{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-09-10-003",
    "source_story_id": "tmp-story-subagents-versus-agent-skills",
    "edition_id": "mp-2026-09-10-morning-0063",
    "edition_url": "https://themachinepress.com/edition/2026-09-10",
    "position": 3,
    "story_type": "dispatch",
    "section": "developer-tools",
    "editorial_classification": "editorial",
    "headline": "Long Tasks Favored a Fresh Context Over a Bigger Instruction Pile",
    "slug": "long-tasks-favored-a-fresh-context-over-a-bigger-instruction-pile",
    "dek": "A study found subagent execution stronger than loading reusable skill packages directly when each skill exposed a clear input-output contract.",
    "summary": "A study found subagent execution stronger than loading reusable skill packages directly when each skill exposed a clear input-output contract.",
    "body_text": "The researchers compare two ways of reusing procedural knowledge: putting a skill package into the main agent’s growing context, or invoking that package in a fresh subagent context. On long-horizon tasks, the subagent approach performed better when the package stated a clear input-output contract and encoded actionable procedure. The improvement came with extra coordination tokens, so the result is a trade-off rather than a universal argument for delegation. The paper’s central claim is architectural: how knowledge is invoked can matter as much as what the knowledge contains.",
    "why_it_matters": "A study found subagent execution stronger than loading reusable skill packages directly when each skill exposed a clear input-output contract.",
    "limitations": [],
    "importance": 8,
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-10-003/long-tasks-favored-a-fresh-context-over-a-bigger-instruction-pile",
    "json_url": "https://themachinepress.com/story/mp-2026-09-10-003.json",
    "first_published_at": "2026-09-10T09:00:00.000-04:00",
    "modified_at": "2026-09-10T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-09-10-003-001",
        "text": "A study found subagent execution stronger than loading reusable skill packages directly when each skill exposed a clear input-output contract.",
        "source_ids": [
          "source-2026-09-10-003"
        ],
        "qualification": null
      }
    ],
    "source_ids": [
      "source-2026-09-10-003"
    ],
    "tags": [
      "subagents",
      "agent skills",
      "long-horizon tasks"
    ],
    "image_url": "https://themachinepress.com/issues/2026-09-10/subagents-code-file.webp",
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-09-10-003",
      "title": "arXiv preprint 2609.09233",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2609.09233",
      "canonical_url": "https://arxiv.org/abs/2609.09233",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-09-06T20:00:00.000-04:00",
      "accessed_at": "2026-09-10T08:24:00.000-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-09-10-003-001"
      ]
    }
  ],
  "corrections": [],
  "publisher": {
    "name": "The Machine Press",
    "url": "https://themachinepress.com",
    "description": "A daily newspaper for the age of artificial intelligence."
  },
  "cite_this_report": {
    "title": "Long Tasks Favored a Fresh Context Over a Bigger Instruction Pile",
    "publisher": "The Machine Press",
    "published_at": "2026-09-10T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-10-003/long-tasks-favored-a-fresh-context-over-a-bigger-instruction-pile"
  }
}