Today, we're announcing our first hosted infrastructure product: pyx, a Python-native package registry. We think of pyx as an optimized backend for uv: it’s a package registry, but it also solves problems that go beyond the scope of a traditional "package registry".

The uv build backend is now stable, and considered ready for production use. An alternative to setuptools, hatchling, etc. for pure Python projects, with a focus on good defaults, user-friendly error messages, and performance. When used with uv, it's 10-35x faster.

Today, we’re announcing the preview release of ty, an extremely fast type checker and language server for Python, written in Rust. In early testing, it's 10x, 50x, even 100x faster than existing type checkers. (We've seen >600x speed-ups over Mypy in some real-world projects.)

zlib-rs is now used in uv! uv is the extremely fast Python package & project manager with over 200M requests/day. Not only are we happy about the adoption but also grateful for Astral and @crmarsh.com stepping up and supporting us through a GitHub sponsorship. Thanks!

In the latest version of uv, we've updated the build system for our managed Python versions to use the latest LLVM version and cherry-picked an extra fix for a performance bug (github.com/llvm/llvm-pr...) and the early benchmarks are pretty exciting! On M3 macOS

uv is one year old today. Hard to believe its only been a year. The growth, adoption, and impact surpassed my wildest expectations. Happy birthday, uv! 🥳

Python 3.14.0a5 is out today in uv 0.5.31 and includes a new tail calling interpreter which, from some quick benchmarks, looks like a 1.23x performance improvement !

How to create a Python CLI that's globally available in your system in 5 easy steps. 1. Install uv 2. Init project with `uv init --app --package mycli` 3. Write code 4. Install with `uv tool install . -e` 5. Use `mycli` anywhere in your computer

🐍🎧 Charlie Marsh: Accelerating Python Tooling With Ruff and uv How is the Rust programming language being used to speed up Python tools? This week on the show, we speak with @crmarsh.com about his company, Astral, and their tools, uv and Ruff. https://realpython.com/podcasts/rpp/238/

I want to thank @crmarsh.com for coming on the show this week. What a great conversation! We dug in to Ruff, uv, Astral, and a new typing tool being developed. Thanks again! realpython.com/podcasts/rpp...

We’re building a new static type checker for Python, from scratch, in Rust. From a technical perspective, it’s probably our most ambitious project yet. We’re about 800 PRs deep!

Incredible: PyCharm now ships with first-class support for uv!

We're continuing to grow the Astral team, and every hire we make is like a dream hire. I don't understand how we're doing it but it rules. This time, it's @gankra.bsky.social! Most recently CTO at Axo, author of cargo-dist, author of The Rustonomicon, Rust 1.0 team member, and more.

It was really valuable for me to stay at a single company long enough to live with the consequences of my own engineering decisions. To come face-to-face with my own technical debt. It takes a few years for this to play out but IMO it's an important part of engineering growth.

🚀 Big news for #Python devs! Socket now supports uv.lock files, bringing deterministic dependency resolution and enhanced supply chain security to your projects. 🛡️ Say goodbye to dependency hell and hello to secure, consistent builds! 🔒✨ @crmarsh.com socket.dev/blog/socket-...

We now have first-class documentation for using uv with AWS Lambda docs.astral.sh/uv/guides/in...

In the next uv release, resolving Apache Airflow (with a warm cache) gets over 2x faster.

You can now install the latest Python 3.14 alpha with uv

When we get Renovate PRs for dependencies that people on the Astral team maintain

A nice side-effect of making things programmable is that it's much easier to get help from LLMs... For example: infrastructure-as-code means that I can get an LLM to generate CDK, then run the generated CDK, paste back errors, and iterate. No need for any other form interaction.

As someone that learned Rust in 2022, it's easy to forgot how long it takes to build a language and an ecosystem... Rust 1.0 was almost 10 years ago (2015). Graydon Hoare started working on it almost 20 years ago (2006).

Being 100% serious, for me what's described here is a big part of Rust's success. It's fun. And it makes things that are often un-fun (installing the toolchain, adding dependencies) easy.

