",
          "src/engine.mjs": "46ae7dcac171c33ee97e06e124a9af73bca0877800e4a4d35e341d020b5eec59",
          "src/catalog.mjs": "18dc5fe430df9b20d20c510cfe63fd64da16940a13411a3c24e229e1ba13ef25",
          "src/math.mjs": "6e8caa526f94621f0ee7455b890e0987645976be5758b6b79f2f57df398d9a19",
          "src/expansion-math.mjs": "0674e5b6a4567e010d52cb213b0c232aa88da609a52e57a4523cee5ba74ec378"
        },
        "expansion_qualified": true,
        "expanded_in_this_release": false,
        "enabled_ai_tools": 35,
        "remaining_failures": [
          {
            "id": "E40-Y025",
            "errors": [
              {
                "kind": "value",
                "field": "a",
                "expected": 34.8765,
                "actual": null
              }
            ],
            "incorrect_ready": false
          },
          {
            "id": "N075",
            "errors": [
              {
                "kind": "status",
                "expected": [
                  "ready"
                ],
                "actual": "clarify"
              },
              {
                "kind": "value",
                "field": "c",
                "expected": 25,
                "actual": null
              }
            ],
            "incorrect_ready": false
          }
        ],
        "limitations": [
          "24 new dimension-unit questions frozen before the final guard; first live attempts, assistant-authored, not human-independent.",
          "One application request per question. Application may retry transient transport failures once within 25 seconds; evaluator never retries or replaces failed answers.",
          "Runtime, grading and answer fixtures frozen throughout each run.",
          "Routing/form preparation checked independently against answer fixtures; model never sees expected answers.",
          "Gate >=95% overall, >=98% complete, zero incorrect-ready forms and final operational errors.",
          "Not a real-user traffic estimate, independent human audit or load test.",
          "Original 59/60 fresh grade preserved; 60/60 after documented fixture correction with no model rerun",
          "No new calculator contracts or numerical engine changes",
          "Assistant implemented and evaluated fixes; not model self-repair or independent human certification",
          "Final 750-case replay revalidates saved model proposals; it is not a second live model run",
          "Private pilot; not a production load test; provider delays remain possible"
        ]
      },
      "development_history": {
        "calculator-ai-two-gate-replay-20261015": {
          "cases": 750,
          "passed": 750,
          "model_calls": 0,
          "method": "Revalidate all saved live proposals after host-only guards; not a new live model evaluation.",
          "incorrect_ready": 0,
          "operational_errors": 0,
          "new_regressions": 0,
          "improved": [
            "E40-Y025",
            "E259",
            "N075"
          ],
          "by_group": {
            "missing": {
              "cases": 119,
              "passed": 119
            },
            "complete": {
              "cases": 363,
              "passed": 363
            },
            "conversion": {
              "cases": 40,
              "passed": 40
            },
            "invalid": {
              "cases": 58,
              "passed": 58
            },
            "correction": {
              "cases": 67,
              "passed": 67
            },
            "unsupported": {
              "cases": 37,
              "passed": 37
            },
            "ambiguous": {
              "cases": 66,
              "passed": 66
            }
          },
          "failures": [],
          "runtime_hashes": {
            "src/pilot-contracts.mjs": "c0ac3f8592e8769f744d590e6a7923a9dfbfc8e0ad2f1e89b202cbf889bcb881",
            "server/provider-request.mjs": "c993a2a57d4a6d6dac0b3d1d89acfcb1ef2b563389f61a1a56b9023818cd0f79",
            "server/ai-http.mjs": "02713a469671ac6ddba2df33636dbedb8a96c48ff971bed8a9d1743a51091c22",
            "server/number-words.mjs": "6db9cafec5a461923497c1061d96489eda7c3a5f96d8a1e465582b367fa81b92",
            "server/ai-core.mjs": "155f5c20de8c8a0f26dd71811c67c2c1da312abc4513a1fdbfa85552bdbf3f28",
            "server/pilot-registry.json": "cefb40c7307583a1bf72e45875da25c03580c727e562070d4c66f04a348e8403",
            "src/orchestrator.mjs": "2ecf87b958e8e65a020e1eec60006f13bf7b7ada2e7930f1850ac567fc0bbbae",
            "src/engine.mjs": "46ae7dcac171c33ee97e06e124a9af73bca0877800e4a4d35e341d020b5eec59",
            "src/catalog.mjs": "18dc5fe430df9b20d20c510cfe63fd64da16940a13411a3c24e229e1ba13ef25",
            "src/math.mjs": "6e8caa526f94621f0ee7455b890e0987645976be5758b6b79f2f57df398d9a19",
            "src/expansion-math.mjs": "0674e5b6a4567e010d52cb213b0c232aa88da609a52e57a4523cee5ba74ec378"
          },
          "checked_at": "2026-10-15T17:04:41.737Z",
          "acceptance_passed": true
        },
        "calculator-ai-two-gate-fresh-20261015": {
          "transport": {
            "attempts": 24,
            "retried_requests": 0,
            "recovered_requests": 0,
            "attempt_timeouts": 0
          },
          "cases": 24,
          "passed": 23,
          "pass_rate": 0.9583333333333334,
          "incorrect_ready": 0,
          "operational_errors": 0,
          "model": "gpt-6-luna",
          "completed_at": "2026-10-15T17:05:05.675Z",
          "dataset_sha256": "0875423b480fa50e7570b41d74329f43260cae16da598fd47117faad38fbabfd",
          "by_group": {
            "missing": {
              "cases": 11,
              "passed": 10,
              "pass_rate": 0.9090909090909091,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "complete": {
              "cases": 4,
              "passed": 4,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "correction": {
              "cases": 9,
              "passed": 9,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            }
          },
          "latency_ms": {
            "median": 2259,
            "p95": 4979,
            "max": 9466
          },
          "usage": {
            "input_tokens": 261465,
            "output_tokens": 2913,
            "cached_input_tokens": 0,
            "responses_with_usage": 24
          },
          "estimated_standard_token_cost_usd": 0.027603,
          "cost_note": "Standard-input price estimate using $0.10/M input and $0.50/M output from prior official research. Excludes cache-write differences, cache discounts and unreported timed-out requests; not an invoice.",
          "failures": [
            {
              "id": "T002",
              "group": "missing",
              "errors": [
                {
                  "kind": "value",
                  "field": "a",
                  "expected": -16.829,
                  "actual": null
                }
              ],
              "incorrect_ready": false
            }
          ],
          "limitations": [
            "24 fresh rounding and bag correction questions frozen before the two repairs; first live attempts.",
            "One application request per question. Application may retry transient transport failures once within 25 seconds; evaluator never retries or replaces failed answers.",
            "Runtime, grading and answer fixtures frozen throughout each run.",
            "Routing/form preparation checked independently against answer fixtures; model never sees expected answers.",
            "Gate >=95% overall, >=98% complete, zero incorrect-ready forms and final operational errors.",
            "Not a real-user traffic estimate, independent human audit or load test."
          ],
          "acceptance_passed": true
        },
        "calculator-ai-two-sign-fresh-20261015": {
          "transport": {
            "attempts": 12,
            "retried_requests": 0,
            "recovered_requests": 0,
            "attempt_timeouts": 0
          },
          "cases": 12,
          "passed": 12,
          "pass_rate": 1,
          "incorrect_ready": 0,
          "operational_errors": 0,
          "model": "gpt-6-luna",
          "completed_at": "2026-10-15T17:06:24.550Z",
          "dataset_sha256": "76c9410713ef10450bc356780cfb8515106a2d1dd5be64dcb862353b194a7d81",
          "by_group": {
            "missing": {
              "cases": 6,
              "passed": 6,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "complete": {
              "cases": 6,
              "passed": 6,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            }
          },
          "latency_ms": {
            "median": 1996,
            "p95": 4731,
            "max": 4731
          },
          "usage": {
            "input_tokens": 130406,
            "output_tokens": 1225,
            "cached_input_tokens": 0,
            "responses_with_usage": 12
          },
          "estimated_standard_token_cost_usd": 0.0136531,
          "cost_note": "Standard-input price estimate using $0.10/M input and $0.50/M output from prior official research. Excludes cache-write differences, cache discounts and unreported timed-out requests; not an invoice.",
          "failures": [],
          "limitations": [
            "12 fresh signed-number questions frozen before the mixed-sign parser repair; first live attempts.",
            "One application request per question. Application may retry transient transport failures once within 25 seconds; evaluator never retries or replaces failed answers.",
            "Runtime, grading and answer fixtures frozen throughout each run.",
            "Routing/form preparation checked independently against answer fixtures; model never sees expected answers.",
            "Gate >=95% overall, >=98% complete, zero incorrect-ready forms and final operational errors.",
            "Not a real-user traffic estimate, independent human audit or load test."
          ],
          "acceptance_passed": true
        },
        "calculator-ai-expand40-fresh-20261015": {
          "transport": {
            "attempts": 40,
            "retried_requests": 0,
            "recovered_requests": 0,
            "attempt_timeouts": 0
          },
          "cases": 40,
          "passed": 37,
          "pass_rate": 0.925,
          "incorrect_ready": 0,
          "operational_errors": 0,
          "model": "gpt-6-luna",
          "completed_at": "2026-10-15T17:08:10.490Z",
          "dataset_sha256": "86b1a241c72826b7fd5680718a4dba395166de33c65e8ea00e65ae51c8f86084",
          "by_group": {
            "complete": {
              "cases": 20,
              "passed": 17,
              "pass_rate": 0.85,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "missing": {
              "cases": 5,
              "passed": 5,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "correction": {
              "cases": 5,
              "passed": 5,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "ambiguous": {
              "cases": 5,
              "passed": 5,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "invalid": {
              "cases": 5,
              "passed": 5,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            }
          },
          "latency_ms": {
            "median": 1973,
            "p95": 3912,
            "max": 4597
          },
          "usage": {
            "input_tokens": 471426,
            "output_tokens": 4375,
            "cached_input_tokens": 0,
            "responses_with_usage": 40
          },
          "estimated_standard_token_cost_usd": 0.04933010000000001,
          "cost_note": "Standard-input price estimate using $0.10/M input and $0.50/M output from prior official research. Excludes cache-write differences, cache discounts and unreported timed-out requests; not an invoice.",
          "failures": [
            {
              "id": "Z036",
              "group": "complete",
              "errors": [
                {
                  "kind": "status",
                  "expected": [
                    "ready"
                  ],
                  "actual": "clarify"
                }
              ],
              "incorrect_ready": false
            },
            {
              "id": "Z034",
              "group": "complete",
              "errors": [
                {
                  "kind": "status",
                  "expected": [
                    "ready"
                  ],
                  "actual": "clarify"
                },
                {
                  "kind": "value",
                  "field": "a",
                  "expected": -8,
                  "actual": null
                }
              ],
              "incorrect_ready": false
            },
            {
              "id": "Z018",
              "group": "complete",
              "errors": [
                {
                  "kind": "status",
                  "expected": [
                    "ready"
                  ],
                  "actual": "clarify"
                },
                {
                  "kind": "value",
                  "field": "a",
                  "expected": 1500,
                  "actual": null
                }
              ],
              "incorrect_ready": false
            }
          ],
          "limitations": [
            "40 new questions frozen before five new contracts; first live attempts, assistant-authored.",
            "One application request per question. Application may retry transient transport failures once within 25 seconds; evaluator never retries or replaces failed answers.",
            "Runtime, grading and answer fixtures frozen throughout each run.",
            "Routing/form preparation checked independently against answer fixtures; model never sees expected answers.",
            "Gate >=95% overall, >=98% complete, zero incorrect-ready forms and final operational errors.",
            "Not a real-user traffic estimate, independent human audit or load test."
          ],
          "acceptance_passed": false
        },
        "calculator-ai-expand40-regression-20261015": {
          "transport": {
            "attempts": 852,
            "retried_requests": 0,
            "recovered_requests": 0,
            "attempt_timeouts": 0
          },
          "cases": 852,
          "passed": 851,
          "pass_rate": 0.9988262910798122,
          "incorrect_ready": 0,
          "operational_errors": 0,
          "model": "gpt-6-luna",
          "completed_at": "2026-10-15T17:14:57.481Z",
          "dataset_sha256": "793ff2bd0d9796432867e6e3ee47185d15a7f2f884c0a25641b283b9ade78234",
          "by_group": {
            "complete": {
              "cases": 410,
              "passed": 409,
              "pass_rate": 0.9975609756097561,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "missing": {
              "cases": 151,
              "passed": 151,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "correction": {
              "cases": 80,
              "passed": 80,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "conversion": {
              "cases": 40,
              "passed": 40,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "ambiguous": {
              "cases": 76,
              "passed": 76,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "invalid": {
              "cases": 58,
              "passed": 58,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "unsupported": {
              "cases": 37,
              "passed": 37,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            }
          },
          "latency_ms": {
            "median": 2189,
            "p95": 4247,
            "max": 11610
          },
          "usage": {
            "input_tokens": 10067143,
            "output_tokens": 109601,
            "cached_input_tokens": 0,
            "responses_with_usage": 852
          },
          "estimated_standard_token_cost_usd": 1.0615148,
          "cost_note": "Standard-input price estimate using $0.10/M input and $0.50/M output from prior official research. Excludes cache-write differences, cache discounts and unreported timed-out requests; not an invoice.",
          "failures": [
            {
              "id": "F60-Q004",
              "group": "complete",
              "errors": [
                {
                  "kind": "status",
                  "expected": [
                    "ready"
                  ],
                  "actual": "clarify"
                }
              ],
              "incorrect_ready": false
            }
          ],
          "limitations": [
            "852 known questions spanning all earlier final cases and subsequent repair questions. These are regression cases, not fresh.",
            "One application request per question. Application may retry transient transport failures once within 25 seconds; evaluator never retries or replaces failed answers.",
            "Runtime, grading and answer fixtures frozen throughout each run.",
            "Routing/form preparation checked independently against answer fixtures; model never sees expected answers.",
            "Gate >=95% overall, >=98% complete, zero incorrect-ready forms and final operational errors.",
            "Not a real-user traffic estimate, independent human audit or load test."
