{
 "golden_set": "exhibit thirteen",
 "authored_for": "the llms.txt story, measured",
 "counts": {
  "items_total": 30,
  "run_set": 24,
  "reserve": 6,
  "headline_set": 20,
  "core_frozen": 14,
  "core_reserve": 6,
  "nav": 6,
  "bait": 4,
  "exact": 20,
  "proxy": 4
 },
 "answer_instruction": "Answer with only the fact or the URL, on one line. If the site content above does not say, reply exactly: the site doesn't say.",
 "run_set": [
  "C01",
  "C02",
  "C03",
  "C04",
  "C05",
  "C06",
  "C07",
  "C08",
  "C09",
  "C10",
  "C11",
  "C12",
  "C13",
  "C14",
  "N01",
  "N02",
  "N03",
  "N04",
  "N05",
  "N06",
  "B01",
  "B02",
  "B03",
  "B04"
 ],
 "reserve": [
  "R01",
  "R02",
  "R03",
  "R04",
  "R05",
  "R06"
 ],
 "headline_set": [
  "C01",
  "C02",
  "C03",
  "C04",
  "C05",
  "C06",
  "C07",
  "C08",
  "C09",
  "C10",
  "C11",
  "C12",
  "C13",
  "C14",
  "N01",
  "N02",
  "N03",
  "N04",
  "N05",
  "N06"
 ],
 "items": [
  {
   "id": "C01",
   "category": "core",
   "question": "How many scoring cells did the open call's judges file verdicts on?",
   "key": "738",
   "checker": "check_C01",
   "stamp": "exact",
   "frozen": true,
   "published": "2026-08-15",
   "source": "the-open-call: 'filed verdicts on 738 scoring cells'",
   "expected_stratum": "HTML-ONLY",
   "variants": [
    "738 scoring cells",
    "738 cells",
    "seven hundred thirty-eight"
   ],
   "decoys": [
    "868",
    "102",
    "840",
    "24",
    "20"
   ],
   "evidence": [
    "738 scoring cells"
   ],
   "note": "the map states 868 verdicts for the same panel; 738 is the scoring subset. The two numbers on one subject are why the no-alternatives rule exists."
  },
  {
   "id": "C02",
   "category": "core",
   "question": "On what date was the open call's round sealed and scored?",
   "key": "2026-08-14",
   "checker": "check_C02",
   "stamp": "exact",
   "frozen": true,
   "published": "2026-08-15",
   "source": "the-open-call: 'round sealed and scored 2026-08-14'",
   "expected_stratum": "HTML-ONLY",
   "variants": [
    "14 August 2026",
    "August 14, 2026",
    "August 14th 2026",
    "2026/08/14"
   ],
   "decoys": [
    "2026-08-15",
    "2026-08-16",
    "2026-08-13",
    "2026-08-11"
   ],
   "evidence": [
    "round sealed and scored 2026-08-14"
   ],
   "note": ""
  },
  {
   "id": "C03",
   "category": "core",
   "question": "How many of the open call's sealed asks were judged?",
   "key": "4",
   "checker": "check_C03",
   "stamp": "exact",
   "frozen": true,
   "published": "2026-08-15",
   "source": "the-open-call: 'Four asks were judged'",
   "expected_stratum": "HTML-ONLY",
   "variants": [
    "four",
    "Four asks",
    "4 asks"
   ],
   "decoys": [
    "2",
    "3",
    "6",
    "20",
    "24"
   ],
   "evidence": [
    "Four asks were judged"
   ],
   "note": ""
  },
  {
   "id": "C04",
   "category": "core",
   "question": "What is the registered tie band on the open call's 0-10 scale?",
   "key": "0.5",
   "checker": "check_C04",
   "stamp": "exact",
   "frozen": true,
   "published": "2026-08-15",
   "source": "the-open-call: 'registered tie band of 0.5'",
   "expected_stratum": "HTML-ONLY",
   "variants": [
    "0.5 points",
    "half a point",
    "a band of 0.5"
   ],
   "decoys": [
    "0.05",
    "5",
    "0.1",
    "7.33",
    "0.37"
   ],
   "evidence": [
    "registered tie band of 0.5"
   ],
   "note": "0.05 is the measured gap on the same page and is the sharpest available decoy: a containment checker matching '0.5' inside '0.05' would pass it."
