RAG is dead, right??
summary
Claims that hybrid retrieval wins: Cursor adds indexing and reports 24% increase in answer accuracy on its composer model, plus 2.6% code retention and 2.2% drop in dissatisfied requests in A/B tests, while Claude Code uses per‑session grep with repeat cost. Embeddings are framed as cached compute so upfront indexing trades CPU for far fewer runtime tokens. The practical build: Cursor’s Merkle‑tree incremental indexing works and reduces repeated embedding overhead; agentic iterative search combines vector, full‑text, regex and filters.
AI-written summary · about
chapters (8)
- 0:00Introduction to the "RAG is dead" discourse
- 1:12Google search volume trends for RAG
- 1:39Defining RAG vs. Agentic Search
- 3:15Cursor's indexing and semantic search approach
- 6:10Contrasting Claude Code (grep) vs. Cursor (indexed)
- 6:40The concept of embeddings as cached compute
- 8:38The shift from simple RAG to Agentic Retrieval
- 9:44Jeff Dean on context windows and stage retrieval
related talks
56mRAG: from dumb implementation to serious results
25mAdvanced RAG Systems: Scaling Retrieval-Augmented Generation in Production
59mThe Evolution of RAG
30mRAG: Tricks from the trenches.
53mBuild RAG from Scratch
27mOptimizing RAG Systems in Practice: Challenges and Proven Solutions
30mGraph RAG: Bringing together graph and vector search to empower retrieval
35mThe Hidden Jewels to Solve Confabulation Challenges in Generative AI’s RAG Promise