I wrote an in depth history on why we ended up in this position jacobtomlinson.dev/posts/2025/p...

Justin Garrison@justingarrison.com · 9mo ago

Python makes package mangers as much as JavaScript makes frameworks

Interesting interview from @talkpython.fm with @crmarsh.com and @carljm.me on Astral’s upcoming type checker “ty” (formally red knot) www.youtube.com/live/XVwpL_c...

ty: Astral's new Type Checker (formerly Red-Knot) - Talk Python Live Stream

YouTube video by Talk Python

youtube.com

I wrote a new blog post about using `conda` without any kind of shell modifications. Just the Python entry point directly. No ~/.bashrc or Windows registry edits! Simply one more entry in PATH. rjai.me/posts/2025-0...

The v1 recipe format gets an official announcement on @conda-forge.org. I’ve converted a bunch of the feedstocks I maintain over the last few months and it’s a real improvement conda-forge.org/blog/2025/02...

On an arm64-based mac, you can create a osx-64 environment with `conda create -n foo --platform osx-64`. However after that `conda install` doesn't have a platform flag. You can get around this with `micromamba install ... --platform osx-64 --prefix <path to env>`

#atuin is one of my favorite apps I started using this year. Here's my shell history year in review.

Super helpful guide for installing PyTorch using pixi from @prefix.dev prefix-dev.github.io/pixi/v0.39.4... This is point of confusion for a lot of people so it’s great that there is official documentation now.

Pytorch Installation - Pixi by prefix.dev

None

prefix-dev.github.io

While I did Advent of Code for several years to play around with Rust, my new December challenge is rowing 200 km on the erg log.concept2.com/challenges/h...

Holiday Challenge

The Holiday Challenge is our largest challenge of the year, with people kicking off the holiday season in style and earning money for charity along the way!

log.concept2.com

I was playing proto-typing @jcristharif.bsky.social's msgspec to (de)serialize a big nested python object. Converting the object to a dict using `msgspec.to_builtins` was 50x faster than my current recursive method and json serialization was 7x faster than `json.dumps` for serializing the dict.

I finally replaced an aging (but still very useable) Macbook Pro from mid-2014, with an M4 Pro. After a few days of tinkering, I think I have it configured. I'm experimenting with a handful of changes that I think are going well so far: