{
  "schemaVersion": 1,
  "reportDate": "2026-10-15",
  "scope": "Sanitized public summary of the Listening Party prompt-finder evaluation and verified release.",
  "provenance": {
    "publicSite": "https://listeningparty.online",
    "deployedApplicationSource": "f497d77fc5c21f0718db8fe90d8f46e710e2dd66",
    "releaseRecordRevision": "abd4329732725481fcc158a14c04bd82ef1cfa62",
    "productionVersion": "be6216e1-e428-4a9d-b50c-2886fc60d6e1",
    "evaluationExecutionSource": "55e7bf73cd7498d0e0e0f8ce93d14146918d0fcd",
    "executionFreezeSha256": "f7a059213e7235e23b361fa1ece8f60538350f9ddeb6bacc1d654c20da662717"
  },
  "decision": {
    "defaultMethod": "BM25",
    "experimentalMethod": "compact GloVe semantic comparator",
    "candidateSelection": null,
    "summary": "No evidence-policy candidate qualified in development. BM25 remains the default."
  },
  "evaluation": {
    "metricDefinitions": {
      "ordinaryHitAt3": "An expected card appears among the first three results for one of 16 ordinary cases.",
      "answerableCoverage": "Any suggestion appears for one of 22 eligible answerable cases; this is not a correctness score.",
      "unsupportedSuggestions": "Any recommendation appears for one of eight no-answer cases."
    },
    "caseSet": {
      "total": 74,
      "authoring": "synthetic, agent-authored English cases",
      "development": {
        "cases": 40,
        "intentFamilies": 20,
        "eligibleAnswerable": 22,
        "ordinary": 16,
        "unsupported": 8
      },
      "holdout": {
        "cases": 34,
        "intentFamilies": 17,
        "eligibleAnswerable": 22,
        "ordinary": 16,
        "unsupported": 8
      }
    },
    "development": {
      "bm25Comparator": {
        "ordinaryHitAt3": "13/16",
        "answerableHitAt3": "16/22",
        "answerableCoverage": "20/22",
        "unsupportedSuggestions": "6/8"
      },
      "semanticComparator": {
        "ordinaryHitAt3": "14/16",
        "answerableHitAt3": "19/22",
        "answerableCoverage": "22/22",
        "unsupportedSuggestions": "8/8"
      },
      "evidencePolicies": {
        "evidence050": {
          "ordinaryHitAt3": "0/16",
          "answerableHitAt3": "0/22",
          "answerableCoverage": "0/22",
          "unsupportedSuggestions": "0/8"
        },
        "evidence067": {
          "ordinaryHitAt3": "0/16",
          "answerableHitAt3": "0/22",
          "answerableCoverage": "0/22",
          "unsupportedSuggestions": "0/8"
        },
        "evidence080": {
          "ordinaryHitAt3": "0/16",
          "answerableHitAt3": "0/22",
          "answerableCoverage": "0/22",
          "unsupportedSuggestions": "0/8"
        }
      }
    },
    "holdout": {
      "policyCandidatesRan": false,
      "bm25Baseline": {
        "ordinaryHitAt3": "12/16",
        "answerableHitAt3": "18/22",
        "answerableCoverage": "22/22",
        "unsupportedSuggestions": "6/8"
      },
      "semanticBaseline": {
        "ordinaryHitAt3": "13/16",
        "answerableHitAt3": "19/22",
        "answerableCoverage": "22/22",
        "unsupportedSuggestions": "8/8"
      }
    }
  },
  "releaseVerification": {
    "liveCheckpointsPassed": 12,
    "viewports": [390, 768, 1440],
    "verifiedBehaviors": [
      "matched-word, source and common-word explanations",
      "search, category, clear and close recovery",
      "keyboard traversal with reduced motion and no horizontal overflow",
      "real-browser optional-model checksum",
      "explicit host deal and separate guest card readback"
    ],
    "slowDownloadCorrection": {
      "previousDeadlineSeconds": 20,
      "currentDeadlineSeconds": 90,
      "regressions": [
        "58-second simulated download completed while keyword results remained usable",
        "stalled download reached the 90-second bound and offered retry"
      ]
    },
    "gameTestSuite": {
      "scope": "full game test suite",
      "passed": 126,
      "skippedOptInProviderTests": 2,
      "typecheckPassed": true,
      "changedFileLintPassed": true,
      "freshCloudflareBuildPassed": true
    }
  },
  "limitations": [
    "The cases are synthetic and agent-authored, not user research or a customer study.",
    "Full blinding was not claimed.",
    "The small baseline difference does not establish a general semantic-search advantage.",
    "No policy candidate ran on holdout because none qualified in development.",
    "The evidence does not verify microphone input, transcription, paid providers, physical devices or multi-device audio.",
    "The finder searches authored cards locally; it does not generate cards or use LangChain."
  ]
}