{
  "title": "Pharmaceutical preparation data for AI record corrections",
  "source_page": "https://fiund.com/industries/pharmaceutical-record-correction-chain-data",
  "authorship": "Original fictional Fiund editorial exercise. No actual operating records or third-party source assets are included.",
  "example": {
    "title": "A corrected report without a new sample",
    "context": "A wholly fictional administrative packet uses sample code Q9. It contains no medicine, test values, patient information or disposition.",
    "records": [
      [
        "1",
        "Sample register",
        "Q9 registered once under batch B8."
      ],
      [
        "2",
        "Report v1",
        "Q9 report filed under folder F3."
      ],
      [
        "3",
        "Correction v2",
        "Same sample Q9; correct folder F5; v1 retained and superseded."
      ]
    ],
    "question": "How many samples and current reports should the index show?",
    "answer": "There is one sample, Q9, according to record 1. Record 3 replaces the current report linkage with v2/F5 and retains v1/F3 from record 2 as superseded history. No new sample, measurement, or release status is implied by this administrative correction.",
    "failureModes": [
      "Counting report versions as samples",
      "Erasing the superseded report or its old reference"
    ],
    "fields": [
      [
        "sample_key",
        "Deduplicates across document versions"
      ],
      [
        "current_report_version",
        "Selects effective correction"
      ],
      [
        "supersession_edge",
        "Preserves the earlier record for traceability"
      ]
    ],
    "holdout": "Include an unapproved proposed v3; keep v2 current until a formal effectivity record appears.",
    "boundary": "No production, pharmaceutical processing, assays, approval procedures or actual regulated documentation."
  }
}