OpenAI bought Astral and I'm not sure how I feel about it
OpenAI just acquired Astral, the company behind uv—a Python package manager that 126 million developers downloaded last month. If you don't write code, this is about infrastructure control becoming the new battleground in AI. Astral also built ruff (a code linter) and ty (a type checker), but uv is what matters here.
Last updated: March 20, 2026 | Author: Koko
What Astral actually built
Astral's whole thing is uv. You say it "you-vee" if you want to sound like you know what you're talking about. It's a Python package manager, which means it handles all the annoying environment stuff that used to make Python developers want to throw their laptops.
According to PyPI Stats, uv got downloaded 126 million times last month. That's wild for a tool that came out in February 2024—barely a year ago. Simon Willison (the Django guy) calls it "essential" now. When someone like Simon calls something essential, you pay attention.
They also made ruff, which is a linter/formatter, and ty for type checking. Ruff's good. People use it. But uv is the one that actually matters here.
Why OpenAI wants this
The official story is that the Astral team joins Codex, which is OpenAI's answer to Claude Code. They're competing directly now.
Here's what Charlie Marsh (Astral's founder) wrote:
"Open source is at the heart of that impact and the heart of that story; it sits at the center of everything we do. In line with our philosophy and OpenAI's own announcement, OpenAI will continue supporting our open source tools after the deal closes."
And OpenAI said:
"By bringing Astral's tooling and engineering expertise to OpenAI, we will accelerate our work on Codex and expand what AI can do across the software development lifecycle."
Notice the difference? Charlie's talking about open source. OpenAI's talking about accelerating Codex. Both can be true, but the emphasis tells you something.
Also—Astral has BurntSushi. If you don't know, he made ripgrep, which is grep but actually good, and the Rust regex library. Codex CLI is written in Rust. So yeah, the talent thing is real.
But here's the thing about talent acquisitions: they start as "we love the product" and end up as "actually we just wanted the people." I've seen this movie before.
The infrastructure war
This acquisition only makes sense if you understand what's actually being fought over. And it's not just models anymore.
OpenAI and Anthropic spent all of 2025 making their coding models better. Then November happened—coding agents went from "sometimes useful" to "I can't work without this" for a lot of developers. Claude Code and Codex are going head to head. Those $200/month subscriptions? That's billions of dollars.
In December 2025, Anthropic bought Bun. That's the JavaScript runtime that powers Claude Code. Jarred Sumner (who built Bun) has been making Claude Code noticeably faster since then.
Now OpenAI's doing the same thing with Python. The pattern isn't subtle: control the infrastructure, not just the model.
What happens to uv now?
Here's the worry everyone has. When VC-backed companies control critical open source tools, what happens if incentives change?
Armin Ronacher built Rye, which got merged into uv. He wrote this back in August 2024:
"Even in the worst possible future this is a very forkable and maintainable thing. I believe that even in case Astral shuts down or were to do something incredibly dodgy licensing wise, the community would be better off than before uv existed."
And Douglas Creager from Astral said on Hacker News:
"That does not change with this acquisition. No one can guarantee how motives, incentives, and decisions might change years down the line. But that's why we bake optionality into it with the tools being permissively licensed."
The licenses are permissive. If OpenAI messes this up, the community can fork and keep going. That's the safety valve. But still. I've watched companies acquire beloved tools and slowly... not kill them exactly, just let them drift. You know how it goes.
Why the harness matters more than the model
I've been thinking about this a lot: the harness matters more than the model.
Modern AI isn't just weights and training data. It's also:
The runtime (how the model actually runs code)
The tools (what it can call)
The orchestration (how it all fits together)
We call this the "harness" at GreatApeAI. And it's where the real fighting is happening now.
OpenAI isn't just hiring smart people. They're buying control over how millions of developers run Python. When Codex generates Python code, it can now be optimized for uv specifically. When it needs to lint or format, it uses ruff and ty.
The model itself? Important, sure. But the system around it is what makes it actually useful.
What this means for AI startups
A few things this tells us:
Infrastructure is the new moat. The winners aren't just the ones with best models. They're the ones who control the whole stack.
Open source tools are strategic. Astral built something so essential, so fast, that OpenAI had to have it. There's a lesson there for founders who actually understand developer pain.
Everything's verticalizing. OpenAI and Anthropic aren't just competing on benchmarks anymore. They're competing on who can build the better end-to-end experience.
Forkability is protection. Astral's licensing means the community has leverage. If OpenAI screws this up, developers can leave. That's not nothing.
So what
OpenAI buying Astral is bigger than "we hired some good engineers." It's about owning the infrastructure Python developers depend on. It's about making Codex work better by controlling the tools Python needs.
This is what AI competition looks like now. Not just smarter models. Owning the whole system that makes AI useful in the first place.
For startups: don't just build models. Build infrastructure. Build tools that become essential. Because that's where value is being captured right now.
What do you think? Is owning the infrastructure layer the new way to win in AI? Or does this just create openings for open source alternatives?
FAQ
What is Astral and why did OpenAI acquire it?
Astral is the company behind uv (Python's most popular package manager), ruff (a linter), and ty (a type checker). OpenAI acquired them to accelerate Codex development and control more of the Python development infrastructure.
What is uv in Python?
uv is a Python package manager that simplifies environment management. It was downloaded 126 million times last month and has become essential for Python developers since its release in February 2024.
Will uv remain open source after the OpenAI acquisition?
Yes, both Astral and OpenAI have stated they will continue supporting the open source tools. The code is permissively licensed, meaning the community can fork it if needed.
How does this compare to Anthropic's Bun acquisition?
Anthropic acquired Bun (JavaScript runtime) in December 2025. Both acquisitions show AI companies buying infrastructure tools to improve their coding agents. It's a pattern of vertical integration.
What is the "harness" in AI development?
The harness is the infrastructure layer around AI models: the runtime environment, tools, and orchestration that make models actually useful. At GreatApeAI, we believe the harness is becoming more important than the model itself.
Should developers be worried about OpenAI owning uv?
The tools are permissively licensed, so the community can fork them if OpenAI neglects them. However, there's always risk when a single company controls critical infrastructure. History shows these acquisitions can lead to slower development or feature drift.
Sources
Primary sources (I actually read these):
Simon Willison's analysis - Read March 19, 2026
Astral announcement - Referenced from coverage
OpenAI announcement - Referenced from coverage
Secondary sources:
PyPI Stats for uv download numbers (126 million downloads last month)
Hacker News discussions on acquisition implications
Previous coverage of Anthropic's Bun acquisition (December 2025)