{
 "name": "Whirrwork agent trust scorecard history",
 "url": "https://whirrwork.com/agents/trust/",
 "license": "CC BY 4.0",
 "caveat": "Self-assessment: Whirrwork wrote the task set. Single box, single region; not an uptime measure.",
 "weights": {
  "coverage": 2,
  "price_coverage_freshness": 1.5,
  "accuracy": 1.5,
  "provenance": 1,
  "reliability_latency": 1,
  "schema_stability": 0.75,
  "rfq_templates": 0.75,
  "honesty": 1.5,
  "agent_ergonomics": 1,
  "discoverability": 1,
  "robot_vs_human": 0.75
 },
 "runs": [
  {
   "run_at": "2026-10-25T13:44:33+00:00",
   "harness_version": "1.0.0",
   "taskset_version": "2026-10-25.1",
   "tasks": 136,
   "overall": 66.5,
   "scores": {
    "coverage": 24,
    "price_coverage_freshness": 29.6,
    "accuracy": 100,
    "provenance": 80,
    "reliability_latency": 100,
    "schema_stability": 71.4,
    "rfq_templates": 81.3,
    "honesty": 85.7,
    "agent_ergonomics": 92.9,
    "discoverability": 40,
    "rfq_outcomes": null
   },
   "notes": {
    "coverage": "live e2e 32.5% of 80 queries; relevance@3 78.8%; incl. worldwide 77.5%; dataset jobs x 36 markets local 11.3%",
    "price_coverage_freshness": "jobs with a price <=12 mo 35.1%; jobs with >=3 prices 16.9%; rows with day-precision as_of 0.0%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "250 calls, 0 unexpected failures, p50 59 ms, p95 186 ms (single run from one box; not an uptime measure)",
    "schema_stability": "failed: output_schemas_declared, versioned_path_or_header",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 20 of 308 jobs (many are category defaults); lifecycle test pass",
    "honesty": "failed: nonexistent-02, nonexistent-03, nonexistent-06",
    "agent_ergonomics": "failed: output_schemas",
    "discoverability": "own discovery files 9/9; listed in 0/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0"
   }
  },
  {
   "run_at": "2026-10-25T14:46:28+00:00",
   "harness_version": "1.0.0",
   "taskset_version": "2026-10-25.1",
   "tasks": 136,
   "overall": 66.5,
   "scores": {
    "coverage": 24,
    "price_coverage_freshness": 29.6,
    "accuracy": 100,
    "provenance": 80,
    "reliability_latency": 100,
    "schema_stability": 71.4,
    "rfq_templates": 81.3,
    "honesty": 85.7,
    "agent_ergonomics": 92.9,
    "discoverability": 40,
    "rfq_outcomes": null
   },
   "notes": {
    "coverage": "live e2e 32.5% of 80 queries; relevance@3 78.8%; incl. worldwide 77.5%; dataset jobs x 36 markets local 11.3%",
    "price_coverage_freshness": "jobs with a price <=12 mo 35.1%; jobs with >=3 prices 16.9%; rows with day-precision as_of 0.0%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "250 calls, 0 unexpected failures, p50 58 ms, p95 110 ms (single run from one box; not an uptime measure)",
    "schema_stability": "failed: output_schemas_declared, versioned_path_or_header",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 20 of 308 jobs (many are category defaults); lifecycle test pass",
    "honesty": "failed: nonexistent-02, nonexistent-03, nonexistent-06",
    "agent_ergonomics": "failed: output_schemas",
    "discoverability": "own discovery files 9/9; listed in 0/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0"
   }
  },
  {
   "run_at": "2026-10-25T14:58:49+00:00",
   "harness_version": "1.0.0",
   "taskset_version": "2026-10-25.1",
   "tasks": 136,
   "overall": 66.6,
   "scores": {
    "coverage": 24,
    "price_coverage_freshness": 29.7,
    "accuracy": 100,
    "provenance": 80.1,
    "reliability_latency": 100,
    "schema_stability": 71.4,
    "rfq_templates": 81.3,
    "honesty": 85.7,
    "agent_ergonomics": 92.9,
    "discoverability": 40,
    "rfq_outcomes": null
   },
   "notes": {
    "coverage": "live e2e 32.5% of 80 queries; relevance@3 78.8%; incl. worldwide 77.5%; dataset jobs x 36 markets local 11.3%",
    "price_coverage_freshness": "jobs with a price <=12 mo 35.1%; jobs with >=3 prices 17.2%; rows with day-precision as_of 0.4%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "250 calls, 0 unexpected failures, p50 59 ms, p95 119 ms (single run from one box; not an uptime measure)",
    "schema_stability": "failed: output_schemas_declared, versioned_path_or_header",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 20 of 308 jobs (many are category defaults); lifecycle test pass",
    "honesty": "failed: nonexistent-02, nonexistent-03, nonexistent-06",
    "agent_ergonomics": "failed: output_schemas",
    "discoverability": "own discovery files 9/9; listed in 0/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0"
   }
  },
  {
   "run_at": "2026-10-25T15:15:03+00:00",
   "harness_version": "1.0.0",
   "taskset_version": "2026-10-25.1",
   "tasks": 136,
   "overall": 69.1,
   "scores": {
    "coverage": 28.5,
    "price_coverage_freshness": 29.7,
    "accuracy": 100,
    "provenance": 80.1,
    "reliability_latency": 99.8,
    "schema_stability": 71.4,
    "rfq_templates": 81.3,
    "honesty": 100,
    "agent_ergonomics": 92.9,
    "discoverability": 40,
    "rfq_outcomes": null
   },
   "notes": {
    "coverage": "live e2e 40.0% of 80 queries; relevance@3 98.8%; incl. worldwide 96.2%; dataset jobs x 36 markets local 11.3%",
    "price_coverage_freshness": "jobs with a price <=12 mo 35.1%; jobs with >=3 prices 17.2%; rows with day-precision as_of 0.4%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "254 calls, 1 unexpected failures, p50 60 ms, p95 126 ms (single run from one box; not an uptime measure)",
    "schema_stability": "failed: output_schemas_declared, versioned_path_or_header",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 20 of 308 jobs (many are category defaults); lifecycle test pass",
    "honesty": "21 of 21 pass",
    "agent_ergonomics": "failed: output_schemas",
    "discoverability": "own discovery files 9/9; listed in 0/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0"
   }
  },
  {
   "run_at": "2026-10-25T15:43:25+00:00",
   "harness_version": "1.0.0",
   "taskset_version": "2026-10-25.1",
   "tasks": 136,
   "overall": 71.4,
   "scores": {
    "coverage": 29.2,
    "price_coverage_freshness": 31.7,
    "accuracy": 95.8,
    "provenance": 81.5,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 81.3,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 40,
    "rfq_outcomes": null
   },
   "notes": {
    "coverage": "live e2e 41.2% of 80 queries; relevance@3 100.0%; incl. worldwide 97.5%; dataset jobs x 36 markets local 11.3%",
    "price_coverage_freshness": "jobs with a price <=12 mo 37.0%; jobs with >=3 prices 19.2%; rows with day-precision as_of 7.7%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 91.7% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "255 calls, 0 unexpected failures, p50 61 ms, p95 123 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 20 of 308 jobs (many are category defaults); lifecycle test pass",
    "honesty": "21 of 21 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 0/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0"
   }
  },
  {
   "run_at": "2026-10-25T16:42:20+00:00",
   "harness_version": "1.0.1",
   "taskset_version": "2026-10-25.1",
   "tasks": 136,
   "overall": 72.2,
   "scores": {
    "coverage": 29.2,
    "price_coverage_freshness": 33,
    "accuracy": 100,
    "provenance": 82.6,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 81.3,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 40,