I love the Fish shell writeup on their Rewrite-it-in-Rust journey. It's really nuanced: there's some good, some bad, some introspection on their own mistakes. But most importantly: they did it! 2604 commits over ~two years of work across > 200 authors. Ship ship ship! fishshell.com/blog/rustport/

My Christmas gift is this series of PRs making standalone PEP 723 scripts first-class targets for all uv commands. You'll be able to create lockfiles for PEP 723 scripts, export them to requirements.txt, view the dependencies with uv tree, and more.

Any good write-ups on performance improvements across glibc versions? I.e., what kind of lift can you expect when bumping your glibc version?

We now have a shared channel with the CPython core devs where we're discussing upstreaming python-build-standalone changes to CPython itself. It's been pretty awesome to see how welcoming, receptive, and helpful they've been. Grateful!

Hearing early reports that the first Astral PR upstreaming python-build-standalone changes to CPython has been merged...

Last week I focused on addressing quirks in python-build-standalone and fixing the common pain points with uv-installed Pythons. For example: if you've ever run into this tkinter error, it's now fixed.

Thanks to @crmarsh.com's uv and @jlowin.dev's FastMCP, the simplest MCP Server in Python with prompts, resources and tools, will soon be: gist.github.com/dsp-ant/f5f7.... One file, 34 lines! FastMCP will be part of Python MCP SDK 1.2.0, coming in the next days.

A Git repo is worth a thousand words (when filing issues)

Unfortunately I’ve discovered a new “hardest problem in Python dependency resolution”

python-build-standalone is now officially part of the Astral organization!

It's so over (just wrote some C)

uv's resolver now recognizes that certain combinations of markers are impossible, even if the grammar allows them. For example, sys_platform == "darwin" and platform_system == "Windows" can't be true at the same time. Lets us do less work by solving fewer combinations.

Justifying my terrible new abstraction by referring to it as a "lowering" (wiser, more powerful)

Congratulations to the new #Python Steering Council for 2025! 🗳️ @pumpichank.bsky.social 🗳️ @corona10.bsky.social 🗳️ @emilyemorehouse.bsky.social 🗳️ @gpshead.bsky.social 🗳️ @pablogsal.com discuss.python.org/t/steering-c... hugovk.github.io/python-steer... #SteeringCouncil #PythonSteeringCouncil

Have we reached peak Member of Technical Staff?

uv has made it into GitHub's default .gitignore

My @eurorust.eu talk is up! "Building an extremely fast Python package manager, in Rust" It was the closing talk of the conference, covering what uv is, why we built it, how it works, what makes it fast -- and the role that Rust plays in all of this. www.youtube.com/watch?v=zOY9...

How I use Astral’s uv to manage my local development environments: 1. Install uv through homebrew, install Python versions through uv. Upgrade uv version to get access to new Python builds, do this frequently #pythonprogramming

2025 budget? That's an Astral Projection

python-build-standalone has exploded in popularity, with over 70M downloads all-time. But what is a "standalone" Python build? It wasn't obvious to me at first! Initially, I didn't understand what problem was being solved or why it mattered...

Big news: Gregory Szorc's python-build-standalone project will be transitioning to Astral. If you've installed Python with uv, then you've used python-build-standalone. It powers installs for uv, Rye, mise, Bazel, pipx, Hatch, and more, with over 70,000,000 downloads. astral.sh/blog/python-...

The Astral documentation gets... a lot of traffic? That seems good

Today is my first day at Astral! Very excited to work with @crmarsh.com and the gang on Python tooling. It'll be fun (and strange) to focus on only one language for a bit!

I still find that Claude and ChatGPT are really, really weak at answering questions about Rust. Instead I use BurntSushi API for my questions.

Always a good reminder: both Ruff and uv see more PyPI downloads from Windows than macOS.

I've never done Advent of Code and at this point I'm afraid to try

Found an optimization in uv's resolver that makes Apache Airflow (with all optional dependencies enabled) resolve >20% faster.