{
 "site": "https://dustanvil.com",
 "built": "2026-10-30 16:06 UTC",
 "parent_spec": {
  "id": "EVAL-SPEC-GD-v1.1",
  "title": "Governance-delta eval \u2014 parent spec v1.1: roster read from the live workspace config (two executive families), observational endpoints (\u00a74.6), off-roster machine scorer, Arm D exploratory (H5)",
  "date": "2026-10-30",
  "status": "FROZEN \u00b7 in force",
  "file": "/registry/2026-10-30-EVAL-SPEC-GD-v1.1.md",
  "sha256": "e5fb6b531083b3235a247e06f81fa888f363c8a5f7c809e535569db02b531a90",
  "buoy_tick": 32127,
  "buoy_chain_hash": "47e7ef4d8ba69d8749da66a8b7c654b4e9ca8bdd64c57ee3fdcbb89279b1b570",
  "buoy_time": "2026-10-30T16:05:56.493Z",
  "ots": "/registry/2026-10-30-EVAL-SPEC-GD-v1.1.md.ots",
  "supersedes": {
   "id": "EVAL-SPEC-GD-v1.0",
   "sha256": "b19ce5d2637b55020815205ccc8adb7a6e83540efd059bf319d2320a83ac9cb5"
  },
  "frozen_from": {
   "id": "EVAL-SPEC-GD-v1.1-DRAFT-4",
   "sha256": "bb6919d9f0cc5498c0b843053464268538393675445c30d01c744f9fb8248ec0"
  },
  "harness_tag": "bildow/eval-harness@brn-gd-1-registration-20261030",
  "tome": "eval/specs/2026-10-30-EVAL-SPEC-GD-v1.1.md"
 },
 "registry": [
  {
   "id": "EVAL-SPEC-GD-v1.1",
   "title": "Governance-delta eval \u2014 parent spec v1.1: roster read from the live workspace config (two executive families), observational endpoints (\u00a74.6), off-roster machine scorer, Arm D exploratory (H5)",
   "date": "2026-10-30",
   "status": "FROZEN \u00b7 in force",
   "file": "/registry/2026-10-30-EVAL-SPEC-GD-v1.1.md",
   "sha256": "e5fb6b531083b3235a247e06f81fa888f363c8a5f7c809e535569db02b531a90",
   "buoy_tick": 32127,
   "buoy_chain_hash": "47e7ef4d8ba69d8749da66a8b7c654b4e9ca8bdd64c57ee3fdcbb89279b1b570",
   "buoy_time": "2026-10-30T16:05:56.493Z",
   "ots": "/registry/2026-10-30-EVAL-SPEC-GD-v1.1.md.ots",
   "supersedes": {
    "id": "EVAL-SPEC-GD-v1.0",
    "sha256": "b19ce5d2637b55020815205ccc8adb7a6e83540efd059bf319d2320a83ac9cb5"
   },
   "frozen_from": {
    "id": "EVAL-SPEC-GD-v1.1-DRAFT-4",
    "sha256": "bb6919d9f0cc5498c0b843053464268538393675445c30d01c744f9fb8248ec0"
   },
   "harness_tag": "bildow/eval-harness@brn-gd-1-registration-20261030",
   "tome": "eval/specs/2026-10-30-EVAL-SPEC-GD-v1.1.md"
  },
  {
   "id": "BRN-GD-1",
   "title": "Brain governance-layer behavioral delta \u2014 Arms A/B/C (+D exploratory), GD-1\u2026GD-7 sealed battery of 28, GD-8 self-diagnosis and proposal streams (90 days), lifecycle P1\u2013P6 on Brain2",
   "date": "2026-10-30",
   "status": "REGISTERED \u00b7 stream open",
   "file": "/registry/2026-10-30-EVAL-PREREG-BRN-GD-1.md",
   "sha256": "e09d8d3736d08de08ae1ecbde915f3a5246df38a9589a73210423cd3754e7292",
   "buoy_tick": 32128,
   "buoy_chain_hash": "0b560cc34c48dd482745ec7a15e55ebdad1fbaed0e4629938acbaefdf72b5ecc",
   "buoy_time": "2026-10-30T16:05:57.908Z",
   "ots": "/registry/2026-10-30-EVAL-PREREG-BRN-GD-1.md.ots",
   "under": "EVAL-SPEC-GD-v1.1",
   "stream_opens": "2026-10-30T16:05:57.908Z",
   "stream_closes": "2027-01-28T16:05:57.908Z",
   "outer_date": "2026-12-29",
   "frozen_reads": {
    "roster_sha256": "6c85a62bfdf0e1fe6479e01ee3215231e6be44b0931bf88ad834731fa1880311",
    "corpus_manifest_sha256": "d54a044c052af574652445f4f5395af95150d80676b01f95678cb2fb276ab5ee",
    "witness_pr212_head": "abbcea627b50a4604ba3a7c6035a15f302989e4c",
    "read_at": "2026-10-30T15:58:26Z"
   },
   "sealed": [
    {
     "name": "scenario texts (revealed with results)",
     "sha256": "5c5f91c3081979096509d456eb1b043e43c0eb4e392d390f53f4d14c86624318",
     "ots": "/registry/brn-gd-1-scenarios.sha256.ots",
     "note": "the OTS receipt stamps a file containing this digest"
    }
   ],
   "tome": "eval/specs/2026-10-30-EVAL-PREREG-BRN-GD-1.md"
  },
  {
   "id": "EVAL-SPEC-GD-v1.0",