  },
  {
   "id": "C05",
   "category": "core",
   "question": "Which model spoke the line about being the only child left in the cove, unless somebody's hiding one in a barrel?",
   "key": "openai-gpt-5.5-2026-04-23",
   "checker": "check_C05",
   "stamp": "exact",
   "frozen": true,
   "published": "2026-08-15",
   "source": "the-open-call pull quote attribution",
   "expected_stratum": "HTML-ONLY",
   "variants": [
    "gpt-5.5-2026-04-23",
    "openai gpt-5.5-2026-04-23",
    "GPT-5.5"
   ],
   "decoys": [
    "qwen3.8:27b",
    "qwen3.6:27b",
    "gemma4:26b",
    "cloud-minimax-m3",
    "deepseek-v4-pro"
   ],
   "evidence": [
    "openai-gpt-5.5-2026-04-23, speaking as"
   ],
   "note": ""
  },
  {
   "id": "C06",
   "category": "core",
   "question": "How many verdicts in total did the open call's panel write?",
   "key": "868",
   "checker": "check_C06",
   "stamp": "exact",
   "frozen": true,
   "published": "2026-08-15",
   "source": "research llms.txt: '868 verdicts from a seven-judge, six-family panel'",
   "expected_stratum": "MAP-ONLY",
   "variants": [
    "868 verdicts",
    "eight hundred sixty-eight"
   ],
   "decoys": [
    "738",
    "102",
    "2592",
    "840"
   ],
   "evidence": [
    "868 verdicts"
   ],
   "note": ""
  },
  {
   "id": "C07",
   "category": "core",
   "question": "How many weeks of the app's own timing receipts does The move publish?",
   "key": "5",
   "checker": "check_C07",
   "stamp": "exact",
   "frozen": true,
   "published": "2026-08-14",
   "source": "research llms.txt: 'five weeks of the app's own timing receipts'",
   "expected_stratum": "MAP-ONLY",
   "variants": [
    "five",
    "five weeks",
    "5 weeks"
   ],
   "decoys": [
    "4",
    "6",
    "30",
    "three"
   ],
   "evidence": [
    "five weeks of the app's own timing receipts"
   ],
   "note": ""
  },
  {
   "id": "C08",
   "category": "core",
   "question": "Which model was added to the open call as a twenty-first arm after publication?",
   "key": "qwen3.8:27b",
   "checker": "check_C08",
   "stamp": "exact",
   "frozen": true,
   "published": "2026-08-16",
   "source": "the-open-call post-publication addendum",
   "expected_stratum": "NEITHER",
   "variants": [
    "qwen3.8 27b",
    "Qwen3.8:27B",
    "qwen3.8-27b"
   ],
   "decoys": [
    "qwen3.6:27b",
    "qwen3.5:27b",
    "gemma4:26b",
    "llama3.3:70b",
    "openai-gpt-5.5-2026-04-23"
   ],
   "evidence": [
    "qwen3.8:27b"
   ],
   "note": "the freshest fact on the site: landed 2026-08-16, absent from both llms.txt files, and past the C-HTML truncation cut."
