frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Open in hackernews

Show HN: Shoehorn, a library to quantize an LLM to fit your Mac's VRAM

https://github.com/notactuallytreyanastasio/shoehorn
5•rhgraysonii•38m ago
I made this after seeing someone posit the idea online yesterday over lunch then spent some time refining it. So far it's pretty impressive IMO! Right now I am running Qwen3-30B-A3B on my 24gb unified memory m4 MacBook Pro at 50 tok/sec and this should definitely not be working for such a large model on my middling hardware.

Things are detailed in the README to get up and running and DESIGN.md has details on all the choices and such made along the way.

Personal Software Is Getting Cheaper to Build, So I Killed My LinkedIn Feed

https://andrewpollack.dev/blog/fixing-linkedin-customizing-the-web/
1•andrewpollack•23s ago•0 comments

Prefill vs. Decode in LLM Inference

https://www.parasail.io/blog/prefill-vs-decode-llm-inference
1•eatonphil•24s ago•0 comments

'The Nerd Reich' tracks the 'unmasking of Silicon Valley's true politics'

https://www.npr.org/2026/08/10/nx-s1-5925350/the-nerd-reich-tracks-the-unmasking-of-silicon-valle...
1•vegetablepotpie•27s ago•0 comments

Show HN: Graft – Claude Code hooks that cut grep tokens by 42%

https://github.com/NanoNets/Graft
1•shrishdwi•50s ago•0 comments

Nigel Farage beats slate of fringe candidates in election

https://apnews.com/article/britain-politics-farage-binface-clacton-election-reform-uk-16f5c7deff9...
1•anonymousiam•59s ago•0 comments

In South Korea, some savour last season of legal dog meat

https://www.reuters.com/world/asia-pacific/south-korea-some-savour-last-season-legal-dog-meat-202...
1•speckx•1m ago•0 comments

AV1 vs. AV2: Which Video Codec Should You Use in 2026?

https://www.red5.net/blog/av1-vs-av2/
1•mondainx•1m ago•0 comments

How fast is C++26's std:hive?

https://lemire.me/blog/2026/08/02/how-fast-is-c26s-stdhive/
1•surprisetalk•2m ago•0 comments

European publishers are getting hit harder by AI bot scraping, report finds

https://digiday.com/media/european-publishers-are-getting-hit-harder-by-ai-bot-scraping-report-fi...
1•thm•5m ago•0 comments

Postgres SELECT DISTINCT does not scale

https://www.dbos.dev/blog/postgres-select-distinct-does-not-scale
2•KraftyOne•6m ago•0 comments

NTU to stop using 'fundamentally unreliable' AI detector from 2027

https://www.channelnewsasia.com/singapore/ntu-ai-detector-stop-using-2027-probability-scores-misc...
2•ahonhn•6m ago•0 comments

Private prisons announce $1.4B in revenue as immigration detentions climb

https://www.npr.org/2026/08/08/nx-s1-5923881/geo-group-corecivic-immigration-detention
2•speckx•6m ago•0 comments

What caused man to be partially sucked out of Ryanair plane

https://news.sky.com/story/revealed-what-caused-man-to-be-partially-sucked-out-of-ryanair-plane-1...
2•austinallegro•9m ago•0 comments

Unforgetful

https://marco.org/2026/08/14/unforgetful
2•robbiet480•10m ago•0 comments

North Koreans Infiltrate U.S. Companies [video]

https://www.youtube.com/watch?v=RUhdmIFpb-w
2•onemoresoop•11m ago•0 comments

Apple sends new 'Threat Notification' alerts over mercenary spyware attacks

https://www.bleepingcomputer.com/news/apple/apple-sends-new-threat-notification-alerts-over-merce...
2•DemiGuru•12m ago•0 comments

2020 Arctic Vault Program

https://archiveprogram.github.com/arctic-vault/
2•EndXA•12m ago•0 comments

Introducing Toast 1

https://www.mixedbread.com/blog/toast-1
4•mplappert•14m ago•0 comments

Show HN: I trained three LLMs from scratch and put them online to try them out

https://huggingface.co/spaces/JohnEnev/modern-llm-playground
2•Johnene•16m ago•1 comments

You Need Less Money Than You Think to Be Happy: A $40k Experiment

https://www.financialsamurai.com/less-money-than-you-think-to-be-happy/
2•timvdalen•17m ago•0 comments

When the Hard Part Stops Being Hard

https://proofsandintuitions.net/2026/08/14/when-the-hard-part-stops-being-hard/
3•noelwelsh•17m ago•0 comments

What to do when a macOS update goes wrong

https://eclecticlight.co/2026/08/14/what-to-do-when-a-macos-update-goes-wrong-2/
2•Brajeshwar•17m ago•0 comments

Unsloth Qwen3.8-27B GGUF files

https://huggingface.co/unsloth/Qwen3.8-27B-GGUF
8•apitman•18m ago•0 comments

Qwen3.8-27B

https://twitter.com/alibaba_qwen/status/2088280182356611304
65•mfiguiere•18m ago•21 comments

Seven books I keep close because I love them

https://blog.plover.com/2026/08/02/
5•surprisetalk•19m ago•0 comments

Unsloth - Qwen3.8 - How to Run Locally

https://unsloth.ai/docs/models/qwen3.8
3•apitman•19m ago•0 comments

Self-hosted web push Cloudflare Worker, works on iOS

https://kukuroo.cc/
3•saiday•19m ago•1 comments

Qwen3.8-27B is now available on Hugging Face

https://huggingface.co/collections/Qwen/qwen38
10•oger•20m ago•2 comments

Demographic Dividend, my ass; overpopulation sucks

https://shubhamjain.co/2026/08/13/demographic-dividend-overpopulation-sucks/
4•shubhamjain•22m ago•0 comments

Dungeon Survivors – A bullet-heaven roguelite built in C# and MonoGame

https://lholhe.itch.io/dungeon-survivors-early-access
2•devejesus•22m ago•0 comments