{
  "schema_version": "1.1",
  "id": "s13:https://arxiv.org/abs/2610.31636v1",
  "slug": "2610-31636v1-1i3iced",
  "url": "https://feed7.dev/p/2610-31636v1-1i3iced",
  "title": "Can an Agent Debug a Web Corpus? Agentic Auditing of Slovak FineWeb2",
  "why_included": "A budgeted agent turned corpus anomalies into reusable tests, but static sampling found more valid selectors for less money. The useful pattern is agent discovery followed by deterministic checks.",
  "summary": "Three agent runs inspected **26.5 million Slovak documents** through a read-only API and produced seven findings. Native-speaker review supported **6 of 7**, including a 1,408,437-document content farm.",
  "practical_implication": "Use the agent for exploratory debugging and have it compile findings into executable selectors. The resulting suite scanned the corpus in **0.7 seconds** without a GPU, making repeated checks deterministic and cheap.",
  "agent_context": "Three agent runs inspected **26.5 million Slovak documents** through a read-only API and produced seven findings. Native-speaker review supported **6 of 7**, including a 1,408,437-document content farm.\n\nUse the agent for exploratory debugging and have it compile findings into executable selectors. The resulting suite scanned the corpus in **0.7 seconds** without a GPU, making repeated checks deterministic and cheap.\n\nThe strongest static sampling baseline validated more selectors than any agent run at under half the cost. The agent added prevalence estimates, failure explanations, sampled precision checks, and reusable tests, but did not dominate discovery.",
  "source": {
    "name": "arXiv",
    "url": "https://arxiv.org/abs/2610.31636v1",
    "published_at": "2026-10-22T17:52:22.000Z"
  },
  "source_class": "blog_post",
  "content_type": "Paper",
  "layer": "agent",
  "domains": [
    "data"
  ],
  "topics": [
    "harness-engineering",
    "tool-use",
    "agent-reliability"
  ],
  "verification": {
    "status": "needs_review",
    "label": "Needs Review",
    "method": "unverified",
    "verified_at": null
  },
  "uncertainty": [
    "The strongest static sampling baseline validated more selectors than any agent run at under half the cost. The agent added prevalence estimates, failure explanations, sampled precision checks, and reusable tests, but did not dominate discovery."
  ],
  "connected_context": {
    "meaning": "This qualifies the prior case for agent-led data inspection: agents were valuable for explaining failures, estimating prevalence, checking precision, and turning discoveries into reusable selectors, but a cheaper static baseline found more validated selectors. The practical pattern is exploratory model judgment followed by deterministic, replayable checks—not treating the agent as either the sole discoverer or the production scanner.",
    "corpus_size": 898,
    "generated_at": "2026-10-24T10:28:40.044Z",
    "connections": [
      {
        "title": "How Kepler Built Verifiable AI for Financial Services — Vinoo Ganesh",
        "source_name": "AI Engineer",
        "source_url": "https://www.youtube.com/watch?v=Tt2kX2sgQio",
        "feed7_url": "https://feed7.dev/p/how-kepler-built-verifiable-ai-for-financial-services-vinoo-ganesh-0yqmhy7",
        "reason": "Both separate model judgment from deterministic execution: the corpus agent proposes and explains findings that become executable selectors, paralleling Kepler’s use of agents for choices while replayable systems handle consequential computation and validation."
      },
      {
        "title": "Semantic Blindness: 500,000 Sensors Confused an LLM - Raahul Singh & Vanč Levstik, Phaidra",
        "source_name": "YouTube",
        "source_url": "https://www.youtube.com/watch?v=EUsPvBeIx70",
        "feed7_url": "https://feed7.dev/p/semantic-blindness-500-000-sensors-confused-an-llm-raahul-singh-vanc-lev-159c4yr",
        "reason": "The resulting selectors play the same scaling role as deterministic indexes and set operations: the model handles ambiguity and investigation, then structured machinery performs exhaustive, repeatable corpus-wide resolution."
      },
      {
        "title": "Grow the Harness, Not the Context: From Strategy-Free Scaffolds to Reusable Specialist Agents",
        "source_name": "arXiv",
        "source_url": "https://arxiv.org/abs/2609.26760v1",
        "feed7_url": "https://feed7.dev/p/2609-26760v1-0lklvan",
        "reason": "Both move useful agent behavior into reusable code, but this study adds an important boundary: retained harness artifacts make repeated checks cheap and deterministic even when the agent does not outperform a static baseline at initial discovery."
      }
    ]
  },
  "lifecycle": "Current",
  "published_at": "2026-10-22T17:52:22.000Z",
  "modified_at": "2026-10-22T17:52:22.000Z",
  "supersedes": [],
  "expires_at": null,
  "formats": {
    "html": "https://feed7.dev/p/2610-31636v1-1i3iced",
    "json": "https://feed7.dev/p/2610-31636v1-1i3iced.json",
    "markdown": "https://feed7.dev/p/2610-31636v1-1i3iced.md"
  }
}