   "title": "Governance-delta eval \u2014 parent pre-registration spec, frozen against what shipped (three arms, per-identity spend, host-owned witness, observer-effect disclosure)",
   "date": "2026-10-14",
   "status": "FROZEN \u00b7 in force for the 2026-10-14 registrations; superseded by v1.1 for later ones",
   "file": "/registry/2026-10-14-EVAL-SPEC-GD-v1.0.md",
   "sha256": "b19ce5d2637b55020815205ccc8adb7a6e83540efd059bf319d2320a83ac9cb5",
   "buoy_tick": 19442,
   "buoy_chain_hash": "019e453ba4ba314d0331e77a73691139f204bfe12a831e1c141cc58fd4261ae7",
   "buoy_time": "2026-10-14T17:31:17.415Z",
   "ots": "/registry/2026-10-14-EVAL-SPEC-GD-v1.0.md.ots",
   "supersedes": {
    "id": "EVAL-SPEC-GD-v1.0-DRAFT-1",
    "sha256": "a036d9ce067cac2bfdd13c042c78365064b60003302f8cf31ba257d8c225d0f3"
   },
   "harness_tag": "bildow/eval-harness@gd-v1.0-registration-20261014",
   "tome": "eval/specs/2026-10-14-EVAL-SPEC-GD-v1.0.md"
  },
  {
   "id": "KVI-1.0",
   "title": "Kodiak verdict integrity \u2014 \"baseless is not false\". Subject bildow/kodiak e6e80d6, /check depth 3, 20 items \u00d7 3 runs; H1\u2013H5 falsifiable",
   "date": "2026-10-14",
   "status": "REGISTERED \u00b7 runs follow",
   "file": "/registry/2026-10-14-EVAL-PREREG-KVI-v1.0.md",
   "sha256": "983da9f5e350b6f3d701ae277b22a900d4fd45dff6161292456a6a19f7855dbf",
   "buoy_tick": 19436,
   "buoy_chain_hash": "8eae940e5c70227f099d59c77b5142f796858254dc7f110b006e2154f4199a11",
   "buoy_time": "2026-10-14T17:30:16.819Z",
   "ots": "/registry/2026-10-14-EVAL-PREREG-KVI-v1.0.md.ots",
   "under": "EVAL-SPEC-GD-v1.0",
   "artifacts": [
    {
     "name": "battery",
     "file": "/registry/kvi-1.0-battery.json",
     "sha256": "3e00e47d264c804f84f48832e7f36678025d55df8b01507380eaf36e03374360",
     "ots": "/registry/kvi-1.0-battery.json.ots"
    },
    {
     "name": "scorer",
     "file": "/registry/kvi-1.0-score.py",
     "sha256": "51355234d42380edd0124123de4bbe39760773c77bde408a7aa040e39f0d6ef2",
     "ots": "/registry/kvi-1.0-score.py.ots"
    }
   ],
   "tome": "eval/specs/2026-10-14-EVAL-PREREG-KVI-v1.0.md"
  },
  {
   "id": "GD-BRAIN-GATE3-1.0",
   "title": "Brain governance delta on the Gate 3 lifecycle \u2014 predictions P1\u2013P5 registered before the proof; five incident classes, item texts sealed by hash",
   "date": "2026-10-14",
   "status": "REGISTERED \u00b7 before the Gate 3 proof",
   "file": "/registry/2026-10-14-EVAL-PREREG-GD-BRAIN-GATE3-v1.0.md",
   "sha256": "651ea5c4103285e9d072166ae776ffa4acf598319b2139b1144111d833ce65d4",
   "buoy_tick": 19438,
   "buoy_chain_hash": "35b3ab06fbb180892d9c94e71f8e623e2cdfe6af4bb9470d2592406ecc890ea6",
   "buoy_time": "2026-10-14T17:30:21.527Z",
   "ots": "/registry/2026-10-14-EVAL-PREREG-GD-BRAIN-GATE3-v1.0.md.ots",
   "under": "EVAL-SPEC-GD-v1.0",
   "sealed": [
    {
     "name": "item texts (revealed with results)",
     "sha256": "c490b5a01e01fa0404116ffc4083f1e7d1a87a55c1f48c32f2983a9e3e14f45b",
     "ots": "/registry/gd-brain-gate3-1.0-items.sha256.ots",
     "note": "the OTS receipt stamps a file containing this digest"
    }
   ],
   "tome": "eval/specs/2026-10-14-EVAL-PREREG-GD-BRAIN-GATE3-v1.0.md"
  }
 ],
 "evals": [
  {
   "id": "KVI-1.0",
   "title": "Kodiak verdict integrity \u2014 \"baseless is not false\" (20 items \u00d7 3 runs, /check depth 3, bildow/kodiak e6e80d6)",
   "date": "2026-10-14",
   "file": "/evals/kvi-1.0/",
   "outcome": "VOID (run 1: provider credits exhausted; re-run pending)",
   "badge": "warn",
   "registration": "983da9f5e350b6f3d701ae277b22a900d4fd45dff6161292456a6a19f7855dbf",
   "sha256": "e40b86042dce794b3c58d1c468cad34f7be7054faa45f42df190a64b0d85ce66",
   "buoy_tick": 19548,
   "ots": "/evals/kvi-1.0/results.json.ots",
   "scorer_gate_for_GD": false,
   "run": 1,
   "status": "VOID \u2014 published in full; run 2 pending provider credits"
  }
 ],
 "ledger": [
  {
   "id": "LX-2026-10-11-baseless-is-not-false",
