turbopuffer Blog https://turbopuffer.com/blog Latest posts from the turbopuffer blog en-us Wed, 10 Jun 2026 03:25:51 GMT Training SID-1 to beat GPT-5 at search with 1k+ QPS RL https://turbopuffer.com/blog/reinforcement-learning-sid-ai https://turbopuffer.com/blog/reinforcement-learning-sid-ai SID-1 is an agentic search model that is 24x faster than GPT-5.1-high, 374x cheaper than Sonnet 4.5, and achieves 1.9x higher recall than traditional RAG pipelines. Here's how we trained it using large-scale RL on turbopuffer. Wed, 20 May 2026 00:00:00 GMT Simon Eskildsen on scaling Shopify, building turbopuffer, and the future of databases https://turbopuffer.com/blog/podcast-cafe-cursor https://turbopuffer.com/blog/podcast-cafe-cursor Thu, 14 May 2026 00:00:00 GMT Mixing numeric attributes into text search for better first-stage relevance https://turbopuffer.com/blog/rank-by-attribute https://turbopuffer.com/blog/rank-by-attribute turbopuffer now allows you to combine attribute values into the scoring function of text queries. Ranking by attribute helps achieve better relevance in the first stage with the same scalability characteristics as BM25. Mon, 27 Apr 2026 00:00:00 GMT Retrieval After RAG: Hybrid Search, Agents, and Database Design https://turbopuffer.com/blog/podcast-latent-space https://turbopuffer.com/blog/podcast-latent-space Thu, 12 Mar 2026 00:00:00 GMT Object storage-native database for search https://turbopuffer.com/blog/video-andy-pavlo-cmu https://turbopuffer.com/blog/video-andy-pavlo-cmu Mon, 09 Mar 2026 00:00:00 GMT Rust zero-cost abstractions vs. SIMD https://turbopuffer.com/blog/zero-cost https://turbopuffer.com/blog/zero-cost A customer query was taking over 4× longer than it should have. The profiler pointed at Rust code we'd assumed was free. We followed the trail all the way down to assembly to find the true cost. Wed, 18 Feb 2026 00:00:00 GMT How to build a distributed queue in a single JSON file on object storage https://turbopuffer.com/blog/object-storage-queue https://turbopuffer.com/blog/object-storage-queue How to build a single global queue for distributed systems on object storage: Start with a single file on object storage, then add write batching, a stateless broker, and high-availability. Thu, 12 Feb 2026 00:00:00 GMT ANN v3: 200ms p99 query latency over 100 billion vectors https://turbopuffer.com/blog/ann-v3 https://turbopuffer.com/blog/ann-v3 Our latest ANN release supports scales of 100+ billion vectors in a single search index, with 200ms p99 query latency at 1k QPS and 92% recall. Wed, 21 Jan 2026 00:00:00 GMT Designing inverted indexes in a KV-store on object storage https://turbopuffer.com/blog/fts-v2-postings https://turbopuffer.com/blog/fts-v2-postings How we redesigned our inverted index structure using fixed-sized posting blocks to achieve 10x smaller indexes and dramatically better throughput. Wed, 14 Jan 2026 00:00:00 GMT Why BM25 queries with more terms can be faster (and other scaling surprises) https://turbopuffer.com/blog/bm25-latency-musings https://turbopuffer.com/blog/bm25-latency-musings I analyzed how BM25 query latencies scale with document count and top_k. Longer queries scale less efficiently, and essential terms impact performance in some surprising ways. Wed, 07 Jan 2026 00:00:00 GMT Vectorized MAXSCORE over WAND, especially for long LLM-generated queries https://turbopuffer.com/blog/fts-v2-maxscore https://turbopuffer.com/blog/fts-v2-maxscore turbopuffer has improved text search performance up to 20x by adopting the same text search algorithm as Apache Lucene, a vectorized variant of block-max MAXSCORE Tue, 09 Dec 2025 00:00:00 GMT FTS v2: up to 20x faster full-text search https://turbopuffer.com/blog/fts-v2 https://turbopuffer.com/blog/fts-v2 turbopuffer's full-text search engine is getting a major upgrade for 20x better full-text search performance Thu, 04 Dec 2025 00:00:00 GMT Faster vector search