Just published a blog post I've been working on for a little while: "Async Rust can be a pleasure to work with (without Send + Sync + 'static)" TL;DR: async @rustlang is actually pleasant to use with structured concurrency and thread-per-core runtimes emschwartz.me/async-rust-can…

I just posted a new blog post: Unnecessary Optimization in Rust: Hamming Distances, SIMD, and Auto-Vectorization emschwartz.me/unnecessary-op…

Thanks @tembo_io for benchmarking @paradedb and Postgres full text search. I happened to be writing about BM25 when I found it so I included it in the second post tembo.io/blog/paradedb-searc…

I just published two blog posts digging into the BM25 full text search algorithm: Understanding the BM25 full text search algorithm: emschwartz.me/understanding-… Comparing full text search algorithms: BM25, TF-IDF, and Postgres emschwartz.me/comparing-full…

Here's a new blog post about why binary vector embeddings are so cool. TL;DR: they retain 95% retrieval accuracy with 32x compression and a ~25x speedup for lookups 🤯 Thanks to the folks at @mixedbreadai for their work on this topic! emschwartz.me/binary-vector-…

Wrote another Rust blog post: Understanding Memory Ordering in Rust I'm reading @m_ou_se's Rust Atomics and Locks book and didn't grok memory ordering the first time through, so I tried to understand it better by re-explaining it emschwartz.me/understanding-…

I had a lot of fun hacking on @TigerBeetleDB yesterday. It seems powerful and it's a neat take on a database (increase the efficiency and simplicity by using a very domain-specific model for financial accounting). I do think the DX can be improved w/ codegen analogous to gRPC

I'm quite excited about this feature Autometrics now includes a `build_info` metric with the version & commit and then writes queries that can help you correlate problems with which version may have introduced them. The PromQL tricks that make it work are also quite fun

I wrote up some of the thinking that underpins the @AutometricsDev project in another blog post. I'd love to hear what observability & devops folks think about this! fiberplane.com/blog/the-case…

This was a fun experience stretching the limits of what Prometheus labels + queries can be used for. Now every @AutometricsDev-instrumented project can use the same alert rules file Thanks to @sinkingpoint for the inspiration on (ab)using labels for slightly unusual purposes 😊

Does anyone know why @OpenTelemetry and @OpenMetricsIO standardized metrics in _slightly_ different ways? - _count vs _total suffixes - seconds vs milliseconds These do not seem like meaningful differences but they make it impossible to support both formats at the same time

Autometrics v0.2 is out now with the following changes: - You can apply `#[autometrics]` to a whole `impl` block (in addition to individual functions) - Supports the `metrics` & `prometheus` crates for tracking metrics - Tracking concurrent requests is now optional