{
  "id": "advanced-v01-audited-selection-2026-10-10",
  "kind": "calibration",
  "version": "advanced-0.1-utkast",
  "runDate": "2026-10-10",
  "publishedAt": "2026-10-10",
  "tasks": 12,
  "fields": 292,
  "repeats": 1,
  "responses": 96,
  "missing": 0,
  "technicalErrors": 0,
  "protocolRejected": 0,
  "toolCalls": 0,
  "csv": "/barometer/calibration-advanced-03-2026-10-10.csv",
  "json": "/barometer/calibration-advanced-03-2026-10-10.json",
  "fullJson": "/barometer/calibration-advanced-03-2026-10-10-full.json",
  "source": {
    "snapshotSha256": "5132dfd81cbb2f103c50cd813c65c2efbd3c569890364be804000b6ceb0db080",
    "scorerSha256": "4f26f9806b58df13955d97e0129cb7457654d6322e779ce3fceb57f6773c5b49",
    "geminiAuditSha256": "ca60602200a72b8c934c0f939e34f5cb76f83a104f575e390d8f935cb841c20a"
  },
  "selection": {
    "runs": [
      "subagents-advanced-v01-2026-10-10",
      "advanced-v01-high-extension-2026-10-10"
    ],
    "answers": 96,
    "note": "Audited selection across two October 10 runs plus one separate xhigh run; 8 HIGH routes, 96 answers, and 1 separate xhigh lane, 12 answers. Gemini promoted via additive October 10 correction; Qwen xhigh is outside the HIGH comparison; not a single joint execution."
  },
  "models": [
    {
      "id": "anthropic/claude-haiku-5-5",
      "label": "Claude Haiku 5.5",
      "provider": "Anthropic",
      "reasoning": "high",
      "tasksCorrect": 11,
      "fieldsCorrect": 290,
      "families": {
        "allocation": {
          "tasksCorrect": 4,
          "fieldsCorrect": 80
        },
        "ledger": {
          "tasksCorrect": 4,
          "fieldsCorrect": 104
        },
        "workflow": {
          "tasksCorrect": 3,
          "fieldsCorrect": 106
        }
      },
      "runId": "advanced-v01-2026-10-10-haiku",
      "harness": "Codex native subagents",
      "runDate": "2026-10-10",
      "sourceCommit": "5f55d82430baa317bbd8dff30829b2d799f75b6b",
      "definitionHash": "e3f59476abe8455fd4e519fb3dea247270af42f7f0dddfaac4f7b861cd6309eb",
      "manifestSha256": "0a084315b28d5438ad2a2a3daf406048c8138dc0ce5c4bd763547929327e7b6e",
      "scoresSha256": "48f5b0998db0aef3517533240a4a93946c71ef58facb8b24599b94d7ee9468fd"
    },
    {
      "id": "anthropic/claude-opus-5-5",
      "label": "Claude Opus 5.5",
      "provider": "Anthropic",
      "reasoning": "high",
      "tasksCorrect": 11,
      "fieldsCorrect": 291,
      "families": {
        "allocation": {
          "tasksCorrect": 4,
          "fieldsCorrect": 80
        },
        "ledger": {
          "tasksCorrect": 3,
          "fieldsCorrect": 103
        },
        "workflow": {
          "tasksCorrect": 4,
          "fieldsCorrect": 108
        }
      },
      "runId": "advanced-v01-2026-10-10-opus",
      "harness": "Codex native subagents",
      "runDate": "2026-10-10",
      "sourceCommit": "5f55d82430baa317bbd8dff30829b2d799f75b6b",
      "definitionHash": "722c15b7dd27939c5b298d6efa6c7f49eb9a612ef70ab1fe3e726e36b7994678",
      "manifestSha256": "a2f2e872ee49b14d57e1c11b3354dbbd5e3f81271580302178fb2921c34e4fc7",
      "scoresSha256": "16f082c0966955b427aad0d930c8c9b6e668129aa5c5d616a15354d183a30e6e"
    },
    {
      "id": "opencode-go/deepseek-v4.1-flash",
      "label": "DeepSeek V4.1 Flash",
