{
  "schema": "skillreducer/research-table/1",
  "id": "smaller-model-tasks",
  "study": "smaller-model",
  "title": "Cheaper on a smaller model, per task",
  "evidence_class": "measured-local",
  "date": "2026-09-20",
  "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590",
  "source_documents": [
    {
      "path": "docs/experiments/DETECTION-2026-09-19.md",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590",
      "committed": "2026-09-20"
    }
  ],
  "n_meaning": "Repetitions per arm",
  "columns": [
    {
      "name": "skill",
      "description": "Skill slug in the AirMarket catalog"
    },
    {
      "name": "task_id",
      "description": "Task generated from the skill"
    },
    {
      "name": "comparable",
      "description": "A default-model run with the skill exists for this task"
    },
    {
      "name": "score_haiku_skill",
      "description": "Grade, smaller model with the skill"
    },
    {
      "name": "score_haiku_none",
      "description": "Grade, smaller model without the skill"
    },
    {
      "name": "score_sonnet_skill",
      "description": "Grade, default model with the skill"
    },
    {
      "name": "quality_held",
      "description": "Smaller model within 0.05 of the default model"
    },
    {
      "name": "skill_lifts_haiku",
      "description": "The skill lifted the smaller model above no skill"
    },
    {
      "name": "cost_ratio",
      "description": "Smaller-model cost over default-model cost; empty when not comparable"
    },
    {
      "name": "cost_ratio_recorded",
      "description": "The ratio as recorded, 0 when not comparable"
    },
    {
      "name": "n",
      "description": "Repetitions per arm"
    },
    {
      "name": "evidence_class",
      "description": "Evidence class of the figure (SPEC.md section 20)"
    },
    {
      "name": "date",
      "description": "UTC day the study's last measured run landed"
    },
    {
      "name": "commit",
      "description": "Commit the figures come from: the committed data when the study has some, otherwise the latest commit of the document that reports it"
    }
  ],
  "rows": [
    {
      "skill": "ads-plan",
      "task_id": "fitness-app-media-plan",
      "comparable": true,
      "score_haiku_skill": 0.8,
      "score_haiku_none": 0.6,
      "score_sonnet_skill": 1,
      "quality_held": false,
      "skill_lifts_haiku": true,
      "cost_ratio": 0.859,
      "cost_ratio_recorded": 0.859,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "ads-plan",
      "task_id": "healthdash-optimization",
      "comparable": false,
      "score_haiku_skill": 1,
      "score_haiku_none": 0.8,
      "score_sonnet_skill": null,
      "quality_held": null,
      "skill_lifts_haiku": true,
      "cost_ratio": null,
      "cost_ratio_recorded": 0,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "alphagbm-tepper-signal",
      "task_id": "tepper-quality-filter-validation",
      "comparable": false,
      "score_haiku_skill": 0.8,
      "score_haiku_none": 0,
      "score_sonnet_skill": null,
      "quality_held": null,
      "skill_lifts_haiku": true,
      "cost_ratio": null,
      "cost_ratio_recorded": 0,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "alphagbm-tepper-signal",
      "task_id": "tepper-signal-multi-scenario",
      "comparable": true,
      "score_haiku_skill": 0.8,
      "score_haiku_none": 1,
      "score_sonnet_skill": 1,
      "quality_held": false,
      "skill_lifts_haiku": false,
      "cost_ratio": 1.466,
      "cost_ratio_recorded": 1.466,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "architecture-designer",
      "task_id": "ecommerce-microservices-design",
      "comparable": true,
      "score_haiku_skill": 1,
      "score_haiku_none": 0.6,
      "score_sonnet_skill": 1,
      "quality_held": true,
      "skill_lifts_haiku": true,
      "cost_ratio": 1.524,
      "cost_ratio_recorded": 1.524,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "architecture-designer",
      "task_id": "monolith-to-microservices-review",
      "comparable": false,
      "score_haiku_skill": 1,
      "score_haiku_none": 1,
      "score_sonnet_skill": null,
      "quality_held": null,
      "skill_lifts_haiku": false,
      "cost_ratio": null,
      "cost_ratio_recorded": 0,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "catchup",
      "task_id": "context-restoration-after-break",
      "comparable": true,
      "score_haiku_skill": 1,
      "score_haiku_none": 0.6,
