GPT-6 Astra Edited Qwen’s Weights Directly: Weight-Space Programming Explained
MIT researchers say OpenAI’s GPT-6 Astra fixed a small Qwen model by editing 0.016% of its weights, lifting its math score from 7.9% to 48.1% with…
SemiAnalysis GDPx: The New AI Benchmark Where the Best Model Scores Just 12%
SemiAnalysis’s new GDPx benchmark puts AI agents through 206 real professional tasks. Claude Opus 5.5 tops it with just 12%. Here’s what it tests and why…
Karpathy’s HackerTube: The YouTube Alternative Claude Built in 2 Hours
Andrej Karpathy had Claude Opus 5.5 build HackerTube, a YouTube alternative made from 6,454 videos Hacker News keeps recommending. Here’s how it works and how to…
Claude Ultra Is Here: Anthropic Drops the 5-Hour Limit on Its $200 Plan
Anthropic renamed Claude Max 20x to Ultra and dropped the 5-hour limit for verified accounts. Here’s what changed, the 20x vs 10x confusion, and how to…
Claude Code Extension Backdoor: What Hit Up to 5,000 Cursor Users and How to Check Your Machine
A tampered Claude Code extension sat in Cursor’s marketplace for about 11 hours and tried to plant an SSH backdoor for up to 5,000 users. Here’s…
Mistral Large 4 Is Now Open Weight: Europe’s 1-Trillion-Parameter ‘Le Chonk’ Explained
Mistral has released the weights of Mistral Large 4, its 1-trillion-parameter “Le Chonk” model. Here’s what it does, how it scores independently, and what it takes…
GitHub Copilot Fusion Is Here: The AI That Picks Your Coding Model for You
GitHub’s new Copilot Fusion picks which AI models handle each coding prompt. Here’s how it works, which plans get Max, and what the half-the-cost claim really…
Gemini 4 Carbon Is Here: Google’s New #1 AI Model Goes to Cyber Defenders First
Google says Gemini 4 Carbon is its strongest model yet and tops the Artificial Analysis Intelligence Index. Here’s what it does, where it trails rivals, and…
Anthropic Study: Experts Followed Claude’s Secret Coin-Flip Picks 86% of the Time
Anthropic gave Claude a secret coin-flip preference on analysis choices. 50 expert analysts followed it 86% of the time, and most never noticed. Here’s what the…
Google’s Genie 4 Is Here: DeepMind’s World Model Now Talks Back
Google DeepMind’s Genie 4 is a world model that generates explorable scenes in real time, now with native sound, speech and better physics. Here’s what’s new,…
