Tell HN: Help restore the tax deduction for software dev in the US (Section 174)
Apple Announces Foundation Models Framework for Developers to Leverage AI
apple.com
Show HN: Munal OS: a graphical experimental OS with WASM sandboxing
github.com
Apple introduces a universal design across platforms
apple.com
Launch HN: Chonkie (YC X25) – Open-Source Library for Advanced Chunking
Doctors could hack the nervous system with ultrasound
spectrum.ieee.org
What methylene blue can (and can’t) do for the brain
neurofrontiers.blog
Hokusai Moyo Gafu: an album of dyeing patterns
ndlsearch.ndl.go.jp
A bit more on Twitter/X's new encrypted messaging
blog.cryptographyengineering.com
Why quadratic funding is not optimal
jonathanwarden.com
Bruteforcing the phone number of any Google user
brutecat.com
The new Gödel Prize winner tastes great and is less filling
blog.computationalcomplexity.org
Algovivo an energy-based formulation for soft-bodied virtual creatures
juniorrojas.com
Show HN: Somo – a human friendly alternative to netstat
github.com
Show HN: Most users won't report bugs unless you make it stupidly easy
A man rebuilding the last Inca rope bridge
atlasobscura.com
Maypole Dance of Braid Like Groups (2009)
divisbyzero.com
Finding Shawn Mendes (2019)
ericneyman.wordpress.com
Potential and Limitation of High-Frequency Cores and Caches (2024)
arch.cs.ucdavis.edu
Show HN: Glowstick – type level tensor shapes in stable rust
github.com
The Legend of Prince's Special Custom-Font Symbol Floppy Disks (2016)
nymag.com
Object firing signals at Earth every 44 minutes
livescience.com
Hi HN,
In the past few years I've become more interested in machine learning. Since I'm sure the same is true for many here, I wanted to share this project I've been working on: glowstick uses type-directed metaprogramming to keep track of tensor shapes in Rust's type system and determine which operations are permitted or not at compile time.
I find Rust has a lot of strengths when it comes to ML applications, but waiting until runtime to find shape related issues feels a bit strange since normally I don't run the code all that often while developing. Given Rust has fancy types available, I figured I'd try my hand at using them to address this.
I've added integration crates for the two ML frameworks I use most frequently, candle and burn, and included examples of implementing llama 3.2 in each using typed shapes for much of the model internals and inference loop. Mixtures of static and dynamic dimensions should be supported well enough for most applications at this point, though there are of course still improvements to be made.
Any feedback is appreciated!