Impact of Meta's Pyrefly on the red-knot roadmap

[mgaitan]

As you may know, a few days ago Meta unveiled Pyrefly—A faster Python type checker written in Rust. The Hacker News thread and the GitHub repo have sparked lively discussion and raised a question many of us are pondering: how does Pyrefly's release affect Astral's Red-Knot project?

Have you had a chance to experiment with Pyrefly? Reviewed its source or assessed its current maturity? I'm especially curious whether any of its architectural ideas or solutions might inspire or accelerate parts of Red Knot's roadmap. Is there room for collaboration on libraries, benchmarks, or CI/CD integrations? Could joint knowledge-sharing sessions benefit both projects?

There's also a "political” angle: does Pyrefly's backing by a Big Tech company feel like a risk to Red Knot's adoption, or rather a sign that Python typing finally matters to the majors? Would a heads-up or even a joint conversation before launch have been welcome? Despite the surprise around Pyrefly, many in the Python community already expressed eagerness to wait for Red Knot's release, trusting Astral's track record of delivering outstanding tools. But does this new "competitor” add pressure to ship Red-Knot sooner?

Thanks for your dedication to advancing Python tooling. I look forward to your thoughts and any perspective you can share about this.

[T-256]

many in the Python community already expressed eagerness to wait for Red Knot's release, trusting Astral's track record of delivering outstanding tools. But does this new "competitor” add pressure to ship Red-Knot sooner?

I personally prefer Astral keep its signatures for red-knot project like its other projects: Correctness and performance!
(I already accepted the trade-off here would be the time.)

[charliermarsh]

Doesn't change much for us! Meta, Google, and Microsoft have all built Python type checkers in the past. We're just focused on building the best tool for our users.

(There's non-zero collaboration: Pyrefly uses our parser and AST infrastructure, and we've had some conversations with folks on the team about our differing approaches -- they did approach us early on to discuss collaborating, but then opted to do their own thing.)

[sa-]

I hope it works out and I'm rooting for astral here - your note about non zero collaboration reminds me of the story about Spegel