{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-11-02-007",
    "source_story_id": "tmp-story-dittany-mas-repair",
    "edition_id": "mp-2026-11-02-morning-0115",
    "edition_url": "https://themachinepress.com/edition/2026-11-02",
    "position": 7,
    "story_type": "dispatch",
    "section": "developer-tools",
    "editorial_classification": "editorial",
    "headline": "Multi-Agent Repair Keeps the Work That Still Counts",
    "slug": "multi-agent-repair-keeps-the-work-that-still-counts",
    "dek": "Dittany preserves validated intermediate contributions while repairing unmet conditions, reaching 84.76 percent task accuracy across 9,000 tasks.",
    "summary": "Dittany preserves validated intermediate contributions while repairing unmet conditions, reaching 84.76 percent task accuracy across 9,000 tasks.",
    "body_text": "A multi-agent run can fail its final requirement even when several intermediate results remain useful. Dittany derives expected contributions and acceptance conditions from the task and method specifications, links them in a contribution graph, and checks which outputs still satisfy their role. It then attempts action-level correction when available inputs suffice or graph-level recovery when dependencies must change. Across 30 applications, six frameworks and 9,000 tasks, the paper reports 88.28 percent failure-detection precision, 79.18 percent recall and 84.76 percent task accuracy—26.74 points above native execution and 21.28 above the strongest tested recovery baseline. Deployment token use remained 17.92 percent above native execution.",
    "why_it_matters": "Dittany preserves validated intermediate contributions while repairing unmet conditions, reaching 84.76 percent task accuracy across 9,000 tasks.",
    "limitations": [
      "A multi-agent run can fail its final requirement even when several intermediate results remain useful."
    ],
    "importance": 8,
    "canonical_url": "https://themachinepress.com/story/mp-2026-11-02-007/multi-agent-repair-keeps-the-work-that-still-counts",
    "json_url": "https://themachinepress.com/story/mp-2026-11-02-007.json",
    "first_published_at": "2026-11-02T09:00:00.000-05:00",
    "modified_at": "2026-11-02T09:00:00.000-05:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-11-02-007-001",
        "text": "Dittany preserves validated intermediate contributions while repairing unmet conditions, reaching 84.76 percent task accuracy across 9,000 tasks.",
        "source_ids": [
          "source-2026-11-02-007"
        ],
        "qualification": "A multi-agent run can fail its final requirement even when several intermediate results remain useful."
      }
    ],
    "source_ids": [
      "source-2026-11-02-007"
    ],
    "tags": [
      "multi-agent systems",
      "runtime repair",
      "contribution graphs",
      "reliability"
    ],
    "image_url": null,
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-11-02-007",
      "title": "arXiv preprint 2610.42144, original v1",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2610.42144v1",
      "canonical_url": "https://arxiv.org/abs/2610.42144v1",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-10-30T13:59:12.000-04:00",
      "accessed_at": "2026-11-02T08:25:20.000-05:00",
      "supports_claim_ids": [
        "claim-mp-2026-11-02-007-001"
      ]
    }
  ],
  "corrections": [],
  "publisher": {
    "name": "The Machine Press",
    "url": "https://themachinepress.com",
    "description": "A daily newspaper for the age of artificial intelligence."
  },
  "cite_this_report": {
    "title": "Multi-Agent Repair Keeps the Work That Still Counts",
    "publisher": "The Machine Press",
    "published_at": "2026-11-02T09:00:00.000-05:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-11-02-007/multi-agent-repair-keeps-the-work-that-still-counts"
  }
}