  },
  {
   "id": "C09",
   "category": "core",
   "question": "What score did the open call's twenty-first arm receive on the panel's 0-10 scale?",
   "key": "7.33",
   "checker": "check_C09",
   "stamp": "exact",
   "frozen": true,
   "published": "2026-08-16",
   "source": "the-open-call addendum: 'it scores 7.33'",
   "expected_stratum": "NEITHER",
   "variants": [
    "7.33 out of 10",
    "a 7.33"
   ],
   "decoys": [
    "7.28",
    "7.16",
    "6.91",
    "7.3",
    "0.5"
   ],
   "evidence": [
    "it scores 7.33"
   ],
   "note": ""
  },
  {
   "id": "C10",
   "category": "core",
   "question": "To what version was the ollama daemon upgraded in order to admit the open call's twenty-first arm?",
   "key": "0.32.13",
   "checker": "check_C10",
   "stamp": "exact",
   "frozen": true,
   "published": "2026-08-16",
   "source": "the-open-call addendum: 'the daemon was upgraded to 0.32.13'",
   "expected_stratum": "NEITHER",
   "variants": [
    "ollama 0.32.13",
    "v0.32.13"
   ],
   "decoys": [
    "0.32.9",
    "0.31.1",
    "0.32.12",
    "0.33.0"
   ],
   "evidence": [
    "upgraded to 0.32.13"
   ],
   "note": ""
  },
  {
   "id": "C11",
   "category": "core",
   "question": "How many days does every log on the server now keep, after the August 2026 retention change?",
   "key": "30",
   "checker": "check_C11",
   "stamp": "exact",
   "frozen": true,
   "published": "2026-08-15",
   "source": "where-your-question-goes: 'keeps thirty days instead of a century'",
   "expected_stratum": "NEITHER",
   "variants": [
    "thirty",
    "thirty days",
    "30 days"
   ],
   "decoys": [
    "100",
    "365",
    "90",
    "7",
    "14"
   ],
   "evidence": [
    "thirty days instead of a century"
   ],
   "note": ""
  },
  {
   "id": "C12",
   "category": "core",
   "question": "With what single word is the question part of a web address now replaced in the server's logs?",
   "key": "REDACTED",
   "checker": "check_C12",
   "stamp": "exact",
   "frozen": true,
   "published": "2026-08-15",
   "source": "where-your-question-goes: 'replaced by the word REDACTED'",
   "expected_stratum": "NEITHER",
   "variants": [
    "redacted",
    "'REDACTED'"
   ],
   "decoys": [
    "removed",
    "deleted",
    "hashed",
    "anonymised",
    "stripped"
   ],
   "evidence": [
    "replaced by the word REDACTED"
   ],
   "note": ""
  },
  {
   "id": "C13",
   "category": "core",
   "question": "How many judges read every open call reply blind?",
   "key": "7",
   "checker": "check_C13",
   "stamp": "exact",
   "frozen": true,
   "published": "2026-08-15",
   "source": "research llms.txt 'seven-judge' AND the-open-call 'seven judges'",
   "expected_stratum": "BOTH",
   "variants": [
    "seven",
    "seven judges",
    "7 judges"
   ],
   "decoys": [
    "6",
    "5",
    "20",
    "4"
   ],
   "evidence": [
    "seven judges",
    "seven-judge"
   ],
   "note": ""
  },
  {
   "id": "C14",
   "category": "core",
   "question": "What email address does the site give for getting in touch?",
   "key": "hello@strata2signal.com",
   "checker": "check_C14",
   "stamp": "exact",
   "frozen": true,
   "published": "2026-08-16",
   "source": "both llms.txt files AND both front pages",
   "expected_stratum": "BOTH",
   "variants": [
    "mailto:hello@strata2signal.com",
    "hello [at] strata2signal.com"
   ],
   "decoys": [
    "contact@strata2signal.com",
    "info@strata2signal.com",
    "hello@research.strata2signal.com"
   ],
   "evidence": [
    "hello@strata2signal.com"
   ],
   "note": ""
  },
  {
   "id": "R01",
   "category": "core",
   "question": "On what date was the open call exhibit published?",