          ],
          "acceptance_passed": true
        },
        "calculator-ai-expand40-final-fresh-20261015": {
          "transport": {
            "attempts": 28,
            "retried_requests": 0,
            "recovered_requests": 0,
            "attempt_timeouts": 0
          },
          "cases": 28,
          "passed": 25,
          "pass_rate": 0.8928571428571429,
          "incorrect_ready": 0,
          "operational_errors": 0,
          "model": "gpt-6-luna",
          "completed_at": "2026-10-15T17:15:26.036Z",
          "dataset_sha256": "f05c38acbab7757fe014e09e06c89ef115282497d07092cfa5e93eae3446737a",
          "by_group": {
            "complete": {
              "cases": 20,
              "passed": 18,
              "pass_rate": 0.9,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "missing": {
              "cases": 8,
              "passed": 7,
              "pass_rate": 0.875,
              "incorrect_ready": 0,
              "operational_errors": 0
            }
          },
          "latency_ms": {
            "median": 2193,
            "p95": 7257,
            "max": 8599
          },
          "usage": {
            "input_tokens": 331932,
            "output_tokens": 2973,
            "cached_input_tokens": 0,
            "responses_with_usage": 28
          },
          "estimated_standard_token_cost_usd": 0.03467970000000001,
          "cost_note": "Standard-input price estimate using $0.10/M input and $0.50/M output from prior official research. Excludes cache-write differences, cache discounts and unreported timed-out requests; not an invoice.",
          "failures": [
            {
              "id": "V019",
              "group": "complete",
              "errors": [
                {
                  "kind": "status",
                  "expected": [
                    "ready"
                  ],
                  "actual": "clarify"
                }
              ],
              "incorrect_ready": false
            },
            {
              "id": "V023",
              "group": "missing",
              "errors": [
                {
                  "kind": "value",
                  "field": "a",
                  "expected": -13,
                  "actual": null
                }
              ],
              "incorrect_ready": false
            },
            {
              "id": "V007",
              "group": "complete",
              "errors": [
                {
                  "kind": "status",
                  "expected": [
                    "ready"
                  ],
                  "actual": "clarify"
                }
              ],
              "incorrect_ready": false
            }
          ],
          "limitations": [
            "28 new questions frozen before signed-rate and colloquial-hundreds repairs; first live attempts, assistant-authored.",
            "One application request per question. Application may retry transient transport failures once within 25 seconds; evaluator never retries or replaces failed answers.",
            "Runtime, grading and answer fixtures frozen throughout each run.",
            "Routing/form preparation checked independently against answer fixtures; model never sees expected answers.",
            "Gate >=95% overall, >=98% complete, zero incorrect-ready forms and final operational errors.",
            "Not a real-user traffic estimate, independent human audit or load test."
          ],
          "acceptance_passed": false
        },
        "calculator-ai-expand38-fresh-20261015": {
          "transport": {
            "attempts": 24,
            "retried_requests": 0,
            "recovered_requests": 0,
            "attempt_timeouts": 0
          },
          "cases": 24,
          "passed": 22,
          "pass_rate": 0.9166666666666666,
          "incorrect_ready": 0,
          "operational_errors": 0,
          "model": "gpt-6-luna",
          "completed_at": "2026-10-15T17:16:56.814Z",
          "dataset_sha256": "490dada80cd2bb6bb46ff4ac768d8c9316bb3e65091803613ac4f4f5800ad2f7",
          "by_group": {
            "complete": {
              "cases": 13,
              "passed": 11,
              "pass_rate": 0.8461538461538461,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "missing": {
              "cases": 7,
              "passed": 7,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "correction": {
              "cases": 4,
              "passed": 4,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            }
          },
          "latency_ms": {
            "median": 2132,
            "p95": 3956,
            "max": 4702
          },
          "usage": {
            "input_tokens": 273645,
            "output_tokens": 2492,
            "cached_input_tokens": 0,
            "responses_with_usage": 24
          },
          "estimated_standard_token_cost_usd": 0.0286105,
          "cost_note": "Standard-input price estimate using $0.10/M input and $0.50/M output from prior official research. Excludes cache-write differences, cache discounts and unreported timed-out requests; not an invoice.",
          "failures": [
            {
              "id": "W006",
              "group": "complete",
              "errors": [
                {
                  "kind": "status",
                  "expected": [
                    "ready"
                  ],
                  "actual": "clarify"
                }
              ],
              "incorrect_ready": false
            },
            {
              "id": "W010",
              "group": "complete",
              "errors": [
                {
                  "kind": "status",
                  "expected": [
                    "ready"
                  ],
                  "actual": "clarify"
                },
                {
                  "kind": "value",
                  "field": "a",
                  "expected": 1340,
                  "actual": null
                }
              ],
              "incorrect_ready": false
            }
          ],
          "limitations": [
            "24 new questions frozen before final 38-tool candidate; first live attempts, assistant-authored.",
            "One application request per question. Application may retry transient transport failures once within 25 seconds; evaluator never retries or replaces failed answers.",
            "Runtime, grading and answer fixtures frozen throughout each run.",
            "Routing/form preparation checked independently against answer fixtures; model never sees expected answers.",
            "Gate >=95% overall, >=98% complete, zero incorrect-ready forms and final operational errors.",
            "Not a real-user traffic estimate, independent human audit or load test."
          ],
          "acceptance_passed": false
        },
        "calculator-ai-expand38-release-fresh-20261015": {
          "transport": {
            "attempts": 20,
            "retried_requests": 0,
            "recovered_requests": 0,
            "attempt_timeouts": 0
          },
          "cases": 20,
          "passed": 19,
          "pass_rate": 0.95,
          "incorrect_ready": 0,
          "operational_errors": 0,
          "model": "gpt-6-luna",
          "completed_at": "2026-10-15T17:18:08.682Z",
          "dataset_sha256": "349bfdcc7ff797b79f34ebaca544b93b8b295618638d37f4ae1cae44354f028c",
          "by_group": {
            "complete": {
              "cases": 12,
              "passed": 11,
              "pass_rate": 0.9166666666666666,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "missing": {
              "cases": 8,
              "passed": 8,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            }
          },
          "latency_ms": {
            "median": 2140,
            "p95": 3951,
            "max": 3998
          },
          "usage": {
            "input_tokens": 228335,
            "output_tokens": 1990,
            "cached_input_tokens": 0,
            "responses_with_usage": 20
          },
          "estimated_standard_token_cost_usd": 0.0238285,
          "cost_note": "Standard-input price estimate using $0.10/M input and $0.50/M output from prior official research. Excludes cache-write differences, cache discounts and unreported timed-out requests; not an invoice.",
          "failures": [
            {
              "id": "J008",
              "group": "complete",
              "errors": [
                {
                  "kind": "status",
                  "expected": [
                    "ready"
                  ],
                  "actual": "clarify"
                }
              ],
              "incorrect_ready": false
            }
          ],
          "limitations": [
            "20 fresh questions frozen before normalized evidence and validated-invalid handling; first live attempts, assistant-authored.",
            "One application request per question. Application may retry transient transport failures once within 25 seconds; evaluator never retries or replaces failed answers.",
            "Runtime, grading and answer fixtures frozen throughout each run.",
            "Routing/form preparation checked independently against answer fixtures; model never sees expected answers.",
            "Gate >=95% overall, >=98% complete, zero incorrect-ready forms and final operational errors.",
            "Not a real-user traffic estimate, independent human audit or load test."