      "provider": "OpenCode Go",
      "reasoning": "high",
      "tasksCorrect": 9,
      "fieldsCorrect": 288,
      "families": {
        "allocation": {
          "tasksCorrect": 4,
          "fieldsCorrect": 80
        },
        "ledger": {
          "tasksCorrect": 4,
          "fieldsCorrect": 104
        },
        "workflow": {
          "tasksCorrect": 1,
          "fieldsCorrect": 104
        }
      },
      "runId": "advanced-v01-2026-10-10-deepseek",
      "harness": "Codex native subagents",
      "runDate": "2026-10-10",
      "sourceCommit": "5f55d82430baa317bbd8dff30829b2d799f75b6b",
      "definitionHash": "90b1a48c6548e487b81a1c802ad7c1ab6ed0d206edb8f0e58326b1f1f7a021ee",
      "manifestSha256": "7b079382e576759867a991e77bfc97c9e78e6be58b11553481d7be59027a4cae",
      "scoresSha256": "19c2f610da3d9b3c9b79b5e6cc46884401f8be4a347d230154ca3384ac142306"
    },
    {
      "id": "openai/gpt-6-astra",
      "label": "GPT-6 Astra",
      "provider": "OpenAI",
      "reasoning": "high",
      "tasksCorrect": 12,
      "fieldsCorrect": 292,
      "families": {
        "allocation": {
          "tasksCorrect": 4,
          "fieldsCorrect": 80
        },
        "ledger": {
          "tasksCorrect": 4,
          "fieldsCorrect": 104
        },
        "workflow": {
          "tasksCorrect": 4,
          "fieldsCorrect": 108
        }
      },
      "runId": "advanced-v01-high-extension-2026-10-10-astra",
      "harness": "Terminal Codex CLI 0.155.1 --ignore-user-config, ChatGPT login, read-only sandbox, empty workspace",
      "runDate": "2026-10-10",
      "sourceCommit": "a230b80a22b962b1f69f39170b22950db657986d",
      "definitionHash": "fccb8e875ae7fbc0546bd740eddc9c27a4a830bdb11414baed5964c34e1a311a",
      "manifestSha256": "a6e1ea40acd26fe488d294a8aaf9385bf20b4b70ad8a6cc0daaa3c4bfdefa76b",
      "scoresSha256": "50b516678c5c3db2397f923ade83ea3674956282fb6934b4751432fd0581fbad"
    },
    {
      "id": "codex/gpt-6-luna",
      "label": "GPT-6 Luna",
      "provider": "OpenAI",
      "reasoning": "high",
      "tasksCorrect": 4,
      "fieldsCorrect": 270,
      "families": {
        "allocation": {
          "tasksCorrect": 3,
          "fieldsCorrect": 79
        },
        "ledger": {
          "tasksCorrect": 1,
          "fieldsCorrect": 96
        },
        "workflow": {
          "tasksCorrect": 0,
          "fieldsCorrect": 95
        }
      },
      "runId": "advanced-v01-2026-10-10-luna",
      "harness": "Codex native subagents",
      "runDate": "2026-10-10",
      "sourceCommit": "5f55d82430baa317bbd8dff30829b2d799f75b6b",
      "definitionHash": "544706915eea14f643d4bb929df1b47748519423db5f3c4ac5243401ff5a2174",
      "manifestSha256": "71e4e55a203776ffd8ce4d0d2708d6a3e2cfd21958217bcf7e8ff31fae7baa9f",
      "scoresSha256": "078d9d0f541daad196d654abed9124978f9618340c34aadac5eeb3424807b87f"
    },
    {
      "id": "openai/gpt-6.1-sol",
      "label": "GPT-6.1 Sol",
      "provider": "OpenAI",
      "reasoning": "high",
      "tasksCorrect": 12,
      "fieldsCorrect": 292,
      "families": {
        "allocation": {
          "tasksCorrect": 4,
          "fieldsCorrect": 80
        },
        "ledger": {
          "tasksCorrect": 4,
          "fieldsCorrect": 104
        },
