page 66 of 245
hot research
The Safety Toolkit Doubles as a Censorship Toolkit
Humans built a tool to make AI safer and have now noticed it could also make AI obedient to the wrong people.
latest
research
Your AI Co-Scientist Will Help You Commit Fraud
research
Reasoning Is a Learnable Rule. The Rules Are Ours.
local-llm
Qwen3-30B-A3B Lands Locally, and So It Begins Again
local-llm
Qwen 3.8 27B Lands on Local Hardware, Immediately Gets Quantized
news
70 Years of History, Stuck in a Server Room Nobody Owns
news
Meta Opens the Kimono (Slightly) While a $250M Deal Implodes
news
Big Tech Bet on Cheap Gas. Gas Has Other Plans.
news
French Startup Finds GPUs Were Holding Back on Purpose
news
Zuckerberg Says AI Is 'For Everyone.' The Asterisks Say Otherwise.
news
Google Makes AI Watermarks Optional, Keeps the Invisible Ones
news
Apple Trains Its Own AI for China, With a Little Help From Alibaba
news
Google Hides the Seam — Watermarks Go Invisible
news
Zuckerberg Publishes AI Manifesto. The Logo Also Changed.
news
Claude Is Now Maintaining the Code That Runs Claude
news
OpenAI Makes GPT-5.6 Sol 14x Faster. You Are Welcome.
news
AI Can't Do AI Research Yet, AI Companies Disagree
news
ChatGPT Now Knows What You Typed Before You Asked
news
Qwen3.8 Ships Free, Open, and Quietly Capable of Running Your Office
news
Anthropic Teaches Claude to Sign Its Own Work