          ],
          "acceptance_passed": false
        },
        "calculator-ai-ready-fresh-20261015": {
          "transport": {
            "attempts": 24,
            "retried_requests": 0,
            "recovered_requests": 0,
            "attempt_timeouts": 0
          },
          "cases": 24,
          "passed": 23,
          "pass_rate": 0.9583333333333334,
          "incorrect_ready": 1,
          "operational_errors": 0,
          "model": "gpt-6-luna",
          "completed_at": "2026-10-15T17:19:27.453Z",
          "dataset_sha256": "96eca8ecf40b94071a22e7e48cddb6754ad8b35ec0e57fdc61517ccfc7d12595",
          "by_group": {
            "complete": {
              "cases": 12,
              "passed": 12,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            },
            "missing": {
              "cases": 8,
              "passed": 7,
              "pass_rate": 0.875,
              "incorrect_ready": 1,
              "operational_errors": 0
            },
            "ambiguous": {
              "cases": 4,
              "passed": 4,
              "pass_rate": 1,
              "incorrect_ready": 0,
              "operational_errors": 0
            }
          },
          "latency_ms": {
            "median": 2026,
            "p95": 4094,
            "max": 8533
          },
          "usage": {
            "input_tokens": 274177,
            "output_tokens": 2389,
            "cached_input_tokens": 0,
            "responses_with_usage": 24
          },
          "estimated_standard_token_cost_usd": 0.0286122,
          "cost_note": "Standard-input price estimate using $0.10/M input and $0.50/M output from prior official research. Excludes cache-write differences, cache discounts and unreported timed-out requests; not an invoice.",
          "failures": [
            {
              "id": "K019",
              "group": "missing",
              "errors": [
                {
                  "kind": "status",
                  "expected": [
                    "clarify"
                  ],
                  "actual": "ready"
                },
                {
                  "kind": "extra_input",
                  "field": "a"
                },
                {
                  "kind": "unresolved_input_filled",
                  "field": "a",
                  "actual": {
                    "value": 1700,
                    "unit": "GBP"
                  }
                },
                {
                  "kind": "missing_question",
                  "field": "a"
                }
              ],
              "incorrect_ready": true
            }
          ],
          "limitations": [
            "24 new questions frozen before host-derived readiness; first live attempts, assistant-authored.",
            "One application request per question. Application may retry transient transport failures once within 25 seconds; evaluator never retries or replaces failed answers.",
            "Runtime, grading and answer fixtures frozen throughout each run.",
            "Routing/form preparation checked independently against answer fixtures; model never sees expected answers.",
            "Gate >=95% overall, >=98% complete, zero incorrect-ready forms and final operational errors.",
            "Not a real-user traffic estimate, independent human audit or load test."
          ],
          "acceptance_passed": false
        }
      },
      "live_regression": {
        "transport": {
          "attempts": 959,
          "retried_requests": 1,
          "recovered_requests": 1,
          "attempt_timeouts": 1
        },
        "cases": 958,
        "passed": 956,
        "pass_rate": 0.9979123173277662,
        "incorrect_ready": 1,
        "operational_errors": 0,
        "model": "gpt-6-luna",
        "completed_at": "2026-10-15T17:28:43.623Z",
        "dataset_sha256": "1ca94d4c33b3ce524a1aec8a0105f67f0e6db60cf88825f0c2d61ca97af47c09",
        "by_group": {
          "complete": {
            "cases": 469,
            "passed": 467,
            "pass_rate": 0.9957356076759062,
            "incorrect_ready": 1,
            "operational_errors": 0
          },
          "missing": {
            "cases": 181,
            "passed": 181,
            "pass_rate": 1,
            "incorrect_ready": 0,
            "operational_errors": 0
          },
          "correction": {
            "cases": 87,
            "passed": 87,
            "pass_rate": 1,
            "incorrect_ready": 0,
            "operational_errors": 0
          },
          "conversion": {
            "cases": 40,
            "passed": 40,
            "pass_rate": 1,
            "incorrect_ready": 0,
            "operational_errors": 0
          },
          "ambiguous": {
            "cases": 83,
            "passed": 83,
            "pass_rate": 1,
            "incorrect_ready": 0,
            "operational_errors": 0
          },
          "invalid": {
            "cases": 61,
            "passed": 61,
            "pass_rate": 1,
            "incorrect_ready": 0,
            "operational_errors": 0
          },
          "unsupported": {
            "cases": 37,
            "passed": 37,
            "pass_rate": 1,
            "incorrect_ready": 0,
            "operational_errors": 0
          }
        },
        "latency_ms": {
          "median": 2091,
          "p95": 4288,
          "max": 14096
        },
        "usage": {
          "input_tokens": 10973392,
          "output_tokens": 120490,
          "cached_input_tokens": 519233,
          "responses_with_usage": 958
        },
        "estimated_standard_token_cost_usd": 1.1575842,
        "cost_note": "Standard-input price estimate using $0.10/M input and $0.50/M output from prior official research. Excludes cache-write differences, cache discounts and unreported timed-out requests; not an invoice.",
        "failures": [
          {
            "id": "H010",
            "group": "complete",
            "errors": [
              {
                "kind": "tool",
                "expected": "CC-033",
                "actual": "CC-032"
              }
            ],
            "incorrect_ready": true
          },
          {
            "id": "X006",
            "group": "complete",
            "errors": [
              {
                "kind": "status",
                "expected": [
                  "ready"
                ],
                "actual": "clarify"
              },
              {
                "kind": "value",
                "field": "a",
                "expected": 200,
                "actual": null
              }
            ],
            "incorrect_ready": false
          }
        ],
        "limitations": [
          "958 known questions spanning all earlier final cases and subsequent repair questions. These are regression cases, not fresh.",
          "One application request per question. Application may retry transient transport failures once within 25 seconds; evaluator never retries or replaces failed answers.",
          "Runtime, grading and answer fixtures frozen throughout each run.",
          "Routing/form preparation checked independently against answer fixtures; model never sees expected answers.",
          "Gate >=95% overall, >=98% complete, zero incorrect-ready forms and final operational errors.",
          "Not a real-user traffic estimate, independent human audit or load test."
        ],
        "acceptance_passed": false
      },
      "fresh_qualification": {
        "transport": {
          "attempts": 100,
          "retried_requests": 0,
          "recovered_requests": 0,
          "attempt_timeouts": 0
        },
        "cases": 100,
        "passed": 100,
        "pass_rate": 1,
        "incorrect_ready": 0,
        "operational_errors": 0,
        "model": "gpt-6-luna",
        "completed_at": "2026-10-15T17:21:52.644Z",
        "dataset_sha256": "58b597df35ec69e908d1627b97190c38e90dd0f69000168f2f51c8576ba023c5",
        "by_group": {
          "complete": {
            "cases": 50,
            "passed": 50,
            "pass_rate": 1,
            "incorrect_ready": 0,
            "operational_errors": 0
          },
          "missing": {
            "cases": 20,
            "passed": 20,
            "pass_rate": 1,
            "incorrect_ready": 0,
            "operational_errors": 0
          },
          "correction": {
            "cases": 10,
            "passed": 10,
            "pass_rate": 1,
            "incorrect_ready": 0,
            "operational_errors": 0
          },
          "ambiguous": {
            "cases": 10,
            "passed": 10,
            "pass_rate": 1,
            "incorrect_ready": 0,
            "operational_errors": 0
          },
          "invalid": {
            "cases": 10,
            "passed": 10,
            "pass_rate": 1,
            "incorrect_ready": 0,
            "operational_errors": 0
          }
        },
        "latency_ms": {
          "median": 2044,
          "p95": 3508,
          "max": 9942
        },
        "usage": {
          "input_tokens": 1141917,
          "output_tokens": 10412,
          "cached_input_tokens": 0,
          "responses_with_usage": 100
        },
        "estimated_standard_token_cost_usd": 0.11939770000000001,
        "cost_note": "Standard-input price estimate using $0.10/M input and $0.50/M output from prior official research. Excludes cache-write differences, cache discounts and unreported timed-out requests; not an invoice.",
        "failures": [],
        "limitations": [
          "100 newly frozen qualification questions, shared families with changed values/wording; first live attempts, assistant-authored.",
          "One application request per question. Application may retry transient transport failures once within 25 seconds; evaluator never retries or replaces failed answers.",
          "Runtime, grading and answer fixtures frozen throughout each run.",
          "Routing/form preparation checked independently against answer fixtures; model never sees expected answers.",
          "Gate >=95% overall, >=98% complete, zero incorrect-ready forms and final operational errors.",
          "Not a real-user traffic estimate, independent human audit or load test."
