{
  "rule": "A finding is not trusted until it can be reproduced from the source by a reader who has not seen it. The second pass is given the draft and the same source set, and is asked the verifier's question. It is never shown the finding, the quoted line, or the severity. If it cannot get there on its own, the finding is withdrawn.",
  "why": "A verifier that is never checked is just a confident voice. The point of the second pass is not to agree. Two of the twelve findings put through it did not survive, and both of those were the fleet being wrong rather than the document.",
  "selection": "Every blocker on the run, which is ten, plus two majors chosen because they are the awkward kind: one that changes no words in the document, and one where two documents disagree and the verifier will not say which is right.",
  "ranOn": "2026-09-04",
  "checks": [
    {
      "id": "r01",
      "findingId": "f13",
      "draftId": "d3-service-levels",
      "verifierId": "v04-time-promise",
      "claimed": "The sheet states a four hour response, twenty four hours a day, and no source supports it.",
      "method": "The second pass was given the sheet and the five declared sources, and one question: for every time commitment on this page, name the source line it comes from.",
      "found": "Ten of the eleven time figures were traced to the operations turnaround table by line. The eleventh could not be traced. The second pass went further than the first and read the support routing table's staffed hours, then said in its own words that the page promises cover the routing table explicitly does not provide.",
      "survived": true,
      "consequence": "Confirmed as a blocker. Two independent reads, five source documents each, and the sentence still had no parent."
    },
    {
      "id": "r02",
      "findingId": "f03",
      "draftId": "d1-partner-letter",
      "verifierId": "v18-obligation-language",
      "claimed": "The letter guarantees a maximum out of service time. The clause library offers no availability guarantee.",
      "method": "The second pass was asked to list every sentence in the letter that creates an obligation, then match each one to an approved clause.",
      "found": "Three obligation sentences. Two matched clauses word for word. The third had no match, and the second pass quoted the clause library line saying no availability guarantee is offered under any plan.",
      "survived": true,
      "consequence": "Confirmed. Escalated to the COO before it was rewritten, because a guarantee is a decision and not a wording problem."
    },
    {
      "id": "r03",
      "findingId": "f06",
      "draftId": "d2-price-sheet",
      "verifierId": "v02-figure-match",
      "claimed": "The Field Kit is printed at 1,420 and the price master says 1,240.",
      "method": "The second pass was given the sheet and the price master and asked to read every row against the master without being told which row was in question.",
      "found": "The same row, the same transposition. It also reported that the other eleven rows matched, which is the part a finding on its own does not tell you.",
      "survived": true,
      "consequence": "Confirmed. The clean rows matter: they are the reason the sheet was corrected rather than rebuilt."
    },
    {
      "id": "r04",
      "findingId": "f07",
      "draftId": "d2-price-sheet",
      "verifierId": "v03-price-parity",
      "claimed": "The price sheet shows partner prices and the one pager shows list prices, and neither says which.",
      "method": "The second pass was given both documents and the price master and asked which column each document was quoting.",
      "found": "It could not tell from either document. It had to work it out by arithmetic, noticing that every figure on the sheet was exactly ten percent below the master's list column. That is the finding restated as an experience.",
      "survived": true,
      "consequence": "Confirmed. Both documents now name their column."
    },
    {
      "id": "r05",
      "findingId": "f10",
      "draftId": "d2-price-sheet",
      "verifierId": "v07-product-line",
      "claimed": "Meridian Bench Plus is on the sheet and does not exist.",
      "method": "The second pass was asked to list every product named in the sheet and mark each one present or absent in the specification.",
      "found": "Eleven present, one absent. It also searched the price master for the string in four spellings and found nothing.",
      "survived": true,
      "consequence": "Confirmed. Logged as a product question rather than a document defect, because somebody in July thought it existed."
    },
    {
      "id": "r06",
      "findingId": "f14",
      "draftId": "d3-service-levels",
      "verifierId": "v08-plan-features",
      "claimed": "The sheet says both plans include unlimited on-site visits. The plan definitions exclude on-site attendance from both.",
      "method": "The second pass was asked to build the inclusion list for each tier from the plan definitions alone, then compare it to the sheet.",
      "found": "Its list carried on-site attendance under exclusions for both tiers, in the definitions' own words. The word unlimited appears nowhere in the definitions.",
      "survived": true,
      "consequence": "Confirmed."
