{
  "schema_version": "1.1",
  "id": "s8:https://www.youtube.com/watch?v=s67bE2Ur3bY",
  "slug": "wearing-the-agent-from-group-chats-to-glasses-sai-krishna-rallabandi-102hk49",
  "url": "https://feed7.dev/p/wearing-the-agent-from-group-chats-to-glasses-sai-krishna-rallabandi-102hk49",
  "title": "Wearing the Agent: From Group Chats to Glasses — Sai Krishna Rallabandi",
  "why_included": "Shared agents need a different harness: action-boundary security, selective group memory, per-user privacy, and routing that decides both who receives information and when the agent should speak.",
  "summary": "An agent deployed among friends and family for **eight months** exposed three group-specific concerns: security at the action boundary, curated shared memory, and recipient-aware routing. The proposed memory stores atomic facts, retrieves selectively, and accounts for time and hierarchy.",
  "practical_implication": "Design permissions around both the person and the room. Separate data from instructions, inspect tool actions deterministically, minimize retained facts, and add explicit policies for who may receive an answer and whether the agent should speak at all.",
  "agent_context": "An agent deployed among friends and family for **eight months** exposed three group-specific concerns: security at the action boundary, curated shared memory, and recipient-aware routing. The proposed memory stores atomic facts, retrieves selectively, and accounts for time and hierarchy.\n\nDesign permissions around both the person and the room. Separate data from instructions, inspect tool actions deterministically, minimize retained facts, and add explicit policies for who may receive an answer and whether the agent should speak at all.\n\nSeveral mechanisms remain proposals, including per-user LoRA adapters for privacy and a classifier for speaking turns. Model-baked permissions are not equivalent to enforceable access control, and the talk gives no measured privacy, security, or retrieval results.",
  "source": {
    "name": "AI Engineer",
    "url": "https://www.youtube.com/watch?v=s67bE2Ur3bY",
    "published_at": "2026-07-29T22:58:19.000Z"
  },
  "source_class": "video",
  "content_type": "Video",
  "layer": "agent",
  "domains": [
    "security"
  ],
  "topics": [
    "agent-memory",
    "agent-reliability",
    "context-engineering"
  ],
  "verification": {
    "status": "source_linked",
    "label": "Source Linked",
    "method": "source_feed",
    "verified_at": null
  },
  "uncertainty": [
    "Several mechanisms remain proposals, including per-user LoRA adapters for privacy and a classifier for speaking turns. Model-baked permissions are not equivalent to enforceable access control, and the talk gives no measured privacy, security, or retrieval results."
  ],
  "connected_context": {
    "meaning": "This moves shared-agent context from a retrieval problem to a room-aware governance problem: useful memory must be minimal, selectively retrieved, and constrained by both speaker and audience. It reinforces bounded context and explicit human authority while narrowing shared-context proposals with recipient-level disclosure and silence policies. The eight-month deployment supplies practical concerns, but not measured assurance that the proposed controls work.",
    "corpus_size": 297,
    "generated_at": "2026-07-31T10:07:13.431Z",
    "connections": [
      {
        "title": "WTF Is the Context Layer? The Missing Infrastructure for Production Agents — Prukalpa Sankar",
        "source_name": "AI Engineer",
        "source_url": "https://www.youtube.com/watch?v=8G_1-3IO4ZQ",
        "feed7_url": "https://feed7.dev/p/wtf-is-the-context-layer-the-missing-infrastructure-for-production-agent-0t47xqf",
        "reason": "Atlan’s shared, portable context layer gains a necessary constraint here: shared facts cannot imply universal visibility, because retrieval and disclosure must depend on the person and room."
      },
      {
        "title": "Citation Needed: Provenance for LLM-Built Knowledge Graphs — Daniel Chalef, Zep AI",
        "source_name": "AI Engineer",
        "source_url": "https://www.youtube.com/watch?v=H7puB0RwJMM",
        "feed7_url": "https://feed7.dev/p/citation-needed-provenance-for-llm-built-knowledge-graphs-daniel-chalef-1iob5t8",
        "reason": "Provenance and deletion policies complement atomic, minimal memory by making retained facts traceable and removable when shared knowledge changes or access must be revoked."
      },
      {
        "title": "UniMem: Complementary Episodic-to-Parametric Memory for Boundary-Agnostic Task Streams",
        "source_name": "arXiv",
        "source_url": "https://arxiv.org/abs/2607.26017v1",
        "feed7_url": "https://feed7.dev/p/2607-26017v1-1opv1da",
        "reason": "UniMem expands where experience can be stored, while this signal exposes the unresolved governance risk of consolidating private or audience-specific facts into less inspectable parameters."
      },
      {
        "title": "AI tools for Forward Deployed Engineering — Vasuman Moza, Varick Agents",
        "source_name": "AI Engineer",
        "source_url": "https://www.youtube.com/watch?v=l0FLhNqBOic",
        "feed7_url": "https://feed7.dev/p/ai-tools-for-forward-deployed-engineering-vasuman-moza-varick-agents-12kjg79",
        "reason": "Both place automation boundaries around real identities and authority; this signal narrows that process-first principle to group conversations, where permission includes whether the agent may answer a particular recipient at all."
      }
    ]
  },
  "lifecycle": "Current",
  "published_at": "2026-07-29T22:58:19.000Z",
  "modified_at": "2026-07-29T22:58:19.000Z",
  "supersedes": [],
  "expires_at": null,
  "formats": {
    "html": "https://feed7.dev/p/wearing-the-agent-from-group-chats-to-glasses-sai-krishna-rallabandi-102hk49",
    "json": "https://feed7.dev/p/wearing-the-agent-from-group-chats-to-glasses-sai-krishna-rallabandi-102hk49.json",
    "markdown": "https://feed7.dev/p/wearing-the-agent-from-group-chats-to-glasses-sai-krishna-rallabandi-102hk49.md"
  }
}