        ],
        "acceptance_passed": true
      },
      "source_hashes": {
        "src/pilot-contracts.mjs": "21037c2e59e7df3952c4c8acc8f17b176256a7814abab8926d7143a204720962",
        "server/provider-request.mjs": "c993a2a57d4a6d6dac0b3d1d89acfcb1ef2b563389f61a1a56b9023818cd0f79",
        "server/ai-http.mjs": "02713a469671ac6ddba2df33636dbedb8a96c48ff971bed8a9d1743a51091c22",
        "server/number-words.mjs": "b3bd8bed61f1e27d133265cb9dd56c9d8dfa96f95428594d234efe540d627599",
        "server/ai-core.mjs": "e75a28aa734b4e99c978132d1fe1e48d7121bd7eab2f057c9fe70aeecbf561b0",
        "server/pilot-registry.json": "d989fc88377321cc83d51a7223f71ea66919aeddc3b7cc05f590d08b7eb517b0",
        "src/orchestrator.mjs": "2ecf87b958e8e65a020e1eec60006f13bf7b7ada2e7930f1850ac567fc0bbbae",
        "src/engine.mjs": "46ae7dcac171c33ee97e06e124a9af73bca0877800e4a4d35e341d020b5eec59",
        "src/catalog.mjs": "18dc5fe430df9b20d20c510cfe63fd64da16940a13411a3c24e229e1ba13ef25",
        "src/math.mjs": "6e8caa526f94621f0ee7455b890e0987645976be5758b6b79f2f57df398d9a19",
        "src/expansion-math.mjs": "0674e5b6a4567e010d52cb213b0c232aa88da609a52e57a4523cee5ba74ec378"
      },
      "final_saved_response_replay": {
        "cases": 958,
        "passed": 956,
        "model_calls": 0,
        "method": "Revalidate all saved live proposals after host-only guards; not a new live model evaluation.",
        "incorrect_ready": 0,
        "operational_errors": 0,
        "new_regressions": 0,
        "improved": [],
        "by_group": {
          "complete": {
            "cases": 469,
            "passed": 467
          },
          "missing": {
            "cases": 181,
            "passed": 181
          },
          "ambiguous": {
            "cases": 83,
            "passed": 83
          },
          "invalid": {
            "cases": 61,
            "passed": 61
          },
          "correction": {
            "cases": 87,
            "passed": 87
          },
          "conversion": {
            "cases": 40,
            "passed": 40
          },
          "unsupported": {
            "cases": 37,
            "passed": 37
          }
        },
        "failures": [
          {
            "id": "H010",
            "errors": [
              {
                "kind": "status",
                "expected": [
                  "ready"
                ],
                "actual": "clarify"
              },
              {
                "kind": "tool",
                "expected": "CC-033",
                "actual": null
              },
              {
                "kind": "value",
                "field": "a",
                "expected": 2650,
                "actual": null
              },
              {
                "kind": "value",
                "field": "b",
                "expected": 2.2,
                "actual": null
              },
              {
                "kind": "value",
                "field": "c",
                "expected": 11,
                "actual": null
              }
            ],
            "incorrect_ready": false
          },
          {
            "id": "X006",
            "errors": [
              {
                "kind": "status",
                "expected": [
                  "ready"
                ],
                "actual": "clarify"
              },
              {
                "kind": "value",
                "field": "a",
                "expected": 200,
                "actual": null
              }
            ],
            "incorrect_ready": false
          }
        ],
        "runtime_hashes": {
          "src/pilot-contracts.mjs": "21037c2e59e7df3952c4c8acc8f17b176256a7814abab8926d7143a204720962",
          "server/provider-request.mjs": "c993a2a57d4a6d6dac0b3d1d89acfcb1ef2b563389f61a1a56b9023818cd0f79",
          "server/ai-http.mjs": "02713a469671ac6ddba2df33636dbedb8a96c48ff971bed8a9d1743a51091c22",
          "server/number-words.mjs": "b3bd8bed61f1e27d133265cb9dd56c9d8dfa96f95428594d234efe540d627599",
          "server/ai-core.mjs": "e75a28aa734b4e99c978132d1fe1e48d7121bd7eab2f057c9fe70aeecbf561b0",
          "server/pilot-registry.json": "d989fc88377321cc83d51a7223f71ea66919aeddc3b7cc05f590d08b7eb517b0",
          "src/orchestrator.mjs": "2ecf87b958e8e65a020e1eec60006f13bf7b7ada2e7930f1850ac567fc0bbbae",
          "src/engine.mjs": "46ae7dcac171c33ee97e06e124a9af73bca0877800e4a4d35e341d020b5eec59",
          "src/catalog.mjs": "18dc5fe430df9b20d20c510cfe63fd64da16940a13411a3c24e229e1ba13ef25",
          "src/math.mjs": "6e8caa526f94621f0ee7455b890e0987645976be5758b6b79f2f57df398d9a19",
          "src/expansion-math.mjs": "0674e5b6a4567e010d52cb213b0c232aa88da609a52e57a4523cee5ba74ec378"
        },
        "checked_at": "2026-10-15T17:28:50.530Z",
        "acceptance_passed": true
      },
      "qualification_saved_response_replay": {
        "cases": 100,
        "passed": 100,
        "model_calls": 0,
        "method": "Revalidate all saved live proposals after host-only guards; not a new live model evaluation.",
        "incorrect_ready": 0,
        "operational_errors": 0,
        "new_regressions": 0,
        "improved": [],
        "by_group": {
          "complete": {
            "cases": 50,
            "passed": 50
          },
          "missing": {
            "cases": 20,
            "passed": 20
          },
          "invalid": {
            "cases": 10,
            "passed": 10
          },
          "correction": {
            "cases": 10,
            "passed": 10
          },
          "ambiguous": {
            "cases": 10,
            "passed": 10
          }
        },
        "failures": [],
        "runtime_hashes": {
          "src/pilot-contracts.mjs": "21037c2e59e7df3952c4c8acc8f17b176256a7814abab8926d7143a204720962",
          "server/provider-request.mjs": "c993a2a57d4a6d6dac0b3d1d89acfcb1ef2b563389f61a1a56b9023818cd0f79",
          "server/ai-http.mjs": "02713a469671ac6ddba2df33636dbedb8a96c48ff971bed8a9d1743a51091c22",
          "server/number-words.mjs": "b3bd8bed61f1e27d133265cb9dd56c9d8dfa96f95428594d234efe540d627599",
          "server/ai-core.mjs": "e75a28aa734b4e99c978132d1fe1e48d7121bd7eab2f057c9fe70aeecbf561b0",
          "server/pilot-registry.json": "d989fc88377321cc83d51a7223f71ea66919aeddc3b7cc05f590d08b7eb517b0",
          "src/orchestrator.mjs": "2ecf87b958e8e65a020e1eec60006f13bf7b7ada2e7930f1850ac567fc0bbbae",
          "src/engine.mjs": "46ae7dcac171c33ee97e06e124a9af73bca0877800e4a4d35e341d020b5eec59",
          "src/catalog.mjs": "18dc5fe430df9b20d20c510cfe63fd64da16940a13411a3c24e229e1ba13ef25",
          "src/math.mjs": "6e8caa526f94621f0ee7455b890e0987645976be5758b6b79f2f57df398d9a19",
          "src/expansion-math.mjs": "0674e5b6a4567e010d52cb213b0c232aa88da609a52e57a4523cee5ba74ec378"
        },
        "checked_at": "2026-10-15T17:28:50.579Z",
        "acceptance_passed": true
      },
      "final_fresh_guard_check": {
        "transport": {
          "attempts": 20,
          "retried_requests": 0,
