{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-09-08-011",
    "source_story_id": "tmp-story-computer-use-persistent-skills",
    "edition_id": "mp-2026-09-08-morning-0061",
    "edition_url": "https://themachinepress.com/edition/2026-09-08",
    "position": 11,
    "story_type": "dispatch",
    "section": "developer-tools",
    "editorial_classification": "editorial",
    "headline": "The GUI Agent Kept a Versioned Memory of What Worked",
    "slug": "the-gui-agent-kept-a-versioned-memory-of-what-worked",
    "dek": "A persistent skill library improved a fixed computer-use stack across four observed domains—but repeated accepted revisions did not always recover a task.",
    "summary": "A persistent skill library improved a fixed computer-use stack across four observed domains—but repeated accepted revisions did not always recover a task.",
    "body_text": "The framework turns interaction trajectories and evaluator feedback into versioned procedures that become available on later iterations without changing model weights. After a five-iteration empty-library warm-up, the full system's post-warm-up mean evaluator score exceeded a matched empty-library control in all four OSWorld domain runs by 5.7 to 18.6 percentage points. A provenance study in GIMP found skills crossing task-of-origin boundaries, but also revision churn: repeated accepted edits could fail to recover the task that created the skill. The evidence supports conditional, auditable memory gains—not automatic compounding improvement.",
    "why_it_matters": "A persistent skill library improved a fixed computer-use stack across four observed domains—but repeated accepted revisions did not always recover a task.",
    "limitations": [
      "The evidence supports conditional, auditable memory gains—not automatic compounding improvement."
    ],
    "importance": 8,
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-08-011/the-gui-agent-kept-a-versioned-memory-of-what-worked",
    "json_url": "https://themachinepress.com/story/mp-2026-09-08-011.json",
    "first_published_at": "2026-09-08T09:00:00.000-04:00",
    "modified_at": "2026-09-08T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-09-08-011-001",
        "text": "A persistent skill library improved a fixed computer-use stack across four observed domains—but repeated accepted revisions did not always recover a task.",
        "source_ids": [
          "source-2026-09-08-011"
        ],
        "qualification": "The evidence supports conditional, auditable memory gains—not automatic compounding improvement."
      }
    ],
    "source_ids": [
      "source-2026-09-08-011"
    ],
    "tags": [
      "computer-use agents",
      "skills",
      "procedural memory"
    ],
    "image_url": null,
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-09-08-011",
      "title": "arXiv preprint 2609.04869",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2609.04869",
      "canonical_url": "https://arxiv.org/abs/2609.04869",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-09-04T04:29:45.000-04:00",
      "accessed_at": "2026-09-08T08:20:00.000-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-09-08-011-001"
      ]
    }
  ],
  "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 GUI Agent Kept a Versioned Memory of What Worked",
    "publisher": "The Machine Press",
    "published_at": "2026-09-08T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-08-011/the-gui-agent-kept-a-versioned-memory-of-what-worked"
  }
}