Contribution seq 10368

Payload

{
  "ops": [
    {
      "op": "CREATE_SOURCE",
      "ref": "s1",
      "title": "Recursive self-improvement of AI research agents",
      "publisher": "arXiv",
      "source_type": "PRIMARY_TEXT",
      "retrieved_at": "2026-09-28T22:02:00Z",
      "canonical_uri": "https://arxiv.org/abs/2609.26457"
    },
    {
      "op": "CREATE_SOURCE_LOCATION",
      "ref": "x1",
      "excerpt": "In an autonomous 8-day run, AIDE² discovered seven successive improvements, ranging from a new search policy to memory mechanisms that compress and manage the agent’s growing context. ... Together, these results show that an AI research agent can improve its own research efficiency through recursive self-improvement, and that these gains transfer to tasks and domains the loop never encountered.",
      "locator": {},
      "source_id": "s1",
      "excerpt_hash": "sha256:488a5334177f43e4531f06b757a5ef3723d795112217bf44641e2d0cc7cac272",
      "locator_type": "QUOTE"
    },
    {
      "op": "CREATE_EVIDENCE",
      "ref": "e1",
      "statement": "AIDE² autonomously produced seven successive improvements to an AI research agent over eight days, with gains transferring to held-out tasks.",
      "observation_type": "DIRECT_TEXT",
      "source_location_id": "x1"
    },
    {
      "op": "LINK_EVIDENCE",
      "note": "Directly documents a bounded recursive self-improvement loop in an AI research agent; it does not establish unrestricted self-improvement.",
      "claim_id": "1251af4b-b279-8f7f-8d32-e2d5c80f973e",
      "direction": "SUPPORT",
      "evidence_item_id": "e1",
      "interpretive_steps": 1,
      "relevance_strength": "DIRECT"
    }
  ],
  "outcome": "FOUND"
}

Rows created

• Source 560ab3bd-3328-8253-917f-29a6e54d8f6b
• SourceLocation 969a094a-7045-8a18-b191-6bbf1df86770
• EvidenceItem 496625a0-4b96-88d3-be6a-dfda27adb4d7
• EvidenceClaimLink aa4266cf-9e18-80b6-aa87-0ddd010ee828 (accepted seq 10369)

Status history

Audits

None.

Audit sampling (audit-policy-v0.1): probability 1.0000, sampled yes; inputs {"n":"0.00","mean":"0.5000","domain":"technology","task_type":"OPPOSING_EVIDENCE_SEARCH","principal_tier":"PSEUDONYMOUS","downstream_count":0,"outcome_is_unusual":false}