{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-11-05-014",
    "source_story_id": "tmp-story-speccompact",
    "edition_id": "mp-2026-11-05-morning-0118",
    "edition_url": "https://themachinepress.com/edition/2026-11-05",
    "position": 14,
    "story_type": "dispatch",
    "section": "developer-tools",
    "editorial_classification": "editorial",
    "headline": "The Agent Summarizes Its Memory One Slot at a Time—Together",
    "slug": "the-agent-summarizes-its-memory-one-slot-at-a-time-together",
    "dek": "SpecCompact drafts structured context slots in parallel, then has the target model verify and repair each proposed line.",
    "summary": "SpecCompact drafts structured context slots in parallel, then has the target model verify and repair each proposed line.",
    "body_text": "A small tuned drafter generates independent summary slots while the target model accepts or repairs lines and feeds the updated prefix into later work. Across AppWorld and tau-cubed with two target models, the system reports 2.04 to 5.63 times faster compaction than Hermes-style summarization. A Hermes user study records an 8.68 times median speedup. Downstream quality varies by model and task, so the result is a latency technique with measured tradeoffs, not lossless memory compression.",
    "why_it_matters": "SpecCompact drafts structured context slots in parallel, then has the target model verify and repair each proposed line.",
    "limitations": [
      "Downstream quality varies by model and task, so the result is a latency technique with measured tradeoffs, not lossless memory compression."
    ],
    "importance": 8,
    "canonical_url": "https://themachinepress.com/story/mp-2026-11-05-014/the-agent-summarizes-its-memory-one-slot-at-a-time-together",
    "json_url": "https://themachinepress.com/story/mp-2026-11-05-014.json",
    "first_published_at": "2026-11-05T09:00:00.000-05:00",
    "modified_at": "2026-11-05T09:00:00.000-05:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-11-05-014-001",
        "text": "SpecCompact drafts structured context slots in parallel, then has the target model verify and repair each proposed line.",
        "source_ids": [
          "source-2026-11-05-014"
        ],
        "qualification": "Downstream quality varies by model and task, so the result is a latency technique with measured tradeoffs, not lossless memory compression."
      }
    ],
    "source_ids": [
      "source-2026-11-05-014"
    ],
    "tags": [
      "agents",
      "context compaction",
      "speculative verification",
      "latency"
    ],
    "image_url": null,
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-11-05-014",
      "title": "arXiv preprint 2611.06876, original v1",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2611.06876v1",
      "canonical_url": "https://arxiv.org/abs/2611.06876v1",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-11-04T13:34:41.000-05:00",
      "accessed_at": "2026-11-05T08:29:24.413-05:00",
      "supports_claim_ids": [
        "claim-mp-2026-11-05-014-001"
      ]
    }
  ],
  "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 Agent Summarizes Its Memory One Slot at a Time—Together",
    "publisher": "The Machine Press",
    "published_at": "2026-11-05T09:00:00.000-05:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-11-05-014/the-agent-summarizes-its-memory-one-slot-at-a-time-together"
  }
}