Open Source AI

Feeds to Scour
SubscribedAll
Scoured 728 posts in 10.2 ms

The latest Gemma 4 models use a training trick to slash their on-device memory footprint

 🔔Plan 9
androidauthority.com·

A new chapter of efficient foundation models for medical imaging

 🖥️Retro Computing

Researchers Build Self-Replicating AI Worm That Operates Entirely on Local, Open-Weight Models

 🔔Plan 9
thehackernews.com·

Re-quantizing a local LLM 14x faster by skipping the tensors that didn't change

 🖥️Retro Computing  Content type: News  Content type: Blog

Local LLMs, Buy a GPU, and the Case for Cognitive Security

 🖥️Retro Computing

Google Shrank Gemma 4 by 72% and Unsloth Fixed the 4-Bit Bug Nobody Else Caught on One 4090, and 4-Bit Shouldn’t Be This Good

 🖥️Retro Computing  Content type: Blog
towardsai.net·

Google releases Gemma 4 12B with encoder-free multimodal architecture

 🔔Plan 9
4sysops.com·

A system programmer’s guide to LLM inference

 🦬Emacs  Content type: Blog

magenta/magenta-realtime: Magenta RealTime 2: An Open-Weights Live Music Model

 Gemini Protocol  Content type: Code
github.com·

HNSW vs LSH: How Elasticsearch hits 0.99 recall@10 at 15,000 QPS — and what it costs

 🖧BSD  Content type: Blog
elastic.co·

local llm on laptop 780M GPU using llama + gemma 4 qat

 🦬Emacs  Content type: Blog
alper.bearblog.dev·

Optimal Post-Training Quantization Scales and Where to Find Them

 🔌Single-Board Computers  Content type: Academic
arxiv.org·

China's Xiaomi MiMo Is Now 15X Faster Than ChatGPT and Claude (4 minute read)

 Gemini Protocol  Content type: News
decrypt.co·

Build a Medical Report Analyzer on Dedicated Inference with Python

 Gemini Protocol
digitalocean.com·

"North Mini Code"; open weights, 30B param, Canadian coding model

 🔌Single-Board Computers  Content type: Blog
cohere.com··Hacker News

AI Serving Platform That Adapts to Your Model

 🏠Self-Hosting  Content type: Blog
databricks.com·

DiffusionGemma

 🖥️Retro Computing
simonwillison.net·

2x GH200 for LLM inference, Part 2: vLLM, DeepSeek V4 Flash, and MTP

 🖥️Retro Computing  Content type: Blog
dnhkng.github.io·

Researchers trained an open source AI search agent, Harness-1, that outperforms GPT-5.4 on recalling relevant information

 🖥️Retro Computing

WWDC 2026: Foundation Models (& Anarlog)

 🔔Plan 9
skushagra.com·
Sign up or log in to see more results

Keyboard Shortcuts

Navigation

Next / previous item
j/k
Open post
oorEnter
Preview post
v

Post Actions

Love post
a
Like post
l
Dislike post
d
Undo reaction
u
Save / unsave
s

Recommendations

Add interest / feed
Enter
Not interested
x

Go to

Home
gh
Interests
gi
Feeds
gf
Likes
gl
History
gy
Changelog
gc
Settings
gs
Browse
gb
Search
/

General

Show this help
?
Submit feedback
!
Close modal / unfocus
Esc

Press ? anytime to show this help