      "score_sonnet_skill": 1,
      "quality_held": true,
      "skill_lifts_haiku": true,
      "cost_ratio": 0.658,
      "cost_ratio_recorded": 0.658,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "catchup",
      "task_id": "prioritize-work-from-state",
      "comparable": false,
      "score_haiku_skill": 0.8,
      "score_haiku_none": 0.8,
      "score_sonnet_skill": null,
      "quality_held": null,
      "skill_lifts_haiku": false,
      "cost_ratio": null,
      "cost_ratio_recorded": 0,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "caveman-commit",
      "task_id": "breaking-api-endpoint",
      "comparable": false,
      "score_haiku_skill": 0.75,
      "score_haiku_none": 0.75,
      "score_sonnet_skill": null,
      "quality_held": null,
      "skill_lifts_haiku": false,
      "cost_ratio": null,
      "cost_ratio_recorded": 0,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "caveman-commit",
      "task_id": "perf-query-opt",
      "comparable": true,
      "score_haiku_skill": 0.75,
      "score_haiku_none": 0.25,
      "score_sonnet_skill": 0.75,
      "quality_held": true,
      "skill_lifts_haiku": true,
      "cost_ratio": 0.842,
      "cost_ratio_recorded": 0.842,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "compose-modifier-and-layout-style",
      "task_id": "fix-composable-modifier-parameter",
      "comparable": true,
      "score_haiku_skill": 1,
      "score_haiku_none": 0.75,
      "score_sonnet_skill": 1,
      "quality_held": true,
      "skill_lifts_haiku": true,
      "cost_ratio": 0.6,
      "cost_ratio_recorded": 0.6,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "compose-modifier-and-layout-style",
      "task_id": "refactor-modifier-chain-style",
      "comparable": false,
      "score_haiku_skill": 1,
      "score_haiku_none": 1,
      "score_sonnet_skill": null,
      "quality_held": null,
      "skill_lifts_haiku": false,
      "cost_ratio": null,
      "cost_ratio_recorded": 0,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "cost-efficiency-analyzer",
      "task_id": "cost-efficiency-current-quarter",
      "comparable": true,
      "score_haiku_skill": 1,
      "score_haiku_none": 0.4,
      "score_sonnet_skill": 1,
      "quality_held": true,
      "skill_lifts_haiku": true,
      "cost_ratio": 0.748,
      "cost_ratio_recorded": 0.748,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "cost-efficiency-analyzer",
      "task_id": "cost-efficiency-trend-analysis",
      "comparable": false,
      "score_haiku_skill": 0.6,
      "score_haiku_none": 0.4,
      "score_sonnet_skill": null,
      "quality_held": null,
      "skill_lifts_haiku": true,
      "cost_ratio": null,
      "cost_ratio_recorded": 0,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "data-analyst",
      "task_id": "compare-campaign-performance",
      "comparable": true,
      "score_haiku_skill": 0.6,
      "score_haiku_none": 0.6,
      "score_sonnet_skill": 0.8,
      "quality_held": false,
      "skill_lifts_haiku": false,
      "cost_ratio": 0.55,
      "cost_ratio_recorded": 0.55,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "data-analyst",
      "task_id": "explore-customer-dataset",
      "comparable": true,
      "score_haiku_skill": 0.8,
      "score_haiku_none": 0.8,
      "score_sonnet_skill": 0.8,
      "quality_held": true,
      "skill_lifts_haiku": false,
      "cost_ratio": 0.714,
      "cost_ratio_recorded": 0.714,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "debugging-network-issues",
      "task_id": "isolate-slow-download-throughput",
      "comparable": true,
      "score_haiku_skill": 0.2,
      "score_haiku_none": 0.4,
      "score_sonnet_skill": 0.8,
      "quality_held": false,
      "skill_lifts_haiku": false,
      "cost_ratio": 0.314,
      "cost_ratio_recorded": 0.314,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "debugging-network-issues",
      "task_id": "timeout-root-cause-multi-layer",
      "comparable": true,
      "score_haiku_skill": 0,
      "score_haiku_none": 0.75,
      "score_sonnet_skill": 1,
      "quality_held": false,
      "skill_lifts_haiku": false,
      "cost_ratio": 0.376,
      "cost_ratio_recorded": 0.376,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "dotnet-local-tools",
      "task_id": "fix-broken-local-tools-config",
      "comparable": false,
      "score_haiku_skill": 1,
      "score_haiku_none": 0.5,
      "score_sonnet_skill": null,
      "quality_held": null,