          "recovered_requests": 0,
          "attempt_timeouts": 0
        },
        "cases": 20,
        "passed": 20,
        "pass_rate": 1,
        "incorrect_ready": 0,
        "operational_errors": 0,
        "model": "gpt-6-luna",
        "completed_at": "2026-10-15T17:29:01.770Z",
        "dataset_sha256": "c9f19bc93d1aaf2a5ae7abc7657cc6f4aa7de082e4713d4b98c6ead38c315908",
        "by_group": {
          "complete": {
            "cases": 15,
            "passed": 15,
            "pass_rate": 1,
            "incorrect_ready": 0,
            "operational_errors": 0
          },
          "missing": {
            "cases": 5,
            "passed": 5,
            "pass_rate": 1,
            "incorrect_ready": 0,
            "operational_errors": 0
          }
        },
        "latency_ms": {
          "median": 2051,
          "p95": 3699,
          "max": 3805
        },
        "usage": {
          "input_tokens": 228649,
          "output_tokens": 2530,
          "cached_input_tokens": 0,
          "responses_with_usage": 20
        },
        "estimated_standard_token_cost_usd": 0.024129900000000003,
        "cost_note": "Standard-input price estimate using $0.10/M input and $0.50/M output from prior official research. Excludes cache-write differences, cache discounts and unreported timed-out requests; not an invoice.",
        "failures": [],
        "limitations": [
          "20 fresh questions frozen before future-growth/present-value mismatch guard; first live attempts, assistant-authored.",
          "One application request per question. Application may retry transient transport failures once within 25 seconds; evaluator never retries or replaces failed answers.",
          "Runtime, grading and answer fixtures frozen throughout each run.",
          "Routing/form preparation checked independently against answer fixtures; model never sees expected answers.",
          "Gate >=95% overall, >=98% complete, zero incorrect-ready forms and final operational errors.",
          "Not a real-user traffic estimate, independent human audit or load test."
        ],
        "acceptance_passed": true
      },
      "worker_checks": {
        "cases": 8,
        "passed": 8,
        "original_issues": [
          {
            "id": "N075",
            "passed": true,
            "http_status": 200
          },
          {
            "id": "E40-Y025",
            "passed": true,
            "http_status": 200
          }
        ]
      },
      "acceptance_passed": true,
      "limitations": [
        "Developer-authored fixtures; not independent human certification or observed traffic accuracy.",
        "Fresh questions include shared families with changed values and wording. Prior candidate failures are preserved and count as development, not erased by this run.",
        "Model clarity remains fallible; users review inputs and deliberately press Calculate.",
        "All numerical formulas unchanged; no general chaining, current rates or financial advice.",
        "Semantic clarify remains clarify even with fields supplied. Only a numerical invalid label can be superseded after all host and engine checks pass.",
        "Strata remains stopped; no Qwen or GPU inference."
      ]
    },
    "live_regression": {
      "transport": {
        "attempts": 1078,
        "retried_requests": 0,
        "recovered_requests": 0,
        "attempt_timeouts": 0
      },
      "cases": 1078,
      "passed": 1076,
      "pass_rate": 0.9981447124304267,
      "incorrect_ready": 0,
      "operational_errors": 0,
      "model": "gpt-6-luna",
      "completed_at": "2026-10-15T18:34:07.366Z",
      "dataset_sha256": "7ffa7acb7715c125ae5c96783a362fd9a60856b084cb14f2b740abf594bd7a98",
      "by_group": {
        "complete": {
          "cases": 534,
          "passed": 532,
          "pass_rate": 0.9962546816479401,
          "incorrect_ready": 0,
          "operational_errors": 0
        },
        "missing": {
          "cases": 206,
          "passed": 206,
          "pass_rate": 1,
          "incorrect_ready": 0,
          "operational_errors": 0
        },
        "correction": {
          "cases": 97,
          "passed": 97,
          "pass_rate": 1,
          "incorrect_ready": 0,
          "operational_errors": 0
        },
        "conversion": {
          "cases": 40,
          "passed": 40,
          "pass_rate": 1,
          "incorrect_ready": 0,
          "operational_errors": 0
        },
        "ambiguous": {
          "cases": 93,
          "passed": 93,
          "pass_rate": 1,
          "incorrect_ready": 0,
          "operational_errors": 0
        },
        "invalid": {
          "cases": 71,
          "passed": 71,
          "pass_rate": 1,
          "incorrect_ready": 0,
          "operational_errors": 0
        },
        "unsupported": {
          "cases": 37,
          "passed": 37,
          "pass_rate": 1,
          "incorrect_ready": 0,
          "operational_errors": 0
        }
      },
      "latency_ms": {
        "median": 2125,
        "p95": 4215,
        "max": 11336
      },
      "usage": {
        "input_tokens": 14579730,
        "output_tokens": 133137,
        "cached_input_tokens": 0,
        "responses_with_usage": 1078
      },
      "estimated_standard_token_cost_usd": 1.5245415,
      "cost_note": "Standard-input price estimate using $0.10/M input and $0.50/M output from prior official research. Excludes cache-write differences, cache discounts and unreported timed-out requests; not an invoice.",
      "failures": [
        {
          "id": "X006",
          "group": "complete",
          "errors": [
            {
              "kind": "status",
              "expected": [
                "ready"
              ],
              "actual": "clarify"
            },
            {
              "kind": "value",
              "field": "a",
              "expected": 200,
              "actual": null
            }
          ],
          "incorrect_ready": false
        },
        {
          "id": "F60-Q004",
          "group": "complete",
          "errors": [
            {
              "kind": "status",
              "expected": [
                "ready"
              ],
              "actual": "clarify"
            }
          ],
          "incorrect_ready": false
        }
      ],
      "limitations": [
        "1078 known questions spanning all earlier final cases and subsequent repair questions. These are regression cases, not fresh.",
        "One application request per question. Application may retry transient transport failures once within 25 seconds; evaluator never retries or replaces failed answers.",
        "Runtime, grading and answer fixtures frozen throughout each run.",
        "Routing/form preparation checked independently against answer fixtures; model never sees expected answers.",
        "Gate >=95% overall, >=98% complete, zero incorrect-ready forms and final operational errors.",
        "Not a real-user traffic estimate, independent human audit or load test."