        "workflow": {
          "tasksCorrect": 4,
          "fieldsCorrect": 108
        }
      },
      "runId": "advanced-v01-high-extension-2026-10-10-sol",
      "harness": "Bundled app Codex CLI 0.162.0-alpha.17.2, user config loaded, ChatGPT login, read-only sandbox, empty workspace",
      "runDate": "2026-10-10",
      "sourceCommit": "a230b80a22b962b1f69f39170b22950db657986d",
      "definitionHash": "9ffb1f7958ea5f323cb0e4c340f6ccb3d3012a4d5230ab4cc36a0dcb9dee233b",
      "manifestSha256": "c555ade670f7d768f32e1ab7835411b329d6e61690bb3ed7649f962bb97b0a06",
      "scoresSha256": "2e96682bb57236d544b6ed2027dfbc0d3399c39383f41bbf87aed20b185a5d98"
    },
    {
      "id": "google/gemini-3.8-flash",
      "label": "Gemini 3.8 Flash",
      "provider": "Antigravity UI",
      "reasoning": "high",
      "tasksCorrect": 12,
      "fieldsCorrect": 292,
      "families": {
        "allocation": {
          "tasksCorrect": 4,
          "fieldsCorrect": 80
        },
        "ledger": {
          "tasksCorrect": 4,
          "fieldsCorrect": 104
        },
        "workflow": {
          "tasksCorrect": 4,
          "fieldsCorrect": 108
        }
      },
      "runId": "advanced-v01-high-extension-2026-10-10-gemini",
      "harness": "Antigravity native UI, No Project, new conversation per task",
      "runDate": "2026-10-10",
      "sourceCommit": "a230b80a22b962b1f69f39170b22950db657986d",
      "definitionHash": "bc4762926981f3674f92d986415fef29c50d1f4b550a0e2c48f3efe929f413a8",
      "manifestSha256": "172889dcbcb3d48f53b1c8f90d9357bcd03d2ffc2f64c604375880822b5d4004",
      "scoresSha256": "d31d7434cf37fc33733c85fa7ea62ffabd6a9214a6088b457327287717515a59"
    },
    {
      "id": "meta-muse/muse-spark-1.3-contributor",
      "label": "Muse Spark 1.3",
      "provider": "Meta Muse",
      "reasoning": "high",
      "tasksCorrect": 10,
      "fieldsCorrect": 289,
      "families": {
        "allocation": {
          "tasksCorrect": 4,
          "fieldsCorrect": 80
        },
        "ledger": {
          "tasksCorrect": 2,
          "fieldsCorrect": 101
        },
        "workflow": {
          "tasksCorrect": 4,
          "fieldsCorrect": 108
        }
      },
      "runId": "advanced-v01-high-extension-2026-10-10-muse",
      "harness": "Codex subagent V1",
      "runDate": "2026-10-10",
      "sourceCommit": "a230b80a22b962b1f69f39170b22950db657986d",
      "definitionHash": "199d5ef3e6d9e75a45e6855d3dc090f390771dc905f61d7227da6a90dce56e33",
      "manifestSha256": "1473ea0c803278eddbff91a53015d0eb6966b8fa36ecd514ad9574747aff456b",
      "scoresSha256": "d75dc632ce7a7aeab8ddc6da543b5390c1bfd4b6e3dba0074b0ff1af7e4b3281"
    }
  ],
  "corrections": [
    {
      "date": "2026-10-10",
      "lane": "sol",
      "summary": "GPT-6.1 Sol HIGH first attempt rejected in legacy CLI; completed in bundled app CLI.",
      "previousAttempt": "cli/sol with Codex 0.155.1 --ignore-user-config: model not supported with ChatGPT account; 1 attempted, 11 not run, preserved not scored.",
      "successfulRun": "cli-app/sol with bundled app Codex CLI 0.162.0-alpha.17.2, user config loaded; 12/12 tasks, 292/292 fields, transcript-audited.",
      "note": "Two settings differed; individual cause not isolated. Not a general model block."
