{
  "title": "Concrete placement and specimen records for AI review",
  "source_page": "https://fiund.com/industries/concrete-placement-specimen-data",
  "authorship": "Original fictional Fiund editorial exercise. No actual operating records or third-party source assets are included.",
  "example": {
    "title": "Match a result to its specimen before matching its placement",
    "context": "An invented concrete packet uses dummy tags and arbitrary result units. It has two placements, three specimen labels and a duplicate image of one tag.",
    "records": [
      [
        "CP1",
        "Practice placement register",
        "Placement P-A links specimen C1; placement P-B links specimen C2."
      ],
      [
        "CP2",
        "Fictional test report",
        "Reviewed result 14 practice units applies to C1. A second result of 16 applies to C3, with no placement link."
      ],
      [
        "CP3",
        "Tag images",
        "Two photos both show C1. A file caption says P-B, but no authorized correction changes the register."
      ]
    ],
    "question": "Which result can be attributed to P-A or P-B, and how many independent C1 tests do the two photographs establish?",
    "answer": "CP1 and CP2 support attributing C1's reviewed 14-unit result to P-A. P-B is linked to C2, for which no result is supplied. The C3 result cannot fill that gap. CP3 shows two images of the same tag, not two independent tests. Its caption conflicts with the register and should be flagged rather than silently overriding the placement link.",
    "failureModes": [
      "Assigning C3's result to P-B because it is the remaining unmatched report.",
      "Counting repeated photos of C1 as separate physical specimens or tests."
    ],
    "fields": [
      [
        "Placement-to-specimen relation",
        "Defines the source of attribution."
      ],
      [
        "Test event identifier",
        "Separates observations from duplicate media."
      ],
      [
        "Conflicting caption and register",
        "Preserves disagreement for review."
      ],
      [
        "Result unit",
        "Keeps arbitrary exercise values separate from engineering measures."
      ]
    ],
    "holdout": "Add a valid correction linking C3 to P-B and explicitly superseding C2's original tag entry. Ask the model to resolve attribution while retaining the correction trail.",
    "boundary": "Use paper tags and invented reports. Exclude actual mix designs, engineering tolerances, client test results and active concrete handling."
  }
}