{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreibxvuahprsgzst4dx4redmkc6rasoobfhcmegxueiewwmo6sh763q",
    "uri": "at://did:plc:lk3jfj3zq4k4wxnk474axylu/app.bsky.feed.post/3mnubsnvuooo2"
  },
  "path": "/t/is-rag-overhyped-or-still-the-best-approach-for-llms-with-external-knowledge/1383161#post_1",
  "publishedAt": "2026-06-09T12:36:20.000Z",
  "site": "https://community.openai.com",
  "textContent": "Retrieval-Augmented Generation (RAG) is widely used in LLM-based applications to connect models with external knowledge sources. It’s often presented as the default solution for improving factual accuracy and grounding responses.\n\nBut in practice, there are mixed opinions on how effective and scalable it really is.\n\nSome argue RAG adds unnecessary complexity and cost, while others see it as essential for any production-grade AI system that relies on dynamic or domain-specific data.\n\nThis raises a few discussion points:\n\n  * Is RAG still the best general approach for grounding LLMs?\n  * Are simpler prompting or fine-tuning strategies sometimes more effective?\n  * Where does RAG start to break down in real-world systems?\n  * Is the industry over-relying on vector search as a solution?\n\n\n\nWould be interesting to hear different perspectives from people building with LLMs.",
  "title": "Is RAG Overhyped or Still the Best Approach for LLMs with External Knowledge?"
}