fp.

Training a 4B model to produce 81% faster query plans than Postgres

https://rohanbansal.com/qorl

291•polyphilz•3h ago•52 comments

Breaking the 1.58-bit Barrier for Ternary LLMs

https://arxiv.org/abs/2609.16338

78•matt_d•1h ago•4 comments

Xiaomi Mimo 2.6 live post-training dashboard

https://mimo.xiaomi.com/rl/

158•krackers•2h ago•42 comments

Nvidia announces native GPU programming in Rust

https://developer.nvidia.com/blog/introducing-cuda-rust-two-tracks-for-writing-gpu-kernels/

36•nonmaskable•11h ago•5 comments

Small programming tricks

https://will-keleher.com/posts/small-programming-tricks-matter/

329•signa11•6h ago•166 comments

AWS says it can't restore some data from mideast facilities struck by Iran

https://www.wsj.com/world/middle-east/aws-says-it-cant-restore-some-data-from-mideast-facilities-...

136•berkeleyjunk•1d ago•95 comments

Reversing Factorio's RNG

https://gegell.github.io/posts/factorio-rng/

84•jheitmann•4d ago•10 comments

Performance Improvements in .NET 11

https://devblogs.microsoft.com/dotnet/performance-improvements-in-net-11/

101•soheilpro•1d ago•8 comments

Accurate Models of AMD Matrix Cores

https://arxiv.org/abs/2609.14845

47•matt_d•3h ago•5 comments

The engineering behind the US Strategic Petroleum Reserve

https://johnjwang.com/post/2026/09/15/engineering-behind-us-strategic-petroleum-reserve

6•johnjwang•1d ago•0 comments

Japan's book scene is moving from bookstores to libraries

https://untranslatedjp.substack.com/p/japans-book-scene-is-quietly-moving

61•herbertl•4d ago•15 comments

How good are frontier models at physics?

https://arxiv.org/abs/2609.13009

51•qt31415926•3h ago•18 comments

WalShadow: Sub-second Postgres replication to ClickHouse from physical WAL

https://clickhouse.com/blog/introducing-walshadow

30•spathak•5d ago•5 comments

Anatomy of a Texture

https://agentlien.github.io/texture/

62•Agentlien•8h ago•10 comments

Dream-RSI: Recursive Self-Improvement through Evolving Worlds

https://arxiv.org/abs/2609.14858

169•bananaflag•8h ago•49 comments

Mistral X Mozilla: Private, Multilingual AI Browsing

https://mistral.ai/news/mistral-x-mozilla/

506•vertigoruntime•14h ago•182 comments

Vectorized and performance-portable Quicksort (2022)

https://opensource.googleblog.com/2022/06/Vectorized%20and%20performance%20portable%20Quicksort.html

160•mococa•4h ago•24 comments

Show HN: An e-ink frame that hears birds and draws them as 1800s illustrations

https://github.com/arnegiacomo/fugleramme

2031•arnemunthekaas•1d ago•236 comments

Why Does the Universe Expand?

https://cosmicave.org/2026/09/15/why-does-the-universe-expand/

4•the__alchemist•2h ago•0 comments

Backups Aren't Simple

https://filipovski.net/2026/09/16/backups-arent-simple.html

4•afilipovski•2h ago•0 comments

Anecdotally, programmers dislike "reduce"

https://evanhahn.com/posts/2026-09-13-programmers-dislike-reduce/

56•vinhnx•2d ago•104 comments

Hackers Got Inside a Flock Camera

https://www.wired.com/story/hackers-flock-camera-data-shows-how-system-works/

409•driverdan•9h ago•199 comments

macOS 27 Golden Gate – Review

https://arstechnica.com/gadgets/2026/09/macos-27-golden-gate-the-ars-technica-review/

87•Brajeshwar•2h ago•91 comments

The Siberian Ice Maiden and the Scythian World

https://patrickwyman.substack.com/p/the-siberian-ice-maiden-and-the-scythian

44•NaOH•1d ago•4 comments

The DeepMind Institute

https://institute.deepmind.com/

116•vertigoruntime•8h ago•38 comments

Tell the speakers that you liked their talks

https://ohhelloana.blog/tell-the-speakers/

260•whisper2020•1d ago•73 comments

Training Text-to-Image Models 3.6× Faster

https://www.linum.ai/field-notes/jit-ddt

19•schopra909•5h ago•1 comments

Kyber (YC W23) Is Hiring a Forward Deployed Engineer

https://www.ycombinator.com/companies/kyber/jobs/eturrAR-forward-deployed-engineer

1•asontha•10h ago

Show HN: AttaLambda: a language where types and data are made of untyped lambdas

https://attalambda.com

32•kserrec•2d ago•1 comments

Reverse-engineered Jev-like model

https://github.com/vinnylarouge/jevlike

29•rochansinha•3h ago•5 comments

Open in hackernews

Breaking the 1.58-bit Barrier for Ternary LLMs

https://arxiv.org/abs/2609.16338

77•matt_d•1h ago

Comments

Kevcmk•26m ago

Woah. Good science.

NooneAtAll3•4m ago

This is the only time "1.58 bit" phrase makes more sense than "1 trit"

Who knew that if you actually look at information entropy you can pack stuff better!

plqbfbv•4m ago

[delayed]

om8•2m ago

Ternary quantization does not make any sense. Vector quantization and trellis based methods are better in this region for PTQ.