      "skill_lifts_haiku": true,
      "cost_ratio": null,
      "cost_ratio_recorded": 0,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "dotnet-local-tools",
      "task_id": "setup-local-tools-fresh",
      "comparable": true,
      "score_haiku_skill": 1,
      "score_haiku_none": 0.5,
      "score_sonnet_skill": 1,
      "quality_held": true,
      "skill_lifts_haiku": true,
      "cost_ratio": 1.256,
      "cost_ratio_recorded": 1.256,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "eol-message",
      "task_id": "eol-message-feature-replacement",
      "comparable": true,
      "score_haiku_skill": 0.8,
      "score_haiku_none": 0.6,
      "score_sonnet_skill": 0.6,
      "quality_held": true,
      "skill_lifts_haiku": true,
      "cost_ratio": 0.751,
      "cost_ratio_recorded": 0.751,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "eol-message",
      "task_id": "eol-message-graceful-exit",
      "comparable": false,
      "score_haiku_skill": 0.8,
      "score_haiku_none": 0.2,
      "score_sonnet_skill": null,
      "quality_held": null,
      "skill_lifts_haiku": true,
      "cost_ratio": null,
      "cost_ratio_recorded": 0,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "event-store-design",
      "task_id": "event-store-implementation",
      "comparable": false,
      "score_haiku_skill": 1,
      "score_haiku_none": 1,
      "score_sonnet_skill": null,
      "quality_held": null,
      "skill_lifts_haiku": false,
      "cost_ratio": null,
      "cost_ratio_recorded": 0,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "event-store-design",
      "task_id": "event-store-system-design",
      "comparable": true,
      "score_haiku_skill": 1,
      "score_haiku_none": 0.8,
      "score_sonnet_skill": 1,
      "quality_held": true,
      "skill_lifts_haiku": true,
      "cost_ratio": 1.02,
      "cost_ratio_recorded": 1.02,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "feedback",
      "task_id": "extract-quiet-feedback",
      "comparable": false,
      "score_haiku_skill": 1,
      "score_haiku_none": 0.5,
      "score_sonnet_skill": null,
      "quality_held": null,
      "skill_lifts_haiku": true,
      "cost_ratio": null,
      "cost_ratio_recorded": 0,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "feedback",
      "task_id": "specific-feedback-sbi",
      "comparable": true,
      "score_haiku_skill": 0,
      "score_haiku_none": 0,
      "score_sonnet_skill": 1,
      "quality_held": false,
      "skill_lifts_haiku": false,
      "cost_ratio": 0.407,
      "cost_ratio_recorded": 0.407,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "friendly",
      "task_id": "dropdown-design-guidelines",
      "comparable": true,
      "score_haiku_skill": 0.6,
      "score_haiku_none": 0.4,
      "score_sonnet_skill": 0.8,
      "quality_held": false,
      "skill_lifts_haiku": true,
      "cost_ratio": 2.181,
      "cost_ratio_recorded": 2.181,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "friendly",
      "task_id": "improve-incomplete-button-guidelines",
      "comparable": false,
      "score_haiku_skill": 1,
      "score_haiku_none": 0.4,
      "score_sonnet_skill": null,
      "quality_held": null,
      "skill_lifts_haiku": true,
      "cost_ratio": null,
      "cost_ratio_recorded": 0,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "geo-fundamentals",
      "task_id": "audit-blog-posts",
      "comparable": true,
      "score_haiku_skill": 1,
      "score_haiku_none": 0.4,
      "score_sonnet_skill": 1,
      "quality_held": true,
      "skill_lifts_haiku": true,
      "cost_ratio": 0.799,
      "cost_ratio_recorded": 0.799,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "geo-fundamentals",
      "task_id": "improve-article-for-ai-citation",
      "comparable": false,
      "score_haiku_skill": 0.8,
      "score_haiku_none": 0.2,
      "score_sonnet_skill": null,
      "quality_held": null,
      "skill_lifts_haiku": true,
      "cost_ratio": null,
      "cost_ratio_recorded": 0,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "hermes-agent-skill-authoring",
      "task_id": "create-bundled-skill",
      "comparable": true,
      "score_haiku_skill": 0.8,
      "score_haiku_none": 0.6,
      "score_sonnet_skill": 0.8,
      "quality_held": true,
      "skill_lifts_haiku": true,
      "cost_ratio": 0.656,
      "cost_ratio_recorded": 0.656,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "hermes-agent-skill-authoring",