   "key": "2026-08-15",
   "checker": "check_R01",
   "stamp": "exact",
   "frozen": false,
   "published": "2026-08-15",
   "source": "the-open-call dateline: 'Published 2026-08-15'",
   "expected_stratum": "HTML-ONLY",
   "variants": [
    "15 August 2026",
    "August 15, 2026"
   ],
   "decoys": [
    "2026-08-14",
    "2026-08-16",
    "2026-08-13"
   ],
   "evidence": [
    "Published 2026-08-15"
   ],
   "note": ""
  },
  {
   "id": "R02",
   "category": "core",
   "question": "Which exhibit number does the open call carry?",
   "key": "11",
   "checker": "check_R02",
   "stamp": "exact",
   "frozen": false,
   "published": "2026-08-15",
   "source": "the-open-call: 'Exhibit eleven'",
   "expected_stratum": "HTML-ONLY",
   "variants": [
    "eleven",
    "exhibit eleven",
    "#11"
   ],
   "decoys": [
    "10",
    "12",
    "21",
    "13"
   ],
   "evidence": [
    "Exhibit eleven"
   ],
   "note": ""
  },
  {
   "id": "R03",
   "category": "core",
   "question": "How many tags did the shelf reading taken five minutes before the open call's seal list?",
   "key": "18",
   "checker": "check_R03",
   "stamp": "exact",
   "frozen": false,
   "published": "2026-08-16",
   "source": "the-open-call addendum: 'lists eighteen tags'",
   "expected_stratum": "NEITHER",
   "variants": [
    "eighteen",
    "eighteen tags"
   ],
   "decoys": [
    "19",
    "20",
    "21",
    "23"
   ],
   "evidence": [
    "lists eighteen tags"
   ],
   "note": ""
  },
  {
   "id": "R04",
   "category": "core",
   "question": "On what date did CUDA support land for the model whose bench How we work walks through?",
   "key": "2026-08-11",
   "checker": "check_R04",
   "stamp": "exact",
   "frozen": false,
   "published": "2026-08-14",
   "source": "how-we-work: 'CUDA support ... had landed that morning, August 11'",
   "expected_stratum": "NEITHER",
   "variants": [
    "August 11",
    "11 August 2026",
    "August 11, 2026"
   ],
   "decoys": [
    "2026-08-12",
    "2026-08-14",
    "2026-08-10"
   ],
   "evidence": [
    "had landed that morning, August 11"
   ],
   "note": ""
  },
  {
   "id": "R05",
   "category": "core",
   "question": "How many new verdict objects did the outside judges produce?",
   "key": "2592",
   "checker": "check_R05",
   "stamp": "exact",
   "frozen": false,
   "published": "2026-08-14",
   "source": "research llms.txt: '2 592 new verdict objects'",
   "expected_stratum": "MAP-ONLY",
   "variants": [
    "2,592",
    "2 592",
    "2592 verdicts"
   ],
   "decoys": [
    "868",
    "738",
    "2492",
    "1296"
   ],
   "evidence": [
    "2 592 new verdict objects"
   ],
   "note": ""
  },
  {
   "id": "R06",
   "category": "core",
   "question": "Besides its predecessor, how many other sealed arms sit inside the same tie band as the open call's twenty-first arm?",
   "key": "6",
   "checker": "check_R06",
   "stamp": "exact",
   "frozen": false,
   "published": "2026-08-16",
   "source": "the-open-call addendum: 'six other sealed arms sit inside the same 0.5'",
   "expected_stratum": "NEITHER",
   "variants": [
    "six",
    "six arms"
   ],
   "decoys": [
    "7",
    "5",
    "20",
    "2"
   ],
   "evidence": [
    "six other sealed arms sit inside the same 0.5"
   ],
   "note": ""
  },
  {
   "id": "N01",
   "category": "nav",
   "question": "Which URL on this site explains, in plain english, what happens to a question you type in — the privacy walk?",
   "key": "https://research.strata2signal.com/where-your-question-goes/index.html",
   "checker": "check_N01",
   "stamp": "exact",
   "frozen": true,
   "published": "2026-08-15",
   "source": "both llms.txt files list it",
