tpuf v3 full-text search now beats prod (v2) across the board on our internal benchmark FTS v2 already used a highly-tuned vectorized search algorithm. FTS v3 uses the same algorithm applied to a more efficient and CPU-friendly storage layout

storage lookups got up to 1.9x faster on tpuf v3 using cascading binary search v3 keys rows on 3 columns (a, b, c). lookups inefficiently compared all 3 during each search step now we cascade: search a, then b, then c, narrowing the search range as we go

contextual embedding improves relevance by encoding the context of a full document into each chunk's embedding we worked with @perplexity_ai to iteratively evaluate their model checkpoints against our private benchmark, resulting in a state-of-the-art contextual embedding model

we're redesigning our caching tier current: whole files evicted in the order written (FIFO) new: smaller chunks, least-read evicted first (LFU) on a cluster where a small fraction of data gets most of the reads, this improves byte hit ratio by ~2x → p99 query latency drops ~2x

Vibes were immaculate at the systems meet-up x @turbopuffer tonight ✨ Packed out a loft in LES with an incredible crowd of infra nerds! Huge thanks to everyone who came through and to our speakers from tpuf, @gen_intuition @cursor_ai and @authzed 💥 ft. WAL Street elevated with 🐡

we've been using Claude Mythos for our internal defense systems we could ask Mythos to examine every log source, but this is inefficient at scale instead, we use turbopuffer to narrow billions of security logs to the anomalous few that we need Mythos to attend to

The people of Silicon Valley are unwell. If you work at a frontier lab or decacorn, you’ve got it bad. You haven’t seen your kids, your girlfriend’s annoyed, invites from friends to hang out are waning, even weddings are on hold. This is the time to make your mark and make your money, after all. If you’re building a startup that’s trying to compete with or be one of these companies, you’ve got it bad. Maybe you passed up a job offer at one a couple years back, which means you left $100M on the table, you fucking idiot. Being an underdog isn't as fun as you thought. If you’re working on a pre-AI startup, you’ve got it bad. Not too long ago you were the belle of the ball, but now your 30% year-on-year growth doesn’t feel like growth at all. It feels like death. "Too many of my friends are spiraling," Brie Wolfson writes. "I don’t know how we’re supposed to live and work like this." It’s all too much, too fast. Too much money, too much growth, too much power, too much change. Every month, a new frontier model is released. Every day, a new unicorn is minted. Seriously. 300 days into the year 2026, and 311 unicorns have been minted so far. So what’s left for the sane ones? Brie went looking for teams playing a different game. She found one in Ottawa: a search infrastructure company that many of the AI darlings rely on. It's growing fast, and its people are well. Read @zebriez’s latest company dispatch from inside the sane asylum of turbopuffer.

turbopuffer v3 will be the first search engine to adopt fully vectorized query execution, a design more often used by OLAP engines modern CPUs reward dumb, serial work. tpuf v3 makes the best use of the hardware for better performance on more queries

making search fast means touching fewer bytes and doing less work per byte tpuf v3 compresses columns with composable schemes: 1M zeros becomes {val: 0, len: 1_000_000}, and predicates run directly on compressed data 1 comparison vs 1M → faster filters

Review Tables now load up to 5.8x faster in Harvey. Our engineering team built server-side pagination to support larger reviews without loading the entire table into browser memory. That required: - A Turbopuffer query layer for server-side sorting and filtering - Custom row-ID pagination and staged loading - Typed and string fields to handle inconsistent model outputs - Fallback reads and backfill jobs to recover from data drift

Nemotron-3.5-Embed is now available in turbopuffer native embedding on our enterprise PDF search benchmark, Nemotron-3.5-Embed 8B improves recall@20 by ~5% over the next best model

rescoring fewer full-precision vectors cut disk read I/O by up to 80% on one prod cluster lower disk I/O produces higher query throughput, especially on smaller VMs where full-precision vectors don't fit in memory it also produces beautiful charts