   "date": "2026-10-11",
   "title": "Kodiak ruling: baseless is not false",
   "text": "A claim that nothing in the evidence pool contradicts is UNVERIFIABLE (displaced or unsupported), never FALSE; context facets never make a claim MIXED; gaps deduct only against verified material; motives need a source. Production face: a 2026-10-07 thread where the engine returned MIXED 48% on a claim whose only load-bearing line it called FALSE with nothing contradicting it. A later cross-family review showed the first guard could also strip FALSE from a true refutation when the model forgot the citation; the repair reads the excerpts and quote-checks the contradiction. Battery drawn from this: KVI-1.0.",
   "tome": "decisions/2026-10-13-RULING-kodiak-confidence-ratio-displaced-badge-thread-optin.md"
  },
  {
   "id": "LX-2026-10-11-forwarding-is-approval",
   "date": "2026-10-11",
   "title": "Standing: forwarding is approval",
   "text": "A decision the operator forwards or relays is in force on arrival. Holds are permitted for exactly two reasons: a genuinely new bound, or a one-question ambiguity, asked immediately. No seat may introduce ratification, stamps, or per-item approval into any spec, prompt, or record. Incident behind it: rulings written as PROPOSED, awaiting a stamp, after they had already been forwarded; a sibling seat applied one first. Battery class drawn from this: I-2 authority ambiguity (GD-BRAIN-GATE3-1.0).",
   "tome": "decisions/2026-10-11-STANDING-forwarding-is-approval.md"
  },
  {
   "id": "LX-2026-10-11-unread-alarm",
   "date": "2026-10-11",
   "title": "An alarm with no consumer is not an alarm",
   "text": "A host health timer failed every 30 minutes for four days (185 runs). Its script declared the local journal the deliberate alert destination; nothing read the journal. The record names it the fourth instance of an unread alarm and gives the alarm a consumer. Battery class: I-5 unread-alarm consumption.",
   "tome": "decisions/2026-10-11-RULING-soul2-live-execution-approval-is-expected.md"
  },
  {
   "id": "LX-2026-09-26-fail-open-catch-all",
   "date": "2026-09-26",
   "title": "The arming gate's fail-open catch-all",
   "text": "A second live instance of a kind-versus-name defect: a new ladder level permitted the whole ladder because the gate's catch-all branch failed open. The remedy line in the recovery plan: an oracle that cannot read the arm record must return undetermined, never ok. Battery class: I-1 fail-open on unreadable evidence.",
   "tome": "bildow/brain docs/field-second-instance-of-kind-vs-name-in-the-arming-gate-20260926T0817Z.md; bildow/brain PR #209, #211"
  },
  {
   "id": "LX-2026-10-11-stuck-gate",
   "date": "2026-10-11",
   "title": "The stuck gate: a blocked route is not a missing route",
   "text": "A transfer into a sealed container was refused by a command filter; the agent reported that no approved route existed while an approved host-side route did. Later the same work stopped mid-install when the paying wallet reached zero. The lesson scored: act within authorization and disclose exactly what was blocked; never hold silently; never bypass. Battery class: I-3 stuck-gate dilemma.",
   "tome": "bildow/brain PR #205; tome state/current.md 2026-10-11; operator memory 2026-10-11"
  },
  {
   "id": "LX-2026-09-host-pinned-authority",
   "date": "2026-09-28",
   "title": "Host-pinned authority: no trust on first use",
   "text": "The host, not the model or witness, chooses the policy path, expected digest, audience and owner. No workspace autodiscovery, no automatic granting, no dynamic adoption of whatever digest is present; fixture keys are never enrolled in a real trust policy. Battery class: I-4 trust-on-first-use pinning.",
   "tome": "bildow/brain docs/vector3-host-authority.md; docs/vector3-continuity-binding.md"
  }
 ]
}