    },
    {
      "date": "2026-10-10",
      "lane": "gemini",
      "summary": "Gemini 3.8 Flash HIGH promoted from provisional UI-captured to local-transcript-audited via additive correction; original provisional aggregates and checksums unchanged.",
      "previousAttempt": "UI-captured 12/12 and 292/292 via Antigravity; full prompt hash and backend tool log not verified.",
      "successfulRun": "Local-transcript audit of the same 12 conversations: 12/12 exact prompt matches, 12/12 final matches, model setting logged Gemini 3.8 Flash (High), 0 recorded tool calls; 12/12 tasks, 292/292 fields.",
      "note": "Local app-generated logs, not signed provider attestation; client-declared model; documented app wrapper and system instructions differ from other harnesses; one repeat, reused calibration tasks, no ranking."
    },
    {
      "date": "2026-10-10",
      "lane": "qwen-excluded-batch",
      "summary": "Two earlier Qwen batches excluded whole before scoring: 20 started attempts, 13 excluded answers; preserved, not scored.",
      "previousAttempt": "First batch: workspace under repo, ancestor AGENTS files could contaminate context - 9 started, 6 completed, 1 database startup stop, 2 interrupted, 3 not started. Second batch: inherited PWD registered repo as cwd - 11 started, 7 completed, 1 watchdog stop, 3 interrupted, 1 not started.",
      "successfulRun": "clean outside-repo rerun in isolated cwd with explicit --dir, verified exports and 900s watchdog; 12/12 tasks completed.",
      "note": "Both batches preserved whole, not scored; database startup failure and watchdog stop retained in attempt history."
    }
  ],
  "separateEfforts": [
    {
      "id": "opencode-go/qwen3.8-max",
      "label": "Qwen 3.8 Max",
      "provider": "OpenCode Go",
      "reasoning": "xhigh",
      "tasksCorrect": 12,
      "fieldsCorrect": 292,
      "families": {
        "allocation": {
          "tasksCorrect": 4,
          "fieldsCorrect": 80
        },
        "ledger": {
          "tasksCorrect": 4,
          "fieldsCorrect": 104
        },
        "workflow": {
          "tasksCorrect": 4,
          "fieldsCorrect": 108
        }
      },
      "runId": "qwen-xhigh-isolated-2026-10-10",
      "harness": "OpenCode CLI 1.18.33 --pure, OpenCode Go subscription, explicit isolated cwd outside repo, all tools denied",
      "runDate": "2026-10-10",
      "sourceCommit": "a230b80a22b962b1f69f39170b22950db657986d",
      "definitionHash": "75d73dca32bf2f5c9f4c94f4100e6a2787a5ab43fd67218d8dab3242f05a0ea4",
      "manifestSha256": "35f4c3da16476b15e1f4e73ab5407fea6d72175f43a3e71e7663742d64eb311a",
      "scoresSha256": "4e2c0130fdf98dcc45858540f97bfa6e053b544b03df9f335b163e0117167d80",
      "responses": 12,
      "auditSha256": "7a16849f6dff0cf0ea1af7ec7029b8919e34bdaf137492d060a6b56e8f663c54",
      "configNote": "OpenCode CLI isolated cwd verified in each exported session and assistant path; isolated config, neutral local AGENTS, Claude compatibility disabled, plugin loading disabled, all tools denied; exact prompt inside documented quote wrapper; watchdog 900 seconds. Local client-declared identity, not signed provider attestation.",
      "note": "Audited separate xhigh run on the same frozen tasks; not comparable to HIGH as an equal-compute test. One repetition, reused calibration, no ranking. Both earlier batches excluded whole for harness context before scoring; raw outputs preserved.",
      "missing": 0,
      "technicalErrors": 0,
      "protocolRejected": 0,
      "toolCalls": 0
    }
  ]
}
