NumPy in rust now exists folks. Let the science people know.
07.03.2026 19:54
👍 66
🔁 8
💬 3
📌 0
NumPy in rust now exists folks. Let the science people know.
i built an entire x86 CPU emulator in CSS (no javascript)
you can write programs in C, compile them to x86 machine code with GCC, and run them inside CSS
lyra.horse/x86css/
It is absurdly improbable that you can hoover up the internet, shred it, then talk to the mulch pile and it talks back.
The Douglas Adams age of technology (2024)
interconnected.org/home/2024/02...
What do LLMs see?
I wrote a lil' tool that extracts the attention matrices out of open models and creates this typing visual, with each token's opacity changing according to its average attention score as the prompt progresses. Dimmer words are considered less important to the model.
"you can just use embedded rust libraries off the shelf for GPUs because it's just a normal no-std library, duh" is SUCH a huge validation for how we've structured the embedded Rust ecosystem.
No OS to port, no complex tooling to set up. If you can run rust code, you can just drop no_std crates in.
we live in the ruins of a greater civilization
Mr. C++
Usually in this last situation it would require a few iterations of agent review, human review, simplification, etc.
Also, there are probably many teams out there where the "review pipeline" is already almost a bottleneck, even without LLM agents.
Oh, I mean, it's debatable more in the sense that it really depends on the situation and context.
Maybe the agent will help fix a super hidden bug with changes in 3 lines of code (trivial review), maybe it will generate tons of boilerplate with complex interactions that would require a long review.
/model claude-opus-4-5-20251101
"We choose to go to the fucking Moon in this fucking decade and do the other fucking things, not because they are fucking easy, but because they are fucking hard."
Another argument for up:
- This fact will create more demand for Rust, thus for Rust developers
Now, both arguments for down are debatable. AI only helps learning if the developer really really wants to learn, the "easy mode" it creates is to not learn much and just depend on the AI tools.
Rust has a linear algebra problem.
See, linear algebra is super important for many of the applications where Rust is most commonly used. Yet, we have an underwhelming set of linear algebra libraries.
Matthew Treinish of IBM Quantum explains why they're underwhelming 👇
![Comic. Conjecture: It’s possible to construct a convincing proof without words, pictures, or content of any kind. Proof: [empty box] [caption] Proofs without words are cool, but we can go further.](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:cz73r7iyiqn26upot4jtjdhk/bafkreiaukmlmzx6xhbcjej4rq5axkf7a25sjywnvodd53dgnebmt4ver5a@jpeg)
Comic. Conjecture: It’s possible to construct a convincing proof without words, pictures, or content of any kind. Proof: [empty box] [caption] Proofs without words are cool, but we can go further.
Proof Without Content
xkcd.com/3201/
Zlib-rs is now feature-complete! We've released v0.6, the first version with a stable and complete API. The blog post has the details.
With thanks to our maintainer Folkert de Vries, our contributors and @sovereign.tech.
trifectatech.org/blog/zlib-rs...
#rust #rustlang
nextest can now rerun failing tests!
cargo nextest self update --version 0.9.123-b.3
# enable functionality, see link
cargo nextest run
If tests fail and/or the run is cancelled, then:
cargo nextest run -R latest
to rebuild and pick up from where the test run left off.
Please test this out!
Oops, I started a new project: Skyreader, an RSS reader on the AT Protocol. Share cool articles like it's 2010 and Google Reader would never die. skyreader.app
www.disnetdev.com/blog/2026-01...
Who here nostalgic for Reader?
Continuing to have AI build a weird game demo a day. Here is: "Make a game where you have to prevent the apocalypse, but the interface is just Jira tickets"
Pretty fun/funny branching storyline, all text is AI created with minor feedback from me. Play: gentle-bienenstitch-01e24b.netlify.app
wikipedia turns 25 today! the last unenshittified major website! backbone of online info! triumph of humanity! powered by urge of unpaid randos to correct each other! somehow mostly reliable! "good thing wikipedia works in practice, because it sure doesn't work in theory" - old wiki adage
@anthropic.com is investing $1.5 million in the PSF, focused on security. These funds will make an enormous impact on the PSF and the security of millions of #Python and @pypi.org users. Please join us in thanking Anthropic for this landmark gift!
Read more on our blog:
Three hours, 2.2 million tokens later with mostly passive prompting hours in the evening, and some overnight "continue" prompts and minijinja is fully ported to go. I linked the single pi session that has all prompts in it. shittycodingagent.ai/session/?29f...
Another Erdos problem this morning:
(just to respond to a few people-- the system does NOT work by trying every possible answer and then checking. There's not enough matter and energy in the universe to solve theorems by trying every possible combination of symbols or whatever)
Plot showing linear speedup with number of cores for different interpn calc methods.
The `interpn` library for Rust and Python has the fastest interpolation algorithms I'm aware of - up to 250x faster than Scipy - and today it got faster!
After years of focusing on single-thread performance, I finally called it good gave it more cores - and the parallel speedup is nearly linear!
On behalf of the packaging maintainers, I’d like to announce packaging 26.0rc1 is out! Please try it out, as it's a huge release. If you'd like to read about the performance work making this the fastest version of packaging ever, see my post:
iscinumpy.dev/post/packagi...
but Swift rounds toward zero to avoid issues of getting results in the wrong quadrant
the one from Rust, 0X40490FDB, is the correctly rounded one and is closer to real pi