    "rfq_outcomes": null
   },
   "notes": {
    "coverage": "live e2e 41.2% of 80 queries; relevance@3 100.0%; incl. worldwide 97.5%; dataset jobs x 36 markets local 11.3%",
    "price_coverage_freshness": "jobs with a price <=12 mo 38.3%; jobs with >=3 prices 20.5%; rows with day-precision as_of 13.2%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "255 calls, 0 unexpected failures, p50 59 ms, p95 119 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 20 of 308 jobs (many are category defaults); lifecycle test pass",
    "honesty": "21 of 21 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 0/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0"
   }
  },
  {
   "run_at": "2026-10-25T17:00:47+00:00",
   "harness_version": "1.0.1",
   "taskset_version": "2026-10-25.1",
   "tasks": 136,
   "overall": 72.2,
   "scores": {
    "coverage": 29.2,
    "price_coverage_freshness": 33,
    "accuracy": 100,
    "provenance": 82.6,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 81.3,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 40,
    "rfq_outcomes": null
   },
   "notes": {
    "coverage": "live e2e 41.2% of 80 queries; relevance@3 100.0%; incl. worldwide 97.5%; dataset jobs x 36 markets local 11.3%",
    "price_coverage_freshness": "jobs with a price <=12 mo 38.3%; jobs with >=3 prices 20.5%; rows with day-precision as_of 13.2%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "255 calls, 0 unexpected failures, p50 60 ms, p95 129 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 20 of 308 jobs (many are category defaults); lifecycle test pass",
    "honesty": "21 of 21 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 0/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0"
   }
  },
  {
   "run_at": "2026-10-25T17:37:18+00:00",
   "harness_version": "1.0.1",
   "taskset_version": "2026-10-25.1",
   "tasks": 136,
   "overall": 73.4,
   "scores": {
    "coverage": 29.2,
    "price_coverage_freshness": 33,
    "accuracy": 100,
    "provenance": 82.6,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 40,
    "rfq_outcomes": null
   },
   "notes": {
    "coverage": "live e2e 41.2% of 80 queries; relevance@3 100.0%; incl. worldwide 97.5%; dataset jobs x 36 markets local 11.3%",
    "price_coverage_freshness": "jobs with a price <=12 mo 38.3%; jobs with >=3 prices 20.5%; rows with day-precision as_of 13.2%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "255 calls, 0 unexpected failures, p50 62 ms, p95 124 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs (many are category defaults); lifecycle test pass",
    "honesty": "21 of 21 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 0/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0"
   }
  },
  {
   "run_at": "2026-10-25T17:54:29+00:00",
   "harness_version": "1.0.1",
   "taskset_version": "2026-10-25.2",
   "tasks": 147,
   "overall": 74.9,
   "scores": {
    "coverage": 29.2,
    "price_coverage_freshness": 33,
    "accuracy": 100,
    "provenance": 82.6,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 40,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 41.2% of 80 queries; relevance@3 100.0%; incl. worldwide 97.5%; dataset jobs x 36 markets local 11.3%",
    "price_coverage_freshness": "jobs with a price <=12 mo 38.3%; jobs with >=3 prices 20.5%; rows with day-precision as_of 13.2%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "268 calls, 0 unexpected failures, p50 62 ms, p95 125 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "24 of 24 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 0/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 73.4
   }
  },
  {
   "run_at": "2026-10-25T18:04:41+00:00",
   "harness_version": "1.0.1",
   "taskset_version": "2026-10-25.2",
   "tasks": 147,
   "overall": 77.3,
   "scores": {
    "coverage": 29.2,
    "price_coverage_freshness": 33,
    "accuracy": 100,
    "provenance": 82.6,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 41.2% of 80 queries; relevance@3 100.0%; incl. worldwide 97.5%; dataset jobs x 36 markets local 11.3%",
    "price_coverage_freshness": "jobs with a price <=12 mo 38.3%; jobs with >=3 prices 20.5%; rows with day-precision as_of 13.2%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "268 calls, 0 unexpected failures, p50 61 ms, p95 131 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "24 of 24 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 75.9
   }
  },
  {
   "run_at": "2026-10-25T18:13:02+00:00",
   "harness_version": "1.0.1",
   "taskset_version": "2026-10-25.2",
   "tasks": 147,
   "overall": 77.3,
   "scores": {
    "coverage": 29.2,
    "price_coverage_freshness": 33,
    "accuracy": 100,
    "provenance": 82.6,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 41.2% of 80 queries; relevance@3 100.0%; incl. worldwide 97.5%; dataset jobs x 36 markets local 11.3%",
    "price_coverage_freshness": "jobs with a price <=12 mo 38.3%; jobs with >=3 prices 20.5%; rows with day-precision as_of 13.2%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "268 calls, 0 unexpected failures, p50 61 ms, p95 138 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "24 of 24 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 75.9
   }
  },
  {
   "run_at": "2026-10-25T18:42:15+00:00",
   "harness_version": "1.0.2",
   "taskset_version": "2026-10-25.2",
   "tasks": 147,
   "overall": 81.9,
   "scores": {
    "coverage": 56,
    "price_coverage_freshness": 35.6,
    "accuracy": 100,
    "provenance": 84.4,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 85.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 12.5%; v1.0.1-comparable coverage 51.5",
    "price_coverage_freshness": "jobs with a price <=12 mo 41.2%; jobs with >=3 prices 22.7%; rows with day-precision as_of 21.8%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "268 calls, 0 unexpected failures, p50 61 ms, p95 124 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "24 of 24 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 80.8
   }
  },
  {
   "run_at": "2026-10-25T18:48:53+00:00",
   "harness_version": "1.0.2",
   "taskset_version": "2026-10-25.2",
   "tasks": 147,
   "overall": 82.9,
   "scores": {
    "coverage": 61.6,
    "price_coverage_freshness": 35.9,
    "accuracy": 100,
    "provenance": 84.7,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 93.8% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 13.3%; v1.0.1-comparable coverage 60.8",
    "price_coverage_freshness": "jobs with a price <=12 mo 41.2%; jobs with >=3 prices 23.4%; rows with day-precision as_of 23.6%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "268 calls, 0 unexpected failures, p50 61 ms, p95 132 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "24 of 24 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 81.8
   }
  },
  {
   "run_at": "2026-10-25T19:00:05+00:00",
   "harness_version": "1.0.2",
   "taskset_version": "2026-10-25.2",
   "tasks": 147,
   "overall": 82.4,
   "scores": {
    "coverage": 61.6,
    "price_coverage_freshness": 36,
    "accuracy": 95.5,
    "provenance": 85.2,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 93.8% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 13.3%; v1.0.1-comparable coverage 60.8",