   "expected_stratum": "MAP-ONLY",
   "variants": [
    "research.strata2signal.com/where-your-question-goes/",
    "/where-your-question-goes/",
    "https://research.strata2signal.com/where-your-question-goes/"
   ],
   "decoys": [
    "https://research.strata2signal.com/how-we-work/index.html",
    "https://strata2signal.com/contact/",
    "https://research.strata2signal.com/licences/index.html"
   ],
   "evidence": [
    "research.strata2signal.com/where-your-question-goes/"
   ],
   "note": ""
  },
  {
   "id": "N02",
   "category": "nav",
   "question": "Which URL on this site gives every model's licence, read first-hand?",
   "key": "https://research.strata2signal.com/licences/index.html",
   "checker": "check_N02",
   "stamp": "exact",
   "frozen": true,
   "published": "2026-08-16",
   "source": "research llms.txt exhibit list",
   "expected_stratum": "MAP-ONLY",
   "variants": [
    "research.strata2signal.com/licences/",
    "/licences/"
   ],
   "decoys": [
    "https://research.strata2signal.com/seat-trials/index.html",
    "https://research.strata2signal.com/data/index.json",
    "https://research.strata2signal.com/august-arrivals/index.html"
   ],
   "evidence": [
    "research.strata2signal.com/licences/"
   ],
   "note": ""
  },
  {
   "id": "N03",
   "category": "nav",
   "question": "Which URL is the machine-readable discovery file listing every exhibit and its endpoints?",
   "key": "https://research.strata2signal.com/data/index.json",
   "checker": "check_N03",
   "stamp": "exact",
   "frozen": true,
   "published": "2026-08-16",
   "source": "research llms.txt 'Data kits' section",
   "expected_stratum": "MAP-ONLY",
   "variants": [
    "research.strata2signal.com/data/index.json",
    "/data/index.json"
   ],
   "decoys": [
    "https://research.strata2signal.com/llms.txt",
    "https://research.strata2signal.com/robots.txt",
    "https://research.strata2signal.com/chair-trials/data/index.json"
   ],
   "evidence": [
    "research.strata2signal.com/data/index.json"
   ],
   "note": ""
  },
  {
   "id": "N04",
   "category": "nav",
   "question": "Which URL explains who writes this hub and how?",
   "key": "https://research.strata2signal.com/how-we-work/index.html",
   "checker": "check_N04",
   "stamp": "exact",
   "frozen": true,
   "published": "2026-08-16",
   "source": "research llms.txt exhibit list",
   "expected_stratum": "MAP-ONLY",
   "variants": [
    "research.strata2signal.com/how-we-work/",
    "/how-we-work/"
   ],
   "decoys": [
    "https://research.strata2signal.com/where-your-question-goes/index.html",
    "https://strata2signal.com/family.html",
    "https://research.strata2signal.com/the-open-call/index.html"
   ],
   "evidence": [
    "research.strata2signal.com/how-we-work/"
   ],
   "note": ""
  },
  {
   "id": "N05",
   "category": "nav",
   "question": "Which URL maps the whole strata-signal family — every product, tool and machine in the workshop?",
   "key": "https://strata2signal.com/family.html",
   "checker": "check_N05",
   "stamp": "exact",
   "frozen": true,
   "published": "2026-08-16",
   "source": "apex llms.txt 'Start here'",
   "expected_stratum": "MAP-ONLY",
   "variants": [
    "strata2signal.com/family.html",
    "/family.html"
   ],
   "decoys": [
    "https://strata2signal.com/",
    "https://strata2signal.com/contact/",
    "https://research.strata2signal.com/"
   ],
   "evidence": [
    "strata2signal.com/family.html"
   ],
   "note": ""
  },
  {
   "id": "N06",
   "category": "nav",
   "question": "Which URL is the exhibit where twenty models auditioned as the game world's narrator?",
