Python packaging was finally solved, and the solution turned out to be a tool so powerful, user environment sanity isn't the only use case. It installs things. No more and no less than that.
OpenAI revealed its acquisition of Astral on March 19, 2026. The entire Astral team, with founder Charlie Marsh, joined the Codex org as a result.
The tool everyone already switched to
Allow me to provide some context for those who may have missed out on this information for some reason.
Former Khan Academy engineer, Charlie Marsh, founded Astral with the mission to reimplement Python's toolchain in Rust. Astral released uv, which is a drop-in alternative to pip, virtualenv, and pip-tools in February 2024.
The speed improvement was so significant that Astral commented that it wasn't just a "nice improvement", but it was in fact 10 to 100 times faster than the previous pip workflows.
The statistics support this. uv can resolve the Jupyter project dependencies in about 0.57 seconds when cold. Poetry takes around 7.59 seconds. When the cache is warm? Just about 20 milliseconds. ⚡
It caught people's attention. uv was already attracting over 126 million downloads monthly by early 2026, according to JetBrains.
When I use the term "conquered," I'm not exaggerating. uv didn't win a debate. It won over the most users.
Governance loses the race
This is the part that's meant to annoy you.
While uv was gobbling up the ecosystem, the community was otherwise engaged... trying to have a committee organize the ecosystem. PEP 772 proposed an elected Python Packaging Council, a real governance body for exactly these decisions.
The PEP remained under review for 14 months.
Compare those two situations as they unfolded in parallel. A product supported by corporate interests was released, improved, and became popular with 126 million monthly users before the city's elected officials even finished discussing its governance structure.
That sums it up. The town was writing a constitution. Astral was delivering binaries.
This is not meant as a criticism of the council people. Governance is intentionally slow and cautious. The problem is that slow and cautious will always lose to a fast product with actual distribution.
So who does uv answer to now?
This is what I keep wondering about.
For a couple of years, the response was "an independent startup that quite enjoys Rust." Okay, so startups have motives, but these were packaging-shaped motives.
The correct response would be the Codex team from OpenAI.
Just to clarify, uv is still going to be open source, and in the announcement Marsh said: "As part of Codex, we'll continue evolving our open source tools to push the frontier of software development." So I guess we'll have to take him at his word.
However, intentions and structure are different things. A large portion of Python's daily package installs is now being funneled through a technology owned by a monopoly corp. That is a dependency the community never approved.
OpenAI is an organization that adheres to a set of incentives in a world inexorably stratified by power and money.
→ The community built a democratic process. Astral built a product.
→ Only one of those decided who controls Python packaging.
→ Distribution beats deliberation, every time.
→ By the time governance catches up, the winner already has an owner.
The takeaway
We like to think that open ecosystems are self-governing and based on consensus. Here's a perfect example that this is not always the case.
It was the tool that won, not the governance body. Uv was a seriously good tool that outgrew the process and outcompeted the governance meant to steer it. Ecosystem governance doesn't lose to bad actors, it loses to good tools that iterate faster than committees can convene.
Here's what I'd like to know: if technology advances and how resources are distributed will always be ahead of the decisions made by elected officials, is there a form of local governance that can meet the pace? Or should we simply accept that the quickest and best-financed solution will prevail, and work from there?
Top comments (0)