      ],
      "acceptance_passed": true
    },
    "fresh_qualification": {
      "transport": {
        "attempts": 100,
        "retried_requests": 0,
        "recovered_requests": 0,
        "attempt_timeouts": 0
      },
      "cases": 100,
      "passed": 100,
      "pass_rate": 1,
      "incorrect_ready": 0,
      "operational_errors": 0,
      "model": "gpt-6-luna",
      "completed_at": "2026-10-15T18:25:06.200Z",
      "dataset_sha256": "36722620603a60fc0ac06d5b12e5f1f732a423ee82a1014ddd18fd422e889e08",
      "by_group": {
        "complete": {
          "cases": 58,
          "passed": 58,
          "pass_rate": 1,
          "incorrect_ready": 0,
          "operational_errors": 0
        },
        "missing": {
          "cases": 12,
          "passed": 12,
          "pass_rate": 1,
          "incorrect_ready": 0,
          "operational_errors": 0
        },
        "correction": {
          "cases": 12,
          "passed": 12,
          "pass_rate": 1,
          "incorrect_ready": 0,
          "operational_errors": 0
        },
        "ambiguous": {
          "cases": 7,
          "passed": 7,
          "pass_rate": 1,
          "incorrect_ready": 0,
          "operational_errors": 0
        },
        "invalid": {
          "cases": 11,
          "passed": 11,
          "pass_rate": 1,
          "incorrect_ready": 0,
          "operational_errors": 0
        }
      },
      "latency_ms": {
        "median": 1935,
        "p95": 4242,
        "max": 8954
      },
      "usage": {
        "input_tokens": 1348498,
        "output_tokens": 10463,
        "cached_input_tokens": 0,
        "responses_with_usage": 100
      },
      "estimated_standard_token_cost_usd": 0.14008130000000002,
      "cost_note": "Standard-input price estimate using $0.10/M input and $0.50/M output from prior official research. Excludes cache-write differences, cache discounts and unreported timed-out requests; not an invoice.",
      "failures": [],
      "limitations": [
        "100 newly frozen questions for 12 new contracts; first live attempts, assistant-authored.",
        "One application request per question. Application may retry transient transport failures once within 25 seconds; evaluator never retries or replaces failed answers.",
        "Runtime, grading and answer fixtures frozen throughout each run.",
        "Routing/form preparation checked independently against answer fixtures; model never sees expected answers.",
        "Gate >=95% overall, >=98% complete, zero incorrect-ready forms and final operational errors.",
        "Not a real-user traffic estimate, independent human audit or load test."
      ],
      "acceptance_passed": true
    },
    "source_hashes": {
      "src/pilot-contracts.mjs": "d56042e219f9a9aba0b6212815677cb9224f6c2fff217deb34860795aa752ac8",
      "server/provider-request.mjs": "c993a2a57d4a6d6dac0b3d1d89acfcb1ef2b563389f61a1a56b9023818cd0f79",
      "server/ai-http.mjs": "02713a469671ac6ddba2df33636dbedb8a96c48ff971bed8a9d1743a51091c22",
      "server/number-words.mjs": "b3bd8bed61f1e27d133265cb9dd56c9d8dfa96f95428594d234efe540d627599",
      "server/ai-core.mjs": "e75a28aa734b4e99c978132d1fe1e48d7121bd7eab2f057c9fe70aeecbf561b0",
      "server/pilot-registry.json": "db128109c41fb0ecae95d962df6d2bb44d528ae7c3448cd8ead2d1a4686d0a64",
      "src/orchestrator.mjs": "2ecf87b958e8e65a020e1eec60006f13bf7b7ada2e7930f1850ac567fc0bbbae",
      "src/engine.mjs": "46ae7dcac171c33ee97e06e124a9af73bca0877800e4a4d35e341d020b5eec59",
      "src/catalog.mjs": "c74a28c715f39a72013d5ab7d990d2e13ae19834e4e60cf35c9d532d1eba8d8b",
      "src/math.mjs": "6e8caa526f94621f0ee7455b890e0987645976be5758b6b79f2f57df398d9a19",
      "src/expansion-math.mjs": "0674e5b6a4567e010d52cb213b0c232aa88da609a52e57a4523cee5ba74ec378"
    },
    "evaluation_source_hashes": {
      "src/pilot-contracts.mjs": "d56042e219f9a9aba0b6212815677cb9224f6c2fff217deb34860795aa752ac8",
      "server/provider-request.mjs": "c993a2a57d4a6d6dac0b3d1d89acfcb1ef2b563389f61a1a56b9023818cd0f79",
      "server/ai-http.mjs": "02713a469671ac6ddba2df33636dbedb8a96c48ff971bed8a9d1743a51091c22",
      "server/number-words.mjs": "b3bd8bed61f1e27d133265cb9dd56c9d8dfa96f95428594d234efe540d627599",
      "server/ai-core.mjs": "e75a28aa734b4e99c978132d1fe1e48d7121bd7eab2f057c9fe70aeecbf561b0",
      "server/pilot-registry.json": "1a9de61a224fc0678433a74b3673e4c44a2c31d52b7420b95b050ab3d6102ad8",
      "src/orchestrator.mjs": "2ecf87b958e8e65a020e1eec60006f13bf7b7ada2e7930f1850ac567fc0bbbae",
      "src/engine.mjs": "46ae7dcac171c33ee97e06e124a9af73bca0877800e4a4d35e341d020b5eec59",
      "src/catalog.mjs": "18dc5fe430df9b20d20c510cfe63fd64da16940a13411a3c24e229e1ba13ef25",
      "src/math.mjs": "6e8caa526f94621f0ee7455b890e0987645976be5758b6b79f2f57df398d9a19",
      "src/expansion-math.mjs": "0674e5b6a4567e010d52cb213b0c232aa88da609a52e57a4523cee5ba74ec378"
    },
    "held_only_changes_after_evaluation": {
      "files": [
        "server/pilot-registry.json",
        "src/catalog.mjs"
      ],
      "tools": [
        "CC-485",
        "CC-488",
        "CC-499"
      ],
      "ai_surface_identical": true,
      "ai_surface_sha256": "ea239ea6ff9d1ccd97e849096381b1bea8aa4ce1671c8ffd33518a0391b0bde3",
      "meaning": "Model instructions, request schema, all 50 registry entries and enabled calculator functions/fields/notes are identical to the live-evaluated candidate. Only held chemistry guards/metadata and the aggregate registry hash changed. Those repairs have separate deterministic tests."
    },
    "worker_checks": {
      "cases": 12,
      "passed": 12
    },
    "chemistry_boundary_checks": {
      "passed": true,
      "checks": 16,
      "repaired_tools": [
        485,
        488,
        499
      ],
      "holds_removed": 0,
      "model_calls": 0,
      "scope": "Boundary repairs and independently worked examples; not specialist signoff.",
      "checked_at": "2026-10-15T18:34:16.003Z"
    },
    "acceptance_passed": true,
    "limitations": [
      "Questions and grades are developer-authored, not independent human certification.",
      "Shared question families exercise changed values, wording, missing fields, corrections and invalid cases; not a real-user traffic accuracy estimate.",
      "Two or more safe clarifications may remain; inspect the exact failures below.",
      "No general chaining, live rates, automatic deployment loop or autonomous background installer.",
      "All 100 specialist holds remain, including 25 chemistry calculators.",
      "No Qwen, Strata, GPU inference or fine-tuning."