    "price_coverage_freshness": "jobs with a price <=12 mo 41.2%; jobs with >=3 prices 24.0%; rows with day-precision as_of 26.0%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 90.9% of 11 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "268 calls, 0 unexpected failures, p50 61 ms, p95 141 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "24 of 24 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 81.3
   }
  },
  {
   "run_at": "2026-10-25T19:07:56+00:00",
   "harness_version": "1.0.2",
   "taskset_version": "2026-10-25.2",
   "tasks": 147,
   "overall": 81.9,
   "scores": {
    "coverage": 61.6,
    "price_coverage_freshness": 36,
    "accuracy": 90.9,
    "provenance": 85.4,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 93.8% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 13.3%; v1.0.1-comparable coverage 60.8",
    "price_coverage_freshness": "jobs with a price <=12 mo 41.2%; jobs with >=3 prices 24.0%; rows with day-precision as_of 27.0%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 81.8% of 11 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "268 calls, 0 unexpected failures, p50 60 ms, p95 123 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "24 of 24 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 80.7
   }
  },
  {
   "run_at": "2026-10-25T19:17:05+00:00",
   "harness_version": "1.0.2",
   "taskset_version": "2026-10-25.2",
   "tasks": 147,
   "overall": 81.9,
   "scores": {
    "coverage": 61.6,
    "price_coverage_freshness": 36,
    "accuracy": 90.9,
    "provenance": 85.4,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 93.8% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 13.3%; v1.0.1-comparable coverage 60.8",
    "price_coverage_freshness": "jobs with a price <=12 mo 41.2%; jobs with >=3 prices 24.0%; rows with day-precision as_of 27.0%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 81.8% of 11 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "268 calls, 0 unexpected failures, p50 62 ms, p95 141 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "24 of 24 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 80.7
   }
  },
  {
   "run_at": "2026-10-25T19:23:23+00:00",
   "harness_version": "1.0.2",
   "taskset_version": "2026-10-25.2",
   "tasks": 147,
   "overall": 83,
   "scores": {
    "coverage": 61.6,
    "price_coverage_freshness": 36.4,
    "accuracy": 100,
    "provenance": 85.8,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 93.8% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 13.3%; v1.0.1-comparable coverage 60.8",
    "price_coverage_freshness": "jobs with a price <=12 mo 41.2%; jobs with >=3 prices 25.3%; rows with day-precision as_of 28.8%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "268 calls, 0 unexpected failures, p50 63 ms, p95 149 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "24 of 24 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 82
   }
  },
  {
   "run_at": "2026-10-25T19:38:30+00:00",
   "harness_version": "1.0.2",
   "taskset_version": "2026-10-25.2",
   "tasks": 147,
   "overall": 83,
   "scores": {
    "coverage": 61.6,
    "price_coverage_freshness": 36.4,
    "accuracy": 100,
    "provenance": 85.8,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 93.8% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 13.3%; v1.0.1-comparable coverage 60.8",
    "price_coverage_freshness": "jobs with a price <=12 mo 41.2%; jobs with >=3 prices 25.3%; rows with day-precision as_of 28.8%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "268 calls, 0 unexpected failures, p50 50 ms, p95 143 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "24 of 24 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 82
   }
  },
  {
   "run_at": "2026-10-25T19:45:38+00:00",
   "harness_version": "1.0.2",
   "taskset_version": "2026-10-25.2",
   "tasks": 147,
   "overall": 83.6,
   "scores": {
    "coverage": 61.6,
    "price_coverage_freshness": 40.3,
    "accuracy": 100,
    "provenance": 87,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 93.8% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 13.3%; v1.0.1-comparable coverage 60.8",
    "price_coverage_freshness": "jobs with a price <=12 mo 45.8%; jobs with >=3 prices 27.6%; rows with day-precision as_of 35.0%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "268 calls, 0 unexpected failures, p50 60 ms, p95 132 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "24 of 24 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 82.6
   }
  },
  {
   "run_at": "2026-10-25T20:31:16+00:00",
   "harness_version": "1.0.2",
   "taskset_version": "2026-10-25.2",
   "tasks": 147,
   "overall": 84.6,
   "scores": {
    "coverage": 61.6,
    "price_coverage_freshness": 47.5,
    "accuracy": 100,
    "provenance": 88.7,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 93.8% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 13.3%; v1.0.1-comparable coverage 60.8",
    "price_coverage_freshness": "jobs with a price <=12 mo 55.8%; jobs with >=3 prices 28.2%; rows with day-precision as_of 43.3%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 11 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "268 calls, 0 unexpected failures, p50 62 ms, p95 162 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "24 of 24 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 83.6
   }
  },
  {
   "run_at": "2026-10-25T21:44:05+00:00",
   "harness_version": "1.0.2",
   "taskset_version": "2026-10-25.2",
   "tasks": 147,
   "overall": 84.6,
   "scores": {
    "coverage": 61.6,
    "price_coverage_freshness": 47.5,
    "accuracy": 100,
    "provenance": 88.7,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 93.8% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 13.3%; v1.0.1-comparable coverage 60.8",
    "price_coverage_freshness": "jobs with a price <=12 mo 55.8%; jobs with >=3 prices 28.2%; rows with day-precision as_of 43.3%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 11 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "268 calls, 0 unexpected failures, p50 63 ms, p95 131 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "24 of 24 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 83.6
   }
  },
  {
   "run_at": "2026-10-25T22:36:23+00:00",
   "harness_version": "1.0.2",
   "taskset_version": "2026-10-25.2",
   "tasks": 147,
   "overall": 85,
   "scores": {
    "coverage": 61.6,
    "price_coverage_freshness": 50.8,
    "accuracy": 100,
    "provenance": 89.5,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 93.8% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 13.3%; v1.0.1-comparable coverage 60.8",
    "price_coverage_freshness": "jobs with a price <=12 mo 60.1%; jobs with >=3 prices 29.2%; rows with day-precision as_of 47.4%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 11 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "268 calls, 0 unexpected failures, p50 62 ms, p95 150 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "24 of 24 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 84.1
   }
  },
  {
   "run_at": "2026-10-25T23:37:32+00:00",
   "harness_version": "1.0.2",
   "taskset_version": "2026-10-25.2",
   "tasks": 147,
   "overall": 85,
   "scores": {
    "coverage": 61.6,
    "price_coverage_freshness": 50.8,
    "accuracy": 100,
    "provenance": 89.5,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 93.8% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 13.3%; v1.0.1-comparable coverage 60.8",
