@GergelyOrosz: This is rare: turbopuffer love blogging + documenting + charting how they rewrite their underlying engine, starting fro…
Summary
Gergely Orosz highlights turbopuffer publicly documenting its in-progress rewrite of the tpuf storage engine (v3), grinding query plans in public rather than only celebrating after the rewrite is finished and fast.
View Cached Full Text
Cached at: 09/30/26, 08:11 PM
This is rare: turbopuffer love blogging + documenting + charting how they rewrite their underlying engine, starting from now when rewrite is done but performance is… ughh
Usually eng blog posts do a celebratory one after it’s ALL finished. Not this team, and I love it:
turbopuffer (@turbopuffer): we’re changing tpuf’s storage architecture so we can keep pushing the search frontier
the new engine (v3) now passes CI, but we have to make it fast before we ship it to prod
we’re grinding query plans in public until v3 wins on everything
follow along:
Similar Articles
@GergelyOrosz: Ever wondered why the founder of @turbopuffer (that is @Sirupsen) was up to 12-16 years ago, when he was getting starte…
A tweet by Gergely Orosz highlights that Simon Eskildsen, founder of turbopuffer, kept a blog from 2010–2014, and shares a comprehensive list of recent talks and podcasts about turbopuffer and vector search.
@GergelyOrosz: So damn excited to share that @turbopuffer is a season sponsor of The Pragmatic Engineer Podcast! I first heard of them…
Gergely Orosz announces turbopuffer as a season sponsor for The Pragmatic Engineer Podcast, highlighting the database's innovative use of object storage and smart caching that helped companies like Cursor scale while cutting costs by 95%.
@vivekgalatage: It's super interesting to know the system architecture of the TPUs. https://henryhmko.github.io/posts/tpu/tpu.html…
A deep dive into Google's TPU architecture, explaining the design philosophy of systolic arrays, pipelining, and ahead-of-time compilation that enables high throughput and energy efficiency.
@GergelyOrosz: So much sw today doesn’t care much about perf. It just feels sluggish. And everyone seems to have accepted it. Except w…
Gergely Orosz discusses how modern software often neglects performance, citing examples like Bun and Linear, and highlights a fast demo of Superlogical for tmux by Mitchell Hashimoto.
@leloykun: [WIP] Blog post on Lean4-to-TileLang Tensor Program Superoptimizer here:
A technical blog post introduces a Lean4-to-TileLang tensor program superoptimizer that automatically generates optimized GPU/TPU kernels and hyperparameter scaling laws, demonstrating performance gains over torch.compile.