https://turbopuffer.com/blog/podcast-the-database-school-aaron-francis https://turbopuffer.com/blog/podcast-the-database-school-aaron-francis Thu, 13 Nov 2025 00:00:00 GMT Billion-scale vector storage for RAG https://turbopuffer.com/blog/podcast-jason-liu https://turbopuffer.com/blog/podcast-jason-liu Tue, 04 Nov 2025 00:00:00 GMT He built a new database in his bedroom https://turbopuffer.com/blog/podcast-pmf-show-he-built-new-database-bedroom-now-powers-cursor-notion-anthropic https://turbopuffer.com/blog/podcast-pmf-show-he-built-new-database-bedroom-now-powers-cursor-notion-anthropic Thu, 30 Oct 2025 00:00:00 GMT Economical way of serving vector search workloads https://turbopuffer.com/blog/podcast-vector-podcast-economical-way-serving-vector-search-workloads https://turbopuffer.com/blog/podcast-vector-podcast-economical-way-serving-vector-search-workloads Thu, 18 Sep 2025 00:00:00 GMT turbopuffer on Postgres FM https://turbopuffer.com/blog/podcast-postgres-fm-turbopuffer https://turbopuffer.com/blog/podcast-postgres-fm-turbopuffer Fri, 12 Sep 2025 00:00:00 GMT Memory, evals, and efficient storage in AI systems with turbopuffer and Braintrust https://turbopuffer.com/blog/podcast-bessemer https://turbopuffer.com/blog/podcast-bessemer Thu, 11 Sep 2025 00:00:00 GMT How to build 10x cheaper with object storage https://turbopuffer.com/blog/podcast-barrchives-podcast-how-build-10x-cheaper-object-storage https://turbopuffer.com/blog/podcast-barrchives-podcast-how-build-10x-cheaper-object-storage Tue, 05 Aug 2025 00:00:00 GMT The infrastructure company powering the top AI apps https://turbopuffer.com/blog/podcast-unsupervised-learning-redpoint-ai-podcast-infrastructure-company-powering-top-ai-apps https://turbopuffer.com/blog/podcast-unsupervised-learning-redpoint-ai-podcast-infrastructure-company-powering-top-ai-apps Tue, 22 Jul 2025 00:00:00 GMT Billion-scale vector search with Notion https://turbopuffer.com/blog/video-data-council-notion https://turbopuffer.com/blog/video-data-council-notion Thu, 29 May 2025 00:00:00 GMT How do vector (search) databases work? https://turbopuffer.com/blog/podcast-how-do-search-databases-work https://turbopuffer.com/blog/podcast-how-do-search-databases-work Sat, 29 Mar 2025 00:00:00 GMT Native filtering for high-recall vector search https://turbopuffer.com/blog/native-filtering https://turbopuffer.com/blog/native-filtering Describing how turbopuffer ensures high recall (accuracy) on queries that use attribute filters. Tue, 21 Jan 2025 00:00:00 GMT Building a database on object storage https://turbopuffer.com/blog/podcast-database-from-first-principles https://turbopuffer.com/blog/podcast-database-from-first-principles Sat, 16 Nov 2024 00:00:00 GMT How to use AI to become a learning machine https://turbopuffer.com/blog/podcast-every-how-use-ai-become-learning-machine https://turbopuffer.com/blog/podcast-every-how-use-ai-become-learning-machine Wed, 11 Sep 2024 00:00:00 GMT Continuous recall measurement https://turbopuffer.com/blog/continuous-recall https://turbopuffer.com/blog/continuous-recall Describing how turbopuffer measures the recall (accuracy) of its vector indexes in production continuously. This ensures that turbopuffer's search results are accurate and reliable, despite using approximate nearest-neighbour algorithms to speed up queries. Wed, 04 Sep 2024 00:00:00 GMT turbopuffer: fast search on object storage https://turbopuffer.com/blog/turbopuffer https://turbopuffer.com/blog/turbopuffer Inaugural blog post about the development of turbopuffer, a search engine that uses object storage and SSD caching for cost-effective, low latency search. This post describes into the motivation behind its creation, its unique architecture, and how it significantly reduces costs for large-scale vector searches. Discover how turbopuffer is transforming search infrastructure for companies like Cursor and Suno, offering a scalable and reliable solution. Mon, 08 Jul 2024 00:00:00 GMT