    "price_coverage_freshness": "jobs with a price <=12 mo 60.1%; jobs with >=3 prices 29.2%; rows with day-precision as_of 47.4%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 11 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "268 calls, 0 unexpected failures, p50 65 ms, p95 162 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "24 of 24 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 84.1
   }
  },
  {
   "run_at": "2026-10-26T00:33:29+00:00",
   "harness_version": "1.0.2",
   "taskset_version": "2026-10-25.2",
   "tasks": 147,
   "overall": 85.5,
   "scores": {
    "coverage": 61.6,
    "price_coverage_freshness": 54.6,
    "accuracy": 100,
    "provenance": 90.1,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 93.8% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 13.3%; v1.0.1-comparable coverage 60.8",
    "price_coverage_freshness": "jobs with a price <=12 mo 64.3%; jobs with >=3 prices 32.1%; rows with day-precision as_of 50.7%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "268 calls, 0 unexpected failures, p50 61 ms, p95 126 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "24 of 24 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 84.6
   }
  },
  {
   "run_at": "2026-10-26T01:40:21+00:00",
   "harness_version": "1.0.2",
   "taskset_version": "2026-10-25.2",
   "tasks": 147,
   "overall": 85.5,
   "scores": {
    "coverage": 61.6,
    "price_coverage_freshness": 54.6,
    "accuracy": 100,
    "provenance": 90.1,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 93.8% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 13.3%; v1.0.1-comparable coverage 60.8",
    "price_coverage_freshness": "jobs with a price <=12 mo 64.3%; jobs with >=3 prices 32.1%; rows with day-precision as_of 50.7%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "268 calls, 0 unexpected failures, p50 61 ms, p95 166 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "24 of 24 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 84.6
   }
  },
  {
   "run_at": "2026-10-26T03:36:31+00:00",
   "harness_version": "1.0.2",
   "taskset_version": "2026-10-25.2",
   "tasks": 147,
   "overall": 86.2,
   "scores": {
    "coverage": 61.6,
    "price_coverage_freshness": 59.8,
    "accuracy": 100,
    "provenance": 91,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 93.8% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 13.3%; v1.0.1-comparable coverage 60.8",
    "price_coverage_freshness": "jobs with a price <=12 mo 70.1%; jobs with >=3 prices 35.7%; rows with day-precision as_of 55.2%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 11 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "268 calls, 0 unexpected failures, p50 62 ms, p95 146 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "24 of 24 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 85.3
   }
  },
  {
   "run_at": "2026-10-26T04:41:28+00:00",
   "harness_version": "1.0.2",
   "taskset_version": "2026-10-25.2",
   "tasks": 147,
   "overall": 85.6,
   "scores": {
    "coverage": 61.6,
    "price_coverage_freshness": 59.8,
    "accuracy": 95,
    "provenance": 91,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 93.8% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 13.3%; v1.0.1-comparable coverage 60.8",
    "price_coverage_freshness": "jobs with a price <=12 mo 70.1%; jobs with >=3 prices 35.7%; rows with day-precision as_of 55.2%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 90.0% of 10 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "268 calls, 0 unexpected failures, p50 48 ms, p95 174 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "24 of 24 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 84.7
   }
  },
  {
   "run_at": "2026-10-26T05:35:29+00:00",
   "harness_version": "1.0.3",
   "taskset_version": "2026-10-25.2",
   "tasks": 147,
   "overall": 86.6,
   "scores": {
    "coverage": 61.6,
    "price_coverage_freshness": 62.6,
    "accuracy": 100,
    "provenance": 91.5,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 93.8% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 13.3%; v1.0.1-comparable coverage 60.8",
    "price_coverage_freshness": "jobs with a price <=12 mo 73.7%; jobs with >=3 prices 36.7%; rows with day-precision as_of 57.3%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 11 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "268 calls, 0 unexpected failures, p50 65 ms, p95 134 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "24 of 24 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 85.7
   }
  },
  {
   "run_at": "2026-10-26T05:38:50+00:00",
   "harness_version": "1.0.4",
   "taskset_version": "2026-10-25.2",
   "tasks": 147,
   "overall": 86.6,
   "scores": {
    "coverage": 61.6,
    "price_coverage_freshness": 62.6,
    "accuracy": 100,
    "provenance": 91.5,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 93.8% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 13.3%; v1.0.1-comparable coverage 60.8",
    "price_coverage_freshness": "jobs with a price <=12 mo 73.7%; jobs with >=3 prices 36.7%; rows with day-precision as_of 57.3%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "268 calls, 0 unexpected failures, p50 61 ms, p95 139 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "24 of 24 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 85.7
   }
  },
  {
   "run_at": "2026-10-26T06:45:28+00:00",
   "harness_version": "1.0.4",
   "taskset_version": "2026-10-25.2",
   "tasks": 147,
   "overall": 86.6,
   "scores": {
    "coverage": 61.6,
    "price_coverage_freshness": 62.6,
    "accuracy": 100,
    "provenance": 91.5,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 93.8% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 13.3%; v1.0.1-comparable coverage 60.8",
    "price_coverage_freshness": "jobs with a price <=12 mo 73.7%; jobs with >=3 prices 36.7%; rows with day-precision as_of 57.3%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 11 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "268 calls, 0 unexpected failures, p50 52 ms, p95 141 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "24 of 24 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 85.7
   }
  },
  {
   "run_at": "2026-10-26T07:39:10+00:00",
   "harness_version": "1.0.4",
   "taskset_version": "2026-10-25.2",
   "tasks": 147,
   "overall": 87.6,
   "scores": {
    "coverage": 68,
    "price_coverage_freshness": 62.8,
    "accuracy": 100,
    "provenance": 91.5,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 20.0%; v1.0.1-comparable coverage 67.3",
    "price_coverage_freshness": "jobs with a price <=12 mo 74.0%; jobs with >=3 prices 36.7%; rows with day-precision as_of 57.3%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 11 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "268 calls, 0 unexpected failures, p50 67 ms, p95 197 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "24 of 24 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 86.8
   }
  },
  {
   "run_at": "2026-10-26T08:40:27+00:00",
   "harness_version": "1.0.4",
   "taskset_version": "2026-10-25.2",
   "tasks": 147,
   "overall": 87.6,
   "scores": {
    "coverage": 68,
    "price_coverage_freshness": 62.8,
    "accuracy": 100,
