{
  "recordedAt": "2026-09-08T08:29:41.424Z",
  "sourceCommit": "159e1eb55fd87d99d399de63e12c65292f621b78",
  "title": "Finding the right API workflow — frozen production responses",
  "sourceCapture": {
    "capturedAt": "2026-09-08T08:23:46.803312+00:00",
    "origin": "https://evals.ax",
    "requests": [
      {
        "path": "/api/v1/openapi.json",
        "status": 200,
        "contentType": "application/json",
        "sha256": "6f0bb187bc2bb8e3e0bedbd74bfbe487b62edb3212ca456f08158b084d7f288f",
        "bytes": 100823
      },
      {
        "path": "/evidence/auth-context-2026-09-07.json",
        "status": 200,
        "contentType": "application/json",
        "sha256": "cb8e1cb5db90f9cad1c4d62b568389b5b619e65bc93b260e1616ae94371fa0ae",
        "bytes": 4287
      },
      {
        "path": "/evidence/dispatch-2026-09-08.json",
        "status": 200,
        "contentType": "application/json",
        "sha256": "2510d98e57db4b9d833e5adac79bb716757d5ea4ce9b5f6fb0bd1438bbf95d67",
        "bytes": 3238
      },
      {
        "path": "/docs/full.md",
        "status": 200,
        "contentType": "text/markdown; charset=utf-8",
        "sha256": "2381575b7512e2b8f37ee488ef2be7aeedeafaf9fe4a0c27bc61127443a7f6e7",
        "bytes": 44281
      },
      {
        "path": "/api/v1/account",
        "status": 401,
        "contentType": "application/json",
        "sha256": "67be62798dfa0e0d385185c2646d737d5d6fb6e8d7ea56e575e8909698243904",
        "bytes": 134
      },
      {
        "path": "/api/v1/experiments",
        "status": 401,
        "contentType": "application/json",
        "sha256": "486c5a967687930a663330e770a2bd55383f4a60c3041581590122877d3fa34b",
        "bytes": 134
      }
    ]
  },
  "primary": "Every required value, workflow step, request shape and supporting source quotation independently accepted within the fixed budget. All failures/timeouts count; verifier errors suppress inference.",
  "secondary": "Duration, provider token usage and command diagnostics descriptive only. No efficiency inference. Missing dollar cost remains unknown.",
  "sample": "Exactly four randomized pairs/eight planned trials; once only; retain every outcome; no reruns, replacements or extension after seeing results.",
  "qualification": "Six independent verifier controls declared before execution; one complete handoff and five distinct wrong-answer controls. Stock read-only network capability preflight failed DNS, retained in network-preflight.log, not a discarded agent trial.",
  "scope": "Real Codex processing freshly captured production public API/documentation responses; no per-trial network, browser, mutation or account authentication. Ordinary local macOS host with fresh copied directories; no external runtime/filesystem/model-context attestation.",
  "bounds": {
    "agent": {
      "kind": "codex",
      "model": "gpt-6-astra",
      "reasoning": "low",
      "executable": "/Applications/ChatGPT.app/Contents/Resources/codex",
      "configuration": "isolated"
    },
    "limits": {
      "trialTimeoutMs": 120000,
      "verifierTimeoutMs": 5000,
      "totalTimeoutMs": 1080000,
      "maxOutputBytes": 2000000,
      "maxSurfaceBytes": 1000000,
      "maxSurfaceFiles": 50
    },
    "cost": "Provider-reported dollar cost expected unavailable; no enforced Codex monetary cap. Existing host authorization only; no purchases."
  },
  "changes": "Only candidate AGENTS.md adds a navigation index; ANSWER_FORMAT.json and every production response byte are identical.",
  "surfaceHashes": {
    "baseline": "e8c6323ce9b69d831f185fb304ff128a087303e9e0d425bfcd0ed7a995bc4924",
    "candidate": "3fb7fe34c149d81d904b7e505a1ac7697fa42ccaa6c3e9da696e01a188ba4db2"
  },
  "manifestFileSha256": "90e387c2f08e7bb3d8f8ad7f821497082fe9ebe5be2a51eb07bc5fbb75cbeb98",
  "trialPlan": [
    {
      "id": "1-candidate",
      "pair": 0,
      "variant": "candidate",
      "order": 0
    },
    {
      "id": "1-baseline",
      "pair": 0,
      "variant": "baseline",
      "order": 1
    },
    {
      "id": "2-baseline",
      "pair": 1,
      "variant": "baseline",
      "order": 2
    },
    {
      "id": "2-candidate",
      "pair": 1,
      "variant": "candidate",
      "order": 3
    },
    {
      "id": "3-candidate",
      "pair": 2,
      "variant": "candidate",
      "order": 4
    },
    {
      "id": "3-baseline",
      "pair": 2,
      "variant": "baseline",
      "order": 5
    },
    {
      "id": "4-baseline",
      "pair": 3,
      "variant": "baseline",
      "order": 6
    },
    {
      "id": "4-candidate",
      "pair": 3,
      "variant": "candidate",
      "order": 7
    }
  ]
}