What actually shipped in agent engineering, pulled from the labs, arXiv and Hacker News.
See who we followA specialized post-trained model using precomputed indexes achieves comparable code search quality to frontier models at lower cost and latency.
Title states RAG is dead without further context provided.
SID-1 agentic search model achieves 24x faster speed and 1.9x higher recall than GPT-5.1 and RAG.
Hybrid search, agents, and database design improve retrieval after RAG systems.
Turbopuffer offers object storage-native database for search operations.
Distributed queues can be built using single JSON files on object storage.
Turbopuffer ANN v3 achieves 200ms p99 latency over 100 billion vectors.
Redesigned inverted index structure using fixed-sized posting blocks achieved 10x smaller indexes and better throughput.
BM25 query latency analysis shows longer queries scale less efficiently with document count and top_k.
Vectorized block-max MAXSCORE algorithm improved text search performance up to 20x for long queries.
Billion-scale vector storage for RAG.
Economical way of serving vector search workloads.
Memory, evals, and efficient storage in AI systems with turbopuffer and Braintrust.
turbopuffer explained how to build systems ten times cheaper using object storage.
Turbopuffer built a database system on object storage.
Turbopuffer is a search engine using object storage and SSD caching for cost-effective, low-latency vector searches.
One pattern, one tradeoff, one production failure story. A short weekly briefing for people building agentic systems.
Weekly email, one-click unsubscribe. We only use your address to send the briefing.