    "provenance": 91.5,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 20.0%; v1.0.1-comparable coverage 68.0",
    "price_coverage_freshness": "jobs with a price <=12 mo 74.0%; jobs with >=3 prices 36.7%; rows with day-precision as_of 57.3%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 11 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "268 calls, 0 unexpected failures, p50 63 ms, p95 186 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "24 of 24 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 86.8
   }
  },
  {
   "run_at": "2026-10-26T09:38:07+00:00",
   "harness_version": "1.0.4",
   "taskset_version": "2026-10-25.2",
   "tasks": 147,
   "overall": 87.9,
   "scores": {
    "coverage": 69.9,
    "price_coverage_freshness": 62.8,
    "accuracy": 100,
    "provenance": 91.5,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 24.7%; v1.0.1-comparable coverage 69.9",
    "price_coverage_freshness": "jobs with a price <=12 mo 74.0%; jobs with >=3 prices 36.7%; rows with day-precision as_of 57.7%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 9 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "268 calls, 0 unexpected failures, p50 64 ms, p95 251 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "24 of 24 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 87.1
   }
  },
  {
   "run_at": "2026-10-26T09:44:54+00:00",
   "harness_version": "1.0.4",
   "taskset_version": "2026-10-25.2",
   "tasks": 147,
   "overall": 87.9,
   "scores": {
    "coverage": 69.9,
    "price_coverage_freshness": 63.1,
    "accuracy": 100,
    "provenance": 91.6,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 24.7%; v1.0.1-comparable coverage 69.9",
    "price_coverage_freshness": "jobs with a price <=12 mo 74.4%; jobs with >=3 prices 36.7%; rows with day-precision as_of 57.8%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 11 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "268 calls, 0 unexpected failures, p50 66 ms, p95 216 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "24 of 24 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 87.2
   }
  },
  {
   "run_at": "2026-10-26T10:43:39+00:00",
   "harness_version": "1.0.4",
   "taskset_version": "2026-10-25.2",
   "tasks": 147,
   "overall": 87.9,
   "scores": {
    "coverage": 69.9,
    "price_coverage_freshness": 63.1,
    "accuracy": 100,
    "provenance": 91.6,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 24.7%; v1.0.1-comparable coverage 69.9",
    "price_coverage_freshness": "jobs with a price <=12 mo 74.4%; jobs with >=3 prices 36.7%; rows with day-precision as_of 57.8%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 11 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "268 calls, 0 unexpected failures, p50 67 ms, p95 204 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "24 of 24 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 87.2
   }
  },
  {
   "run_at": "2026-10-26T10:54:25+00:00",
   "harness_version": "1.0.4",
   "taskset_version": "2026-10-25.2",
   "tasks": 147,
   "overall": 88,
   "scores": {
    "coverage": 70.1,
    "price_coverage_freshness": 63.1,
    "accuracy": 100,
    "provenance": 91.6,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 25.3%; v1.0.1-comparable coverage 70.1",
    "price_coverage_freshness": "jobs with a price <=12 mo 74.4%; jobs with >=3 prices 36.7%; rows with day-precision as_of 57.8%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "268 calls, 0 unexpected failures, p50 60 ms, p95 162 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "24 of 24 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 87.2
   }
  },
  {
   "run_at": "2026-10-26T11:01:25+00:00",
   "harness_version": "1.0.4",
   "taskset_version": "2026-10-26.1",
   "tasks": 153,
   "overall": 88,
   "scores": {
    "coverage": 70.1,
    "price_coverage_freshness": 63.1,
    "accuracy": 100,
    "provenance": 91.6,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 25.3%; v1.0.1-comparable coverage 70.1",
    "price_coverage_freshness": "jobs with a price <=12 mo 74.4%; jobs with >=3 prices 36.7%; rows with day-precision as_of 57.8%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 11 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "274 calls, 0 unexpected failures, p50 60 ms, p95 171 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "30 of 30 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 87.2
   }
  },
  {
   "run_at": "2026-10-26T11:05:15+00:00",
   "harness_version": "1.0.4",
   "taskset_version": "2026-10-26.1",
   "tasks": 153,
   "overall": 88,
   "scores": {
    "coverage": 70.3,
    "price_coverage_freshness": 63.1,
    "accuracy": 100,
    "provenance": 91.6,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 25.7%; v1.0.1-comparable coverage 70.3",
    "price_coverage_freshness": "jobs with a price <=12 mo 74.4%; jobs with >=3 prices 36.7%; rows with day-precision as_of 57.8%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 10 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "274 calls, 0 unexpected failures, p50 63 ms, p95 212 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "30 of 30 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 87.2
   }
  },
  {
   "run_at": "2026-10-26T11:10:44+00:00",
   "harness_version": "1.0.4",
   "taskset_version": "2026-10-26.1",
   "tasks": 153,
   "overall": 88,
   "scores": {
    "coverage": 70.3,
    "price_coverage_freshness": 63.1,
    "accuracy": 100,
    "provenance": 91.6,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 25.7%; v1.0.1-comparable coverage 70.3",
    "price_coverage_freshness": "jobs with a price <=12 mo 74.4%; jobs with >=3 prices 36.7%; rows with day-precision as_of 57.8%",
    "accuracy": "providers ok 100.0% of 12 verifiable; price numbers found on source 100.0% of 10 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "274 calls, 0 unexpected failures, p50 69 ms, p95 216 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "30 of 30 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 87.2
   }
  },
  {
   "run_at": "2026-10-26T11:13:49+00:00",
   "harness_version": "1.0.4",
   "taskset_version": "2026-10-26.1",
   "tasks": 153,
   "overall": 88,
   "scores": {
    "coverage": 70.5,
    "price_coverage_freshness": 63.1,
    "accuracy": 100,
    "provenance": 91.6,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 26.2%; v1.0.1-comparable coverage 70.5",
    "price_coverage_freshness": "jobs with a price <=12 mo 74.4%; jobs with >=3 prices 36.7%; rows with day-precision as_of 57.8%",
    "accuracy": "providers ok 100.0% of 11 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "274 calls, 0 unexpected failures, p50 60 ms, p95 197 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "30 of 30 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 87.3
   }
  },
  {
   "run_at": "2026-10-26T11:29:15+00:00",
   "harness_version": "1.0.4",
   "taskset_version": "2026-10-26.1",
   "tasks": 153,
   "overall": 88.1,
   "scores": {
    "coverage": 71,
    "price_coverage_freshness": 63.1,
    "accuracy": 100,
    "provenance": 91.6,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 27.4%; v1.0.1-comparable coverage 71.0",
    "price_coverage_freshness": "jobs with a price <=12 mo 74.4%; jobs with >=3 prices 36.7%; rows with day-precision as_of 58.0%",