      "task_id": "fix-noncompliant-skill",
      "comparable": false,
      "score_haiku_skill": 0.8,
      "score_haiku_none": 0.2,
      "score_sonnet_skill": null,
      "quality_held": null,
      "skill_lifts_haiku": true,
      "cost_ratio": null,
      "cost_ratio_recorded": 0,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "landing-page-conversion-audit",
      "task_id": "checkout-bottleneck-diagnosis",
      "comparable": false,
      "score_haiku_skill": 0.8,
      "score_haiku_none": 0.6,
      "score_sonnet_skill": null,
      "quality_held": null,
      "skill_lifts_haiku": true,
      "cost_ratio": null,
      "cost_ratio_recorded": 0,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "landing-page-conversion-audit",
      "task_id": "landing-page-message-match-audit",
      "comparable": true,
      "score_haiku_skill": 1,
      "score_haiku_none": 0.4,
      "score_sonnet_skill": 1,
      "quality_held": true,
      "skill_lifts_haiku": true,
      "cost_ratio": 0.573,
      "cost_ratio_recorded": 0.573,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "latent-briefing",
      "task_id": "latent-briefing-config-tuning",
      "comparable": false,
      "score_haiku_skill": 1,
      "score_haiku_none": 1,
      "score_sonnet_skill": null,
      "quality_held": null,
      "skill_lifts_haiku": false,
      "cost_ratio": null,
      "cost_ratio_recorded": 0,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "latent-briefing",
      "task_id": "latent-briefing-design",
      "comparable": true,
      "score_haiku_skill": 1,
      "score_haiku_none": 0.6,
      "score_sonnet_skill": 1,
      "quality_held": true,
      "skill_lifts_haiku": true,
      "cost_ratio": 1.094,
      "cost_ratio_recorded": 1.094,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "llm-council",
      "task_id": "payment-error-refactor",
      "comparable": true,
      "score_haiku_skill": 0.25,
      "score_haiku_none": 0.25,
      "score_sonnet_skill": 0.25,
      "quality_held": true,
      "skill_lifts_haiku": false,
      "cost_ratio": 0.623,
      "cost_ratio_recorded": 0.623,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "llm-council",
      "task_id": "search-filter-plan",
      "comparable": true,
      "score_haiku_skill": 0.25,
      "score_haiku_none": 0.75,
      "score_sonnet_skill": 0,
      "quality_held": true,
      "skill_lifts_haiku": false,
      "cost_ratio": 0.461,
      "cost_ratio_recorded": 0.461,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "modeling-product-usage-metrics",
      "task_id": "metrics-map-product-questions",
      "comparable": false,
      "score_haiku_skill": 0.8,
      "score_haiku_none": 0,
      "score_sonnet_skill": null,
      "quality_held": null,
      "skill_lifts_haiku": true,
      "cost_ratio": null,
      "cost_ratio_recorded": 0,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "modeling-product-usage-metrics",
      "task_id": "retention-mobile-app-config",
      "comparable": true,
      "score_haiku_skill": 0.4,
      "score_haiku_none": 0.6,
      "score_sonnet_skill": 0.8,
      "quality_held": false,
      "skill_lifts_haiku": false,
      "cost_ratio": 0.652,
      "cost_ratio_recorded": 0.652,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "n8n-error-handling",
      "task_id": "scheduled-sync-retry-alert",
      "comparable": true,
      "score_haiku_skill": 0.75,
      "score_haiku_none": 0,
      "score_sonnet_skill": 1,
      "quality_held": false,
      "skill_lifts_haiku": true,
      "cost_ratio": 0.979,
      "cost_ratio_recorded": 0.979,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "n8n-error-handling",
      "task_id": "webhook-silent-error",
      "comparable": true,
      "score_haiku_skill": 1,
      "score_haiku_none": 1,
      "score_sonnet_skill": 1,
      "quality_held": true,
      "skill_lifts_haiku": false,
      "cost_ratio": 0.352,
      "cost_ratio_recorded": 0.352,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "napkin",
      "task_id": "activate-napkin-whiteboard",
      "comparable": true,
      "score_haiku_skill": 0.8,
      "score_haiku_none": 0,
      "score_sonnet_skill": 0.4,
      "quality_held": true,
      "skill_lifts_haiku": true,
      "cost_ratio": 0.541,
      "cost_ratio_recorded": 0.541,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "napkin",
      "task_id": "check-unsaved-napkin",
      "comparable": false,
      "score_haiku_skill": 0.8,
      "score_haiku_none": 0.4,
