- 1Attention Sinks and Compression Valleys in LLMs (arxiv.org)
- 1Part 2 - AI Chat Evaluation of the Formal Language in He Xin's PEPC System ()
- 1Hand tool rewrites ancient Egyptian history (popsci.com)
- 2A note about personal security (werd.io)
- 1Part 1 - AI Chat Evaluation of the Formal Language in He Xin's PEPC System ()
- 1A Note on File History in Emacs (brainbaking.com)
- 1Revisionist History – Aliens, Secrets and Conspiracies (steveblank.com)
- 1Show HN: cbt (C++ Build Tool) (github.com)
- 1Open model StepFun-3.5 is #1 on MathArena, an uncheatable math benchmark (twitter.com)
- 1Show HN: Bitcoin, GEB, and Bach's fugues share the same structural move (falsework.dev)
- 1Functional Programming in M4 (minnie.tuhs.org)
- 2AI makes it easier to build the wrong thing faster (newsletter.masilotti.com)
- 1Show HN: I built a macOS desktop toy that patrols while you work (airwolfspace.com)
- 1Poison at Play: Unsafe lead levels found in half of New Orleans playgrounds (veritenews.org)
- 4Unresponsive Buttons on My Fastest Hardware (blog.jim-nielsen.com)
- 85AI-First Company Memos (the-ai-native.company)
- 1How to Test ProxySQL Read/Write Split with Sysbench (rendiment.io)
- 5The singularity won't be gentle – by Nate Silver (natesilver.net)
- 1A New Computer Could Replace Electricity with Light (popularmechanics.com)
- 1Show HN: Health.md - Apple Health → Markdown (healthmd.isolated.tech)
- 1PicoClaw: Ultra-Efficient AI Assistant in Go (github.com)
- 1AITools.coffee – GitHub metrics observatory tracking 27K+ open-source AI repos (aitools.coffee)
- 3Show HN: AI Agent Built From Scratch (No Frameworks) (medium.com)
- 4Databases should contain their own Metadata – Use SQL Everywhere (floedb.ai)
- 3Seeking Order in Chaos (garrit.xyz)
- 2Show HN: Funxy – A typed scripting language that embeds into Go apps (github.com)
- 1The jarring experience of developing today (its.beer)
- 5Kiro: DeepSeek, MiniMax, and Qwen now available as open weight model options (kiro.dev)
- 1Terence Tao: Why I Co-Founded SAIR (youtube.com)
- 1Maia 200: The AI accelerator built for inference (blogs.microsoft.com)