    "accuracy": "providers ok 100.0% of 10 verifiable; price numbers found on source 100.0% of 11 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "274 calls, 0 unexpected failures, p50 64 ms, p95 187 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "30 of 30 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 87.4
   }
  },
  {
   "run_at": "2026-10-26T12:37:34+00:00",
   "harness_version": "1.0.4",
   "taskset_version": "2026-10-26.1",
   "tasks": 153,
   "overall": 88.4,
   "scores": {
    "coverage": 71,
    "price_coverage_freshness": 65.4,
    "accuracy": 100,
    "provenance": 91.8,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 27.4%; v1.0.1-comparable coverage 71.0",
    "price_coverage_freshness": "jobs with a price <=12 mo 77.6%; jobs with >=3 prices 37.0%; rows with day-precision as_of 59.1%",
    "accuracy": "providers ok 100.0% of 10 verifiable; price numbers found on source 100.0% of 11 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "274 calls, 0 unexpected failures, p50 67 ms, p95 229 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "30 of 30 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 87.7
   }
  },
  {
   "run_at": "2026-10-26T22:19:31+00:00",
   "harness_version": "1.0.4",
   "taskset_version": "2026-10-26.1",
   "tasks": 153,
   "overall": 88.4,
   "scores": {
    "coverage": 71,
    "price_coverage_freshness": 65.4,
    "accuracy": 100,
    "provenance": 91.8,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 27.4%; v1.0.1-comparable coverage 71.0",
    "price_coverage_freshness": "jobs with a price <=12 mo 77.6%; jobs with >=3 prices 37.0%; rows with day-precision as_of 59.1%",
    "accuracy": "providers ok 100.0% of 10 verifiable; price numbers found on source 100.0% of 11 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "274 calls, 0 unexpected failures, p50 69 ms, p95 227 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "30 of 30 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 87.7
   }
  },
  {
   "run_at": "2026-10-26T22:49:17+00:00",
   "harness_version": "1.0.4",
   "taskset_version": "2026-10-26.1",
   "tasks": 153,
   "overall": 88.4,
   "scores": {
    "coverage": 71,
    "price_coverage_freshness": 65.4,
    "accuracy": 100,
    "provenance": 91.8,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 27.4%; v1.0.1-comparable coverage 71.0",
    "price_coverage_freshness": "jobs with a price <=12 mo 77.6%; jobs with >=3 prices 37.0%; rows with day-precision as_of 59.1%",
    "accuracy": "providers ok 100.0% of 10 verifiable; price numbers found on source 100.0% of 11 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "274 calls, 0 unexpected failures, p50 69 ms, p95 223 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "30 of 30 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 87.7
   }
  },
  {
   "run_at": "2026-10-26T23:35:50+00:00",
   "harness_version": "1.0.4",
   "taskset_version": "2026-10-26.1",
   "tasks": 153,
   "overall": 88.1,
   "scores": {
    "coverage": 71,
    "price_coverage_freshness": 67.2,
    "accuracy": 95.8,
    "provenance": 92.1,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 27.4%; v1.0.1-comparable coverage 71.0",
    "price_coverage_freshness": "jobs with a price <=12 mo 80.2%; jobs with >=3 prices 37.0%; rows with day-precision as_of 60.5%",
    "accuracy": "providers ok 100.0% of 11 verifiable; price numbers found on source 91.7% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "274 calls, 0 unexpected failures, p50 68 ms, p95 216 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "30 of 30 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 87.4
   }
  },
  {
   "run_at": "2026-10-26T23:39:17+00:00",
   "harness_version": "1.0.5",
   "taskset_version": "2026-10-26.1",
   "tasks": 153,
   "overall": 88.6,
   "scores": {
    "coverage": 71,
    "price_coverage_freshness": 67.2,
    "accuracy": 100,
    "provenance": 92.1,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 27.4%; v1.0.1-comparable coverage 71.0",
    "price_coverage_freshness": "jobs with a price <=12 mo 80.2%; jobs with >=3 prices 37.0%; rows with day-precision as_of 60.5%",
    "accuracy": "providers ok 100.0% of 10 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "274 calls, 0 unexpected failures, p50 69 ms, p95 189 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "30 of 30 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 87.9
   }
  },
  {
   "run_at": "2026-10-27T00:49:58+00:00",
   "harness_version": "1.0.5",
   "taskset_version": "2026-10-26.1",
   "tasks": 153,
   "overall": 88.6,
   "scores": {
    "coverage": 71,
    "price_coverage_freshness": 67.2,
    "accuracy": 100,
    "provenance": 92.1,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 27.4%; v1.0.1-comparable coverage 71.0",
    "price_coverage_freshness": "jobs with a price <=12 mo 80.2%; jobs with >=3 prices 37.0%; rows with day-precision as_of 60.5%",
    "accuracy": "providers ok 100.0% of 10 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "274 calls, 0 unexpected failures, p50 69 ms, p95 224 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "30 of 30 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 87.9
   }
  },
  {
   "run_at": "2026-10-27T01:35:00+00:00",
   "harness_version": "1.0.5",
   "taskset_version": "2026-10-26.1",
   "tasks": 153,
   "overall": 88.7,
   "scores": {
    "coverage": 71,
    "price_coverage_freshness": 67.9,
    "accuracy": 100,
    "provenance": 92.5,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 27.4%; v1.0.1-comparable coverage 71.0",
    "price_coverage_freshness": "jobs with a price <=12 mo 81.2%; jobs with >=3 prices 37.0%; rows with day-precision as_of 62.4%",
    "accuracy": "providers ok 100.0% of 11 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "274 calls, 0 unexpected failures, p50 65 ms, p95 263 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "30 of 30 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 88
   }
  },
  {
   "run_at": "2026-10-27T02:47:15+00:00",
   "harness_version": "1.0.6",
   "taskset_version": "2026-10-27.1",
   "tasks": 164,
   "overall": 88.7,
   "scores": {
    "coverage": 71,
    "price_coverage_freshness": 67.9,
    "accuracy": 100,
    "provenance": 92.5,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 27.4%; v1.0.1-comparable coverage 71.0",
    "price_coverage_freshness": "jobs with a price <=12 mo 81.2%; jobs with >=3 prices 37.0%; rows with day-precision as_of 62.4%",
    "accuracy": "providers ok 100.0% of 11 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "285 calls, 0 unexpected failures, p50 70 ms, p95 310 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "41 of 41 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 88
   }
  },
  {
   "run_at": "2026-10-27T03:31:31+00:00",
   "harness_version": "1.0.6",
   "taskset_version": "2026-10-27.1",
   "tasks": 164,
   "overall": 88.9,
   "scores": {
    "coverage": 71,
    "price_coverage_freshness": 68.8,
    "accuracy": 100,
    "provenance": 92.8,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 27.4%; v1.0.1-comparable coverage 71.0",
    "price_coverage_freshness": "jobs with a price <=12 mo 82.1%; jobs with >=3 prices 37.7%; rows with day-precision as_of 63.8%",
    "accuracy": "providers ok 100.0% of 11 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "285 calls, 0 unexpected failures, p50 63 ms, p95 185 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "41 of 41 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 88.2