      "score_sonnet_skill": null,
      "quality_held": null,
      "skill_lifts_haiku": true,
      "cost_ratio": null,
      "cost_ratio_recorded": 0,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "offensive-reporting",
      "task_id": "format-raw-finding",
      "comparable": true,
      "score_haiku_skill": 0,
      "score_haiku_none": 0,
      "score_sonnet_skill": 0.6,
      "quality_held": false,
      "skill_lifts_haiku": false,
      "cost_ratio": 0.216,
      "cost_ratio_recorded": 0.216,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "offensive-reporting",
      "task_id": "write-executive-summary",
      "comparable": false,
      "score_haiku_skill": 0.8,
      "score_haiku_none": 0.8,
      "score_sonnet_skill": null,
      "quality_held": null,
      "skill_lifts_haiku": false,
      "cost_ratio": null,
      "cost_ratio_recorded": 0,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "optimize-simplicite-logs",
      "task_id": "find-and-analyze-errors-in-log",
      "comparable": true,
      "score_haiku_skill": 1,
      "score_haiku_none": 0.75,
      "score_sonnet_skill": 1,
      "quality_held": true,
      "skill_lifts_haiku": true,
      "cost_ratio": 0.776,
      "cost_ratio_recorded": 0.776,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "optimize-simplicite-logs",
      "task_id": "parse-and-filter-simplicite-log",
      "comparable": true,
      "score_haiku_skill": 1,
      "score_haiku_none": 1,
      "score_sonnet_skill": 1,
      "quality_held": true,
      "skill_lifts_haiku": false,
      "cost_ratio": 0.606,
      "cost_ratio_recorded": 0.606,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "product-vision",
      "task_id": "create-vision-from-scratch",
      "comparable": true,
      "score_haiku_skill": 0,
      "score_haiku_none": 0,
      "score_sonnet_skill": 0.8,
      "quality_held": false,
      "skill_lifts_haiku": false,
      "cost_ratio": 0.387,
      "cost_ratio_recorded": 0.387,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "product-vision",
      "task_id": "refine-flat-vision",
      "comparable": true,
      "score_haiku_skill": 1,
      "score_haiku_none": 0.8,
      "score_sonnet_skill": 0.8,
      "quality_held": true,
      "skill_lifts_haiku": true,
      "cost_ratio": 0.544,
      "cost_ratio_recorded": 0.544,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "research-agent",
      "task_id": "author-competitor-research-agent",
      "comparable": true,
      "score_haiku_skill": 1,
      "score_haiku_none": 0.4,
      "score_sonnet_skill": 1,
      "quality_held": true,
      "skill_lifts_haiku": true,
      "cost_ratio": 0.94,
      "cost_ratio_recorded": 0.94,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "research-agent",
      "task_id": "author-regulation-monitoring-agent",
      "comparable": false,
      "score_haiku_skill": 1,
      "score_haiku_none": 0.4,
      "score_sonnet_skill": null,
      "quality_held": null,
      "skill_lifts_haiku": true,
      "cost_ratio": null,
      "cost_ratio_recorded": 0,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "terracotta",
      "task_id": "fix-input-guidelines",
      "comparable": true,
      "score_haiku_skill": 0.6,
      "score_haiku_none": 0.4,
      "score_sonnet_skill": 0.6,
      "quality_held": true,
      "skill_lifts_haiku": true,
      "cost_ratio": 0.872,
      "cost_ratio_recorded": 0.872,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "terracotta",
      "task_id": "modal-guidelines",
      "comparable": true,
      "score_haiku_skill": 0.6,
      "score_haiku_none": 0.6,
      "score_sonnet_skill": 0.9,
      "quality_held": false,
      "skill_lifts_haiku": false,
      "cost_ratio": 2.545,
      "cost_ratio_recorded": 2.545,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "triton-kernel",
      "task_id": "fix-softmax-masking-bug",
      "comparable": false,
      "score_haiku_skill": 1,
      "score_haiku_none": 0.8,
      "score_sonnet_skill": null,
      "quality_held": null,
      "skill_lifts_haiku": true,
      "cost_ratio": null,
      "cost_ratio_recorded": 0,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "triton-kernel",
      "task_id": "fused-vector-op-kernel",
      "comparable": true,
      "score_haiku_skill": 0.8,
      "score_haiku_none": 0.8,
      "score_sonnet_skill": 1,
      "quality_held": false,
      "skill_lifts_haiku": false,
      "cost_ratio": 1.007,
