4
Section 4: The Rust-Powered Infrastructure Revolution 10:15 Miles: This is probably the most significant "structural" trend in the entire ecosystem right now. If you are a developer in 2026, and you are using the top-tier tools, you are actually running a lot of Rust code, whether you realize it or not.
10:30 Lena: It is everywhere. I was looking at a report from Uvik Software that estimated between twenty-five and thirty-three percent of all new native packages on PyPI are now being written in Rust instead of C. That is a massive shift from just a few years ago.
10:44 Miles: And it is being driven by some of the most fundamental tools we use. Take Astral, the company behind `uv` and `Ruff`. They have basically rewritten the "developer experience" layer of Python in Rust. And it was such a big deal that OpenAI actually acquired Astral in March 2026.
11:01 Lena: Wait, OpenAI bought a tooling company? That really signals how strategic this stuff has become. If you are building massive AI models, you can't afford to have your developers waiting for `pip` to resolve dependencies for ten minutes.
4:04 Miles: Exactly. And `uv` is the poster child for this. It is a Rust-based package, project, and Python version manager that is—and I am not exaggerating—ten to a hundred times faster than `pip`. It replaces `pip`, `pip-tools`, `virtualenv`, and even some of what `Poetry` does, all in one binary. It is so dominant now that `Rye`, which was another experimental manager, was archived just a couple of months ago, in February 2026, with everyone being told to migrate to `uv`.
11:42 Lena: I have switched to `uv` for my personal projects, and the speed difference is jarring. It is like going from a dial-up modem to fiber. And `Ruff` is the same story for linting and formatting, right?
11:53 Miles: `Ruff` is incredible. It replaces about six or seven different tools—`flake8`, `Black`, `isort`, `pyupgrade`—and it does it all in a single pass that is often a hundred times faster. Most big projects like `Pandas`, `FastAPI`, and `SciPy` have already switched to it. It has essentially consolidated a very fragmented part of the Python stack into one incredibly fast Rust engine.
12:15 Lena: It is interesting because this "Python API, Rust engine" pattern seems to be the secret sauce for Python's longevity. We get the beautiful, readable syntax of Python on top, but the heavy lifting, the "boring" stuff like dependency resolution or linting millions of lines of code, is handled by compiled Rust.
12:33 Miles: And it is not just tooling. We are seeing it in data libraries like `Polars` and `Pydantic v2`. `Pydantic v2` is five to fifty times faster than the original version because its core was rewritten in Rust. Since `Pydantic` is the backbone of things like `FastAPI` and most modern AI agent frameworks, that performance gain ripples through the entire ecosystem.
12:56 Lena: It really changes the "Python is slow" narrative. If your "slow" parts are actually highly optimized Rust, then for ninety-nine percent of use cases, Python is more than fast enough.
4:04 Miles: Exactly. And even the type checking is heading this way. `mypy` is still the most deployed, and `Pyright` is the favorite for IDEs like VS Code, but Astral has an alpha tool called `ty` which is a Rust-based type checker. It is not production-ready yet—it is on our "Watch List"—but the goal is to have type checking that is fast enough to run without even needing a cache.
13:27 Lena: It is a brave new world for the toolchain. But as much as I love talking about how fast my linter is, I think we need to look at the "other" big performance story—the Just-In-Time compiler in CPython itself. I saw some news recently that the 3.15 JIT is actually "back on track."