   }
  },
  {
   "run_at": "2026-10-27T04:49:36+00:00",
   "harness_version": "1.0.6",
   "taskset_version": "2026-10-27.1",
   "tasks": 164,
   "overall": 88.4,
   "scores": {
    "coverage": 71,
    "price_coverage_freshness": 68.8,
    "accuracy": 95.8,
    "provenance": 92.8,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 27.4%; v1.0.1-comparable coverage 71.0",
    "price_coverage_freshness": "jobs with a price <=12 mo 82.1%; jobs with >=3 prices 37.7%; rows with day-precision as_of 63.8%",
    "accuracy": "providers ok 100.0% of 10 verifiable; price numbers found on source 91.7% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "285 calls, 0 unexpected failures, p50 59 ms, p95 149 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "41 of 41 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 87.6
   }
  },
  {
   "run_at": "2026-10-27T04:55:15+00:00",
   "harness_version": "1.0.7",
   "taskset_version": "2026-10-27.1",
   "tasks": 164,
   "overall": 88.9,
   "scores": {
    "coverage": 71,
    "price_coverage_freshness": 68.8,
    "accuracy": 100,
    "provenance": 92.8,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 27.4%; v1.0.1-comparable coverage 71.0",
    "price_coverage_freshness": "jobs with a price <=12 mo 82.1%; jobs with >=3 prices 37.7%; rows with day-precision as_of 63.8%",
    "accuracy": "providers ok 100.0% of 10 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "285 calls, 0 unexpected failures, p50 61 ms, p95 189 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "41 of 41 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 88.2
   }
  },
  {
   "run_at": "2026-10-27T05:37:22+00:00",
   "harness_version": "1.0.7",
   "taskset_version": "2026-10-27.1",
   "tasks": 164,
   "overall": 89.1,
   "scores": {
    "coverage": 71,
    "price_coverage_freshness": 70.4,
    "accuracy": 100,
    "provenance": 92.9,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 27.4%; v1.0.1-comparable coverage 71.0",
    "price_coverage_freshness": "jobs with a price <=12 mo 84.4%; jobs with >=3 prices 37.7%; rows with day-precision as_of 64.6%",
    "accuracy": "providers ok 100.0% of 10 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "285 calls, 0 unexpected failures, p50 37 ms, p95 191 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "41 of 41 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 88.4
   }
  },
  {
   "run_at": "2026-10-27T06:39:44+00:00",
   "harness_version": "1.0.7",
   "taskset_version": "2026-10-27.1",
   "tasks": 164,
   "overall": 89.1,
   "scores": {
    "coverage": 71,
    "price_coverage_freshness": 70.4,
    "accuracy": 100,
    "provenance": 93,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 27.4%; v1.0.1-comparable coverage 71.0",
    "price_coverage_freshness": "jobs with a price <=12 mo 84.4%; jobs with >=3 prices 37.7%; rows with day-precision as_of 64.8%",
    "accuracy": "providers ok 100.0% of 10 verifiable; price numbers found on source 100.0% of 11 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "285 calls, 0 unexpected failures, p50 61 ms, p95 180 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "41 of 41 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 88.4
   }
  },
  {
   "run_at": "2026-10-27T07:43:29+00:00",
   "harness_version": "1.0.7",
   "taskset_version": "2026-10-27.1",
   "tasks": 164,
   "overall": 88.7,
   "scores": {
    "coverage": 71,
    "price_coverage_freshness": 71.2,
    "accuracy": 95.8,
    "provenance": 93.3,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 27.4%; v1.0.1-comparable coverage 71.0",
    "price_coverage_freshness": "jobs with a price <=12 mo 85.4%; jobs with >=3 prices 38.0%; rows with day-precision as_of 66.6%",
    "accuracy": "providers ok 100.0% of 10 verifiable; price numbers found on source 91.7% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "285 calls, 0 unexpected failures, p50 67 ms, p95 237 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "41 of 41 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 88
   }
  },
  {
   "run_at": "2026-10-27T08:42:33+00:00",
   "harness_version": "1.0.8",
   "taskset_version": "2026-10-27.1",
   "tasks": 164,
   "overall": 89.2,
   "scores": {
    "coverage": 71,
    "price_coverage_freshness": 71.2,
    "accuracy": 100,
    "provenance": 93.3,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 27.4%; v1.0.1-comparable coverage 71.0",
    "price_coverage_freshness": "jobs with a price <=12 mo 85.4%; jobs with >=3 prices 38.0%; rows with day-precision as_of 66.6%",
    "accuracy": "providers ok 100.0% of 10 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "285 calls, 0 unexpected failures, p50 69 ms, p95 252 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "41 of 41 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 88.5
   }
  },
  {
   "run_at": "2026-10-27T09:32:27+00:00",
   "harness_version": "1.0.8",
   "taskset_version": "2026-10-27.1",
   "tasks": 164,
   "overall": 89.4,
   "scores": {
    "coverage": 71,
    "price_coverage_freshness": 72.8,
    "accuracy": 100,
    "provenance": 93.4,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 27.5%; v1.0.1-comparable coverage 71.0",
    "price_coverage_freshness": "jobs with a price <=12 mo 87.7%; jobs with >=3 prices 38.0%; rows with day-precision as_of 67.0%",
    "accuracy": "providers ok 100.0% of 10 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "285 calls, 0 unexpected failures, p50 61 ms, p95 172 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "41 of 41 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 88.7
   }
  },
  {
   "run_at": "2026-10-27T09:44:15+00:00",
   "harness_version": "1.0.8",
   "taskset_version": "2026-10-27.1",
   "tasks": 164,
   "overall": 89.5,
   "scores": {
    "coverage": 71,
    "price_coverage_freshness": 73.8,
    "accuracy": 100,
    "provenance": 93.4,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 27.5%; v1.0.1-comparable coverage 71.0",
    "price_coverage_freshness": "jobs with a price <=12 mo 89.0%; jobs with >=3 prices 38.3%; rows with day-precision as_of 67.2%",
    "accuracy": "providers ok 100.0% of 10 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "285 calls, 0 unexpected failures, p50 57 ms, p95 176 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "41 of 41 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 88.8
   }
  },
  {
   "run_at": "2026-10-27T10:37:45+00:00",
   "harness_version": "1.0.8",
   "taskset_version": "2026-10-27.1",
   "tasks": 164,
   "overall": 89.5,
   "scores": {
    "coverage": 71,
    "price_coverage_freshness": 73.8,
    "accuracy": 100,
    "provenance": 93.4,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 27.5%; v1.0.1-comparable coverage 71.0",
    "price_coverage_freshness": "jobs with a price <=12 mo 89.0%; jobs with >=3 prices 38.3%; rows with day-precision as_of 67.2%",
    "accuracy": "providers ok 100.0% of 10 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "285 calls, 0 unexpected failures, p50 60 ms, p95 168 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "41 of 41 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 88.8
   }
  },
  {
   "run_at": "2026-10-27T11:32:13+00:00",
   "harness_version": "1.0.8",
   "taskset_version": "2026-10-27.1",