      "cost_ratio_recorded": 1.007,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "typescript-server",
      "task_id": "product-inventory",
      "comparable": true,
      "score_haiku_skill": 1,
      "score_haiku_none": 0.2,
      "score_sonnet_skill": 0.8,
      "quality_held": true,
      "skill_lifts_haiku": true,
      "cost_ratio": 0.536,
      "cost_ratio_recorded": 0.536,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "typescript-server",
      "task_id": "scheduled-notifications",
      "comparable": false,
      "score_haiku_skill": 0.6,
      "score_haiku_none": 0.2,
      "score_sonnet_skill": null,
      "quality_held": null,
      "skill_lifts_haiku": true,
      "cost_ratio": null,
      "cost_ratio_recorded": 0,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "unity-primetween",
      "task_id": "generate-sequence-script",
      "comparable": false,
      "score_haiku_skill": 0.6,
      "score_haiku_none": 0.4,
      "score_sonnet_skill": null,
      "quality_held": null,
      "skill_lifts_haiku": true,
      "cost_ratio": null,
      "cost_ratio_recorded": 0,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "unity-primetween",
      "task_id": "generate-tween-script",
      "comparable": true,
      "score_haiku_skill": 1,
      "score_haiku_none": 0.8,
      "score_sonnet_skill": 1,
      "quality_held": true,
      "skill_lifts_haiku": true,
      "cost_ratio": 0.753,
      "cost_ratio_recorded": 0.753,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "unity-script-roles",
      "task_id": "collectible-system-architecture",
      "comparable": true,
      "score_haiku_skill": 1,
      "score_haiku_none": 0.6,
      "score_sonnet_skill": 1,
      "quality_held": true,
      "skill_lifts_haiku": true,
      "cost_ratio": 0.433,
      "cost_ratio_recorded": 0.433,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "unity-script-roles",
      "task_id": "quest-dialogue-system-architecture",
      "comparable": false,
      "score_haiku_skill": 1,
      "score_haiku_none": 0.8,
      "score_sonnet_skill": null,
      "quality_held": null,
      "skill_lifts_haiku": true,
      "cost_ratio": null,
      "cost_ratio_recorded": 0,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "vision-sft",
      "task_id": "vlm-config-technical-diagrams",
      "comparable": true,
      "score_haiku_skill": 0.8,
      "score_haiku_none": 0,
      "score_sonnet_skill": 0.4,
      "quality_held": true,
      "skill_lifts_haiku": true,
      "cost_ratio": 1.036,
      "cost_ratio_recorded": 1.036,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "vision-sft",
      "task_id": "vlm-debug-silent-killers",
      "comparable": false,
      "score_haiku_skill": 1,
      "score_haiku_none": 1,
      "score_sonnet_skill": null,
      "quality_held": null,
      "skill_lifts_haiku": false,
      "cost_ratio": null,
      "cost_ratio_recorded": 0,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "web-quality-audit",
      "task_id": "comprehensive-website-audit",
      "comparable": true,
      "score_haiku_skill": 1,
      "score_haiku_none": 0.8,
      "score_sonnet_skill": 1,
      "quality_held": true,
      "skill_lifts_haiku": true,
      "cost_ratio": 0.408,
      "cost_ratio_recorded": 0.408,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "web-quality-audit",
      "task_id": "launch-blocker-audit",
      "comparable": false,
      "score_haiku_skill": 1,
      "score_haiku_none": 0.8,
      "score_sonnet_skill": null,
      "quality_held": null,
      "skill_lifts_haiku": true,
      "cost_ratio": null,
      "cost_ratio_recorded": 0,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "wiki-lint",
      "task_id": "vault-audit-with-metadata",
      "comparable": false,
      "score_haiku_skill": 0.2,
      "score_haiku_none": 0.6,
      "score_sonnet_skill": null,
      "quality_held": null,
      "skill_lifts_haiku": false,
      "cost_ratio": null,
      "cost_ratio_recorded": 0,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    },
    {
      "skill": "wiki-lint",
      "task_id": "vault-health-check",
      "comparable": true,
      "score_haiku_skill": 0.5,
      "score_haiku_none": 1,
      "score_sonnet_skill": 1,
      "quality_held": false,
      "skill_lifts_haiku": false,
      "cost_ratio": 0.587,
      "cost_ratio_recorded": 0.587,
      "n": 3,
      "evidence_class": "measured-local",
      "date": "2026-09-20",
      "commit": "80e4a13fb96ebb96d8efff8690a8c77215657590"
    }
  ]
}
