thread of one happy fellow's beliefs, let go

Multithreading in OCaml 🙈

On Dean Jones' 210 at Chepauk.

Here is a rather lengthy write-up (in English) explaining why I chose #OCaml as my main programming language for my personal and professional projects! xvw.lol/en/articles/wh… I hope you find this interesting, and I welcome any feedback!

Tiger Emerging from Bamboo, by Katayama Yokoku, late 18th century

Hit 1000 km this morning for the year! On track for 1000 miles by the end of the year if I keep up my weekly mileage. 🏃‍♂️ I've not been consistent for the last few months. Need to up my game for the inevitable loss of time over the NE monoon.

New Raven alpha drops in 3 weeks! 🚀 Coming up: dataframes, tokenizers, transformers, forward-mode differenciation, vmap, and a complete linear algebra suite. Now's a really good time to jump in to contribute: tons of good first issues (just pick a failing test and fix it)

Join KC Sivaramakrishnan as he shares how the OCaml community evolves while maintaining legacy compatibility from OCaml 5’s concurrency & parallelism to cutting-edge tooling. 🎟️ Get tickets now: fossunited.org/dashboard/buy-… #unitedbyfoss #indiafoss #fossunited #indiafoss2025

Finally got around to trying OCaml multicore for something practical - the "find all references" command for the ReScript editor tooling. Here's a bench of it resolving references for a function in a 350k LoC / 1800 files repo: - Current (sync): 1.4s - Lwt: 1.6s - Eio: 0.55s

Super fancy use of effect handlers and domains to speed up Merlin, the editor service that provides advanced IDE features for OCaml: discuss.ocaml.org/t/ann-an-exper…

In this study ntrs.nasa.gov/citations/2014… by the NASA (2014), they compare formal verification with an interactive theorem prover ⚔️ model checking ⚔️ abstract interpretation. => Theorem prover requires the most knowledge, model checking only specifying, abstract interp nothing!

A new episode of Signals and Threads just dropped! This one is an interview with Chris Lattner, talking about Mojo, a new-ish language for GPU programming that's aiming to be an alternative to the CUDA stack. signalsandthreads.com/why-ml-needs-a…

Now on HN frontpage 🟧

The latency killer in distributed systems isn't the network, it's coordination. When two services have to agree on something, you're adding round-trip time, queueing delays, retry logic, and timeout handling. Eliminate coordination with partitioning or relaxed consistency.