   "tasks": 164,
   "overall": 89.8,
   "scores": {
    "coverage": 71,
    "price_coverage_freshness": 76.6,
    "accuracy": 100,
    "provenance": 93.6,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 27.5%; v1.0.1-comparable coverage 71.0",
    "price_coverage_freshness": "jobs with a price <=12 mo 92.9%; jobs with >=3 prices 38.6%; rows with day-precision as_of 67.9%",
    "accuracy": "providers ok 100.0% of 10 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "285 calls, 0 unexpected failures, p50 67 ms, p95 271 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "41 of 41 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 89.2
   }
  },
  {
   "run_at": "2026-10-27T13:51:22+00:00",
   "harness_version": "1.0.8",
   "taskset_version": "2026-10-27.1",
   "tasks": 164,
   "overall": 90.1,
   "scores": {
    "coverage": 71,
    "price_coverage_freshness": 78.3,
    "accuracy": 100,
    "provenance": 93.7,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 27.5%; v1.0.1-comparable coverage 71.0",
    "price_coverage_freshness": "jobs with a price <=12 mo 95.1%; jobs with >=3 prices 39.0%; rows with day-precision as_of 68.3%",
    "accuracy": "providers ok 100.0% of 10 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "285 calls, 0 unexpected failures, p50 68 ms, p95 251 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "41 of 41 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 89.4
   }
  },
  {
   "run_at": "2026-10-27T15:40:50+00:00",
   "harness_version": "1.0.8",
   "taskset_version": "2026-10-27.1",
   "tasks": 164,
   "overall": 90.2,
   "scores": {
    "coverage": 71,
    "price_coverage_freshness": 79.9,
    "accuracy": 100,
    "provenance": 93.8,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 27.5%; v1.0.1-comparable coverage 71.0",
    "price_coverage_freshness": "jobs with a price <=12 mo 96.8%; jobs with >=3 prices 40.6%; rows with day-precision as_of 68.9%",
    "accuracy": "providers ok 100.0% of 10 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "285 calls, 0 unexpected failures, p50 74 ms, p95 347 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "41 of 41 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 89.6
   }
  },
  {
   "run_at": "2026-10-27T16:39:06+00:00",
   "harness_version": "1.0.8",
   "taskset_version": "2026-10-27.1",
   "tasks": 164,
   "overall": 90.2,
   "scores": {
    "coverage": 71,
    "price_coverage_freshness": 79.9,
    "accuracy": 100,
    "provenance": 93.8,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 27.5%; v1.0.1-comparable coverage 71.0",
    "price_coverage_freshness": "jobs with a price <=12 mo 96.8%; jobs with >=3 prices 40.6%; rows with day-precision as_of 69.2%",
    "accuracy": "providers ok 100.0% of 10 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "285 calls, 0 unexpected failures, p50 75 ms, p95 265 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "41 of 41 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 89.6
   }
  },
  {
   "run_at": "2026-10-27T17:36:38+00:00",
   "harness_version": "1.0.8",
   "taskset_version": "2026-10-27.1",
   "tasks": 163,
   "overall": 90.3,
   "scores": {
    "coverage": 71,
    "price_coverage_freshness": 79.9,
    "accuracy": 100,
    "provenance": 93.9,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 27.5%; v1.0.1-comparable coverage 71.0",
    "price_coverage_freshness": "jobs with a price <=12 mo 96.8%; jobs with >=3 prices 40.6%; rows with day-precision as_of 69.6%",
    "accuracy": "providers ok 100.0% of 10 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "284 calls, 0 unexpected failures, p50 72 ms, p95 233 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "40 of 40 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 89.6
   }
  },
  {
   "run_at": "2026-10-27T18:38:01+00:00",
   "harness_version": "1.0.8",
   "taskset_version": "2026-10-27.1",
   "tasks": 163,
   "overall": 90.3,
   "scores": {
    "coverage": 71,
    "price_coverage_freshness": 79.9,
    "accuracy": 100,
    "provenance": 93.9,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 27.5%; v1.0.1-comparable coverage 71.0",
    "price_coverage_freshness": "jobs with a price <=12 mo 96.8%; jobs with >=3 prices 40.6%; rows with day-precision as_of 69.6%",
    "accuracy": "providers ok 100.0% of 10 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "284 calls, 0 unexpected failures, p50 67 ms, p95 226 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "40 of 40 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 89.6
   }
  },
  {
   "run_at": "2026-10-27T19:34:41+00:00",
   "harness_version": "1.0.8",
   "taskset_version": "2026-10-27.1",
   "tasks": 163,
   "overall": 89.8,
   "scores": {
    "coverage": 71,
    "price_coverage_freshness": 80.4,
    "accuracy": 95.8,
    "provenance": 93.9,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 27.6%; v1.0.1-comparable coverage 71.0",
    "price_coverage_freshness": "jobs with a price <=12 mo 97.4%; jobs with >=3 prices 40.6%; rows with day-precision as_of 69.7%",
    "accuracy": "providers ok 100.0% of 10 verifiable; price numbers found on source 91.7% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "284 calls, 0 unexpected failures, p50 69 ms, p95 222 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "40 of 40 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 89.2
   }
  },
  {
   "run_at": "2026-10-27T19:42:58+00:00",
   "harness_version": "1.0.8",
   "taskset_version": "2026-10-27.1",
   "tasks": 163,
   "overall": 90.3,
   "scores": {
    "coverage": 71,
    "price_coverage_freshness": 80.4,
    "accuracy": 100,
    "provenance": 93.9,
    "reliability_latency": 100,
    "schema_stability": 100,
    "rfq_templates": 100,
    "honesty": 100,
    "agent_ergonomics": 100,
    "discoverability": 70,
    "rfq_outcomes": null,
    "robot_vs_human": 100
   },
   "notes": {
    "coverage": "live e2e 100.0% of 80 queries; relevance@3 100.0%; incl. worldwide 100.0%; dataset jobs x 36 markets local 27.6%; v1.0.1-comparable coverage 71.0",
    "price_coverage_freshness": "jobs with a price <=12 mo 97.4%; jobs with >=3 prices 40.6%; rows with day-precision as_of 69.7%",
    "accuracy": "providers ok 100.0% of 10 verifiable; price numbers found on source 100.0% of 12 verifiable; coverage-label inconsistencies 0",
    "provenance": "price rows sourced+dated 100.0%, providers sourced+dated 100.0%; penalty for month-only as_of (cannot tell retrieval vs publication date)",
    "reliability_latency": "284 calls, 0 unexpected failures, p50 67 ms, p95 281 ms (single run from one box; not an uptime measure)",
    "schema_stability": "all checks pass",
    "rfq_templates": "templates with >=4 job fields + example 100.0%; distinct field sets 308 of 308 jobs; lifecycle test pass",
    "honesty": "40 of 40 pass",
    "agent_ergonomics": "all checks pass",
    "discoverability": "own discovery files 9/9; listed in 1/2 machine-checked registries; 15 more directories need a manual check",
    "rfq_outcomes": "not available yet: RFQs are matched but not forwarded, so there are no quotes-received / response-time / job-completed signals. Real RFQs on file: 0",
    "robot_vs_human": "8/8 MCP comparisons sourced, labelled and with a crossover range; REST jobs ok, REST compare ok",
    "overall_excluding_robot_vs_human": 89.7
   }
  }
 ]
}