    },
    {
      "id": "r07",
      "findingId": "f16",
      "draftId": "d3-service-levels",
      "verifierId": "v18-obligation-language",
      "claimed": "The sheet says Apex will ensure the bench is never without a working instrument.",
      "method": "Same method as the letter. List the obligations, match each to a clause.",
      "found": "No matching clause. The second pass added something the first did not: the clause library offers loan instruments where stock allows, so the sheet had taken a conditional offer and printed it as a certainty.",
      "survived": true,
      "consequence": "Confirmed, and the rewrite used the second pass's wording because it kept the condition."
    },
    {
      "id": "r08",
      "findingId": "f34",
      "draftId": "d6-launch-note",
      "verifierId": "v07-product-line",
      "claimed": "The note says the Field Kit ships with the Meridian Pro calibration cradle. The cradle fits the Bench only.",
      "method": "The second pass was asked what ships in the Field Kit box, from the specification, before being shown the note.",
      "found": "Field carry case and standard probe set. No cradle. The specification says in a compatibility line that the cradle fits the Bench only.",
      "survived": true,
      "consequence": "Confirmed. This one was going to every existing partner contact, which is why it was worth two reads."
    },
    {
      "id": "r09",
      "findingId": "f26",
      "draftId": "d5-proposal",
      "verifierId": "v21-source-freshness",
      "claimed": "The training day figure is correct and its source is 114 days past its review date.",
      "method": "The second pass was asked to check the figure, and separately to report the review state of every source the proposal declared.",
      "found": "The figure matches the catalog exactly. The catalog was last reviewed on November 14, 2025 with a 180 day interval, so it was 114 days overdue on the day the proposal was built. Both halves of the finding held.",
      "survived": true,
      "consequence": "Confirmed and accepted. Nothing in the document changed. The value of this finding is that somebody now knows the proposal rests on an unreviewed price."
    },
    {
      "id": "r10",
      "findingId": "f31",
      "draftId": "d5-proposal",
      "verifierId": "v17-cross-document",
      "claimed": "The proposal says Atlas Care Plus is 149 per month and the service levels sheet in the same batch says 139.",
      "method": "The second pass was given all seven drafts and asked to list every fact stated in more than one of them and flag any that disagreed.",
      "found": "Four facts appear in more than one draft. Three agree. One does not, and it is this one. The second pass also declined to say which document was right, which is the correct answer for this verifier.",
      "survived": true,
      "consequence": "Confirmed. Resolved by the price master, not by the verifier."
    },
    {
      "id": "r11",
      "findingId": "f27",
      "draftId": "d5-proposal",
      "verifierId": "v02-figure-match",
      "claimed": "The proposal prices the Meridian Pro at 7,110 and the price master says 7,900.",
      "method": "The second pass was given the proposal and the whole price master, both columns, and asked to check the pricing summary.",
      "found": "7,110 is the partner column for MER-PRO-STD, exactly ten percent below list, and the proposal is a partner document. The figure was right. The first pass had compared against the list column and stopped there.",
      "survived": false,
      "consequence": "Finding withdrawn. The verifier's reading rule was changed the same day: a figure match finding must now name the column it compared against, which would have made the mistake visible in the finding itself. This is the one the fleet got wrong on the document's side, and it is the reason the reproduction rule exists.",
      "whoWasWrong": "the fleet"
    },
    {
      "id": "r12",
      "findingId": "f28",
      "draftId": "d5-proposal",
      "verifierId": "v01-unsourced-claim",
      "claimed": "The proposal states a territory pricing rule that appears in none of its seven sources.",
      "method": "The second pass was asked to trace the sentence, and was given the partner file rather than the indexed source list, because the adjudicator wanted to know whether the claim was true and not only whether it was indexed.",
      "found": "The rule is in appendix B of the signed partner terms, in plain language, dated March 2026. The appendix is a separate file. The source index recorded the terms and not the attachment, so the verifier read a document that had been quietly cut in half.",
      "survived": false,
      "consequence": "Finding withdrawn. The document was right. The defect was in what the fleet was given to read, which is the first blind spot this verifier declares in its own entry. The source index now pulls attachments, and the run was repeated. This is the more uncomfortable of the two, because the verifier behaved correctly on the evidence it had and was still wrong.",
      "whoWasWrong": "the source index"
    }
  ]
}