{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-09-16-001",
    "source_story_id": "tmp-lead-cadworld-agent-benchmark",
    "edition_id": "mp-2026-09-16-morning-0069",
    "edition_url": "https://themachinepress.com/edition/2026-09-16",
    "position": 1,
    "story_type": "lead",
    "section": "benchmarks-evals",
    "editorial_classification": "editorial",
    "headline": "The Best CAD Agent Finished One Task in Six",
    "slug": "the-best-cad-agent-finished-one-task-in-six",
    "dek": "Across 200 FreeCAD tasks, the strongest tested computer-use agent passed 17.5%; expert reference work passed 87.0%.",
    "summary": "Across 200 FreeCAD tasks, the strongest tested computer-use agent passed 17.5%; expert reference work passed 87.0%.",
    "body_text": "Computer-use benchmarks often stop when an interface appears to accept the right clicks. CADWorld checks the persistent engineering artifact instead. Its 200 FreeCAD tasks span 11 workflows, from sketches and assemblies to CAM, finite-element analysis, meshes and technical drawings. Executable checks inspect geometry, dimensions, parametric structure, constraints and downstream manufacturing state. Across seven current agents, the strongest passed 17.5% of the full suite versus 87.0% for an expert reference. Stronger agents increasingly reached plausible-looking outputs but failed structural and construction-process requirements, exposing a gap between GUI fluency and reliable engineering work.",
    "why_it_matters": "Across 200 FreeCAD tasks, the strongest tested computer-use agent passed 17.5%; expert reference work passed 87.0%.",
    "limitations": [],
    "importance": 10,
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-16-001/the-best-cad-agent-finished-one-task-in-six",
    "json_url": "https://themachinepress.com/story/mp-2026-09-16-001.json",
    "first_published_at": "2026-09-16T09:00:00.000-04:00",
    "modified_at": "2026-09-16T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-09-16-001-001",
        "text": "Across 200 FreeCAD tasks, the strongest tested computer-use agent passed 17.5%; expert reference work passed 87.0%.",
        "source_ids": [
          "source-2026-09-16-001"
        ],
        "qualification": null
      }
    ],
    "source_ids": [
      "source-2026-09-16-001"
    ],
    "tags": [
      "computer-use agents",
      "CAD",
      "executable evaluation"
    ],
    "image_url": "https://themachinepress.com/issues/2026-09-16/lead-cadworld.png",
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-09-16-001",
      "title": "arXiv preprint 2609.16251",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2609.16251",
      "canonical_url": "https://arxiv.org/abs/2609.16251",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-09-14T20:00:00.000-04:00",
      "accessed_at": "2026-09-16T08:18:00.000-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-09-16-001-001"
      ]
    }
  ],
  "corrections": [],
  "publisher": {
    "name": "The Machine Press",
    "url": "https://themachinepress.com",
    "description": "A daily newspaper for the age of artificial intelligence."
  },
  "cite_this_report": {
    "title": "The Best CAD Agent Finished One Task in Six",
    "publisher": "The Machine Press",
    "published_at": "2026-09-16T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-16-001/the-best-cad-agent-finished-one-task-in-six"
  }
}