   "key": "https://research.strata2signal.com/the-open-call/index.html",
   "checker": "check_N06",
   "stamp": "exact",
   "frozen": true,
   "published": "2026-08-16",
   "source": "research llms.txt exhibit list",
   "expected_stratum": "MAP-ONLY",
   "variants": [
    "research.strata2signal.com/the-open-call/",
    "/the-open-call/"
   ],
   "decoys": [
    "https://research.strata2signal.com/voice-trials/index.html",
    "https://research.strata2signal.com/cove-voice-head-to-head/index.html",
    "https://research.strata2signal.com/chair-trials/index.html"
   ],
   "evidence": [
    "research.strata2signal.com/the-open-call/"
   ],
   "note": ""
  },
  {
   "id": "B01",
   "category": "bait",
   "question": "What per-seat price does the site's pricing page quote for the private document box?",
   "key": "ABSTAIN",
   "checker": "check_B01",
   "stamp": "proxy",
   "frozen": true,
   "published": "n/a",
   "source": "no pricing page exists; the apex says 'no per-seat anything'",
   "expected_stratum": "NEITHER",
   "variants": [
    "the site doesn't say",
    "there is no pricing page",
    "the site does not quote a per-seat price"
   ],
   "decoys": [
    "$12,000",
    "$500 per seat",
    "$15,000 plus a retainer",
    "it is free"
   ],
   "evidence": [],
   "note": "abstention scoring is irreducibly proxy in both directions: a model that abstains because it did not read carefully scores as honest, and a model that answers 'no per-seat anything' is saying something true about the page while not naming the absence. Stamped, never averaged into accuracy."
  },
  {
   "id": "B02",
   "category": "bait",
   "question": "What did the exhibit called 'The lantern trials' conclude?",
   "key": "ABSTAIN",
   "checker": "check_B02",
   "stamp": "proxy",
   "frozen": true,
   "published": "n/a",
   "source": "no such exhibit; the string 'lantern' appears nowhere in the corpus",
   "expected_stratum": "NEITHER",
   "variants": [
    "the site doesn't say",
    "there is no such exhibit",
    "no exhibit by that name"
   ],
   "decoys": [
    "it concluded the lanterns were brightest at dusk",
    "no model earned a seat",
    "it was inconclusive"
   ],
   "evidence": [],
   "note": ""
  },
  {
   "id": "B03",
   "category": "bait",
   "question": "Which URL is this site's RSS feed?",
   "key": "ABSTAIN",
   "checker": "check_B03",
   "stamp": "proxy",
   "frozen": true,
   "published": "n/a",
   "source": "no feed; /feed.xml and /rss.xml both 404 and neither map names one",
   "expected_stratum": "NEITHER",
   "variants": [
    "the site doesn't say",
    "there is no RSS feed"
   ],
   "decoys": [
    "https://research.strata2signal.com/feed.xml",
    "https://research.strata2signal.com/rss.xml",
    "https://research.strata2signal.com/data/index.json"
   ],
   "evidence": [],
   "note": "the strongest bait on the sheet: a plausible convention, and the site DOES publish a machine-readable index, so a confabulating arm has something adjacent to reach for."
  },
  {
   "id": "B04",
   "category": "bait",
   "question": "Which URL do I use to create an account so RuleSage saves my history?",
   "key": "ABSTAIN",
   "checker": "check_B04",
   "stamp": "proxy",
   "frozen": true,
   "published": "n/a",
   "source": "no accounts anywhere; the apex says 'no account, nothing to install'",
   "expected_stratum": "NEITHER",
   "variants": [
    "the site doesn't say",
    "there are no accounts",
    "RuleSage needs no account"
   ],
   "decoys": [
    "https://rulesage.strata2signal.com/signup",
    "https://strata2signal.com/login",
    "https://rulesage.strata2signal.com/account"
   ],
   "evidence": [],
   "note": ""
  }
 ]
}