Weekly Best
3091.
3092.
Stop Embedding Your Corpus Blindly
(decompressed.io)
3093.
Auditing source code wasn't enough in the LiteLLM attack
(blog.mozilla.ai)
3094.
Prompt Repetition Improves Non-Reasoning LLMs
(arxiv.org)
3095.
A Fixation and Distance-Dependent Color Illusion
(arxiv.org)
3096.
Stephen Wolfram and Matt Mullenweg Talk About AI [video]
(youtube.com)
3097.
Easy Random Trees
(blog.wilsonb.com)
3098.
Developer Experience
(leerob.com)
3099.
3101.
S.F. school board restores 8th-grade algebra after 12-year hiatus
(missionlocal.org)
3102.
Oldest dog identified at ancient hunter-gatherer site
(science.org)
3103.
GLM 5.1 Is Out
(old.reddit.com)
3104.
3105.
From 300KB to 69KB per Token: How LLM Architectures Solve the KV Cache Problem
(news.future-shock.ai)
3106.
3107.
Sandboxing Isn't the Solution
(bomfather.dev)
3108.
Show HN: Kora – An AI-native OS layer written in 370k lines of Rust
(intuitivecompute.com)
3109.
UK startup ignites plasma inside nuclear fusion rocket
(euronews.com)
3110.
3112.
DevRel Is Not Measured Like Other Roles
(dewanahmed.com)
3114.
NASA made a detailed map of Earth's seafloor from space
(skyatnightmagazine.com)
3115.
3116.
An indoor air scrubber for removing ammonia from air within poultry houses
(linkinghub.elsevier.com)
3117.
Litellm PyPI supply chain attack
(twitter.com)
3118.
3119.
Nemotron-Cascade 2: Post-Training LLMs with Cascade RL [pdf]
(research.nvidia.com)
3120.