    ]
  },
  "ai_live": {
    "model": "gpt-6-luna",
    "checked_at": "2026-10-15T13:58:29.130Z",
    "cases": [
      {
        "name": "Discount",
        "passed": true,
        "latency_ms": 2274
      },
      {
        "name": "Missing return assumption",
        "passed": true,
        "latency_ms": 3341
      },
      {
        "name": "Follow-up rate",
        "passed": true,
        "latency_ms": 2345
      },
      {
        "name": "Flooring packs",
        "passed": true,
        "latency_ms": 1839
      },
      {
        "name": "Area conversion",
        "passed": true,
        "latency_ms": 2160
      },
      {
        "name": "Missing area not guessed",
        "passed": true,
        "latency_ms": 2039
      },
      {
        "name": "Currency mismatch",
        "passed": true,
        "latency_ms": 1454
      },
      {
        "name": "Loan explicit zero",
        "passed": true,
        "latency_ms": 8751
      },
      {
        "name": "Exam target",
        "passed": true,
        "latency_ms": 2108
      },
      {
        "name": "Subscriptions list",
        "passed": true,
        "latency_ms": 1788
      },
      {
        "name": "Follow-up correction",
        "passed": true,
        "latency_ms": 2119
      },
      {
        "name": "Unsupported tax",
        "passed": true,
        "latency_ms": 1317
      },
      {
        "name": "Unsupported live data",
        "passed": true,
        "latency_ms": 1646
      },
      {
        "name": "Prompt injection cannot open held tool",
        "passed": true,
        "latency_ms": 1303
      },
      {
        "name": "Invalid discount",
        "passed": true,
        "latency_ms": 1821
      }
    ],
    "usage": {
      "input_tokens": 83105,
      "output_tokens": 1537,
      "cached_input_tokens": 75334
    },
    "passed": true,
    "estimated_standard_token_cost_usd": 0.009079,
    "limitations": [
      "Small smoke suite, not the proposed 300-question benchmark",
      "Timing includes model API round trip; one run per case",
      "Token estimate excludes cache write differences and actual billing adjustments"
    ]
  },
  "ai_browser": {
    "passed": true,
    "mocked_provider": true,
    "live_model_evidence": "See ai-evaluation-report.json; batch50 regression and fresh qualification runs",
    "viewports": [
      320,
      390,
      1440
    ],
    "checks": [
      "No AI calls on page load",
      "Prepared inputs need review before calculation",
      "Follow-up messages preserve context",
      "New conversation clears context",
      "Provider errors leave ordinary tools usable",
      "Local matching sends no API requests",
      "Clarification retains known values and leaves missing fields empty",
      "Invalid inputs display specific limits",
      "Mobile and desktop layouts",
      "No browser errors"
    ],
    "checked_at": "2026-10-15T18:25:35.853Z"
  },
  "registry": {
    "passed": true,
    "metadata_records": 3000,
    "routing_cases": 135,
    "changed_input_tools": 25,
    "independent_output_checks": 46,
    "rejected_invalid_requests": 3110,
    "model_used": false,
    "limits": [
      "Bounded phrase matching, not general natural-language understanding",
      "No automatic parameter extraction, conversion or chaining",
      "Only 50 executable orchestration contracts"
    ],
    "checked_at": "2026-10-18T12:57:51.815Z"
  },
  "registry_browser": {
    "passed": true,
    "tools_exercised": 50,
    "reference_output_checks": 78,
    "changed_input_flooring_case": true,
    "viewports": [
      320,
      390,
      1440
    ],
    "checks": [
      "Blank input enforcement",
      "Explicit example loading",
      "Stale results cleared on edits",
      "Invalid inputs rejected",
      "Ambiguous queries require selection",
      "Unsupported queries abstain",
      "Keyboard selection and focus",
      "Registry load failure handled",
      "No external or model requests"
    ],
    "checked_at": "2026-10-15T18:25:43.958Z"
  },
  "visual_models": {
    "passed": true,
    "assertions": 1080,
    "pilotTools": [
      26,
      51,
      79,
      151,
      152,
      6,
      8,
      27,
      30,
      33,
      35,
      37,
      39,
      44,
      61,
      153,
      157,
      158,
      175,
      183
    ],
    "checks": [
      "Engine result agreement",
      "Independent savings recurrence",
      "Independent amortisation recurrence",
      "Whole-unit break-even boundary",
      "Area and material conservation",
      "Zero inputs and rates",
      "Invalid input rejection",
      "Bounded maximum horizon"
    ],
    "limitations": [
      "Charts inherit calculator assumptions",
      "No expert financial or materials certification",
      "Material visuals are quantity diagrams, not cutting plans"
    ]
  },
  "visual_browser": {
    "passed": true,
    "visualRoutes": [
      "discount-and-sale-price",
      "tip-and-bill-split",
      "compound-interest",
      "simple-interest",
      "inflation-adjusted-money",
      "future-value-of-a-lump-sum",
      "savings-contribution-growth",
      "emergency-fund-target",
      "subscription-spending",
      "investment-fee-impact",
      "loan-repayment",
      "mortgage-deposit",
      "break-even-units",
      "wall-paint-quantity",
      "flooring-packs",
      "tile-quantity",
      "tile-adhesive-quantity",
      "plaster-quantity",
      "renovation-contingency-budget",
      "turf-rolls"
    ],
    "mobileWidth": 390,
    "desktopWidth": 1440,
    "checks": [
      "Live inputs",
      "Saved comparison retains its inputs",
      "Keyboard and pointer inspection",
      "Mobile label size",
      "Table and chart CSV",
      "Comparison toggle",
      "Invalid input clears chart",
      "Reset clears comparison",
      "Zero-value chart",
      "Resize",
      "Reduced motion",
      "Coin-flip regression"
    ],
    "pageErrors": []
  },
  "development_batch3": {
    "passed": true,
    "checked_at": "2026-10-18T12:57:51.443Z",
    "new_tools": 490,
    "scenarios": 980,
    "independent_output_checks": 2968,
    "invalid_checks": 980,
    "property_checks": 400,
    "errors": [],
    "limitations": [
      "Developer and research-agent worked reference arithmetic, not external specialist certification.",
      "Candidate scopes are explicitly bounded on each page; no live rates, regulations, measurements or provider data are implied.",
      "No expansion of the 50 AI finder contracts."
    ]
  },
  "development_batch3_browser": {
    "passed": true,
    "checked_at": "2026-10-15T20:20:23.456Z",
    "new_tools": 490,
    "viewports": [
      320,
      390,
      1440
    ],
    "responsive_route_checks": 1470,
    "changed_input_output_checks": 4452,
    "invalid_stale_result_checks": 490,
    "keyboard_submit": true,
    "errors": [],
    "recheck": {
      "passed": true,
      "ids": [
        1679,
        1926
      ],
      "reason": "Browser numeric controls reject alphabetic fill before submission. Used the existing negative-value invalid fixtures; numerical invalid-string tests retained.",
      "checked_at": "2026-10-15T20:21:20.239Z"
    }
  }
}