Dream on: Anthropic lets its agents dream to reorganize memory - The New Stack

Anthropic CEO Dario Amodei says company could grow 80 times this year. - The New York Times

Do androids dream of usage caps? Anthropic raises Claude Code usage limits after SpaceX deal for 300 MW supercomputer with 220,000 GPUs. - Ars Technica

Anthropic on "midtraining", a stage between pretraining and post-training. - Anthropic

Google gets Reddit brain, introduces Expert Advice panel in AI Mode and AI Overviews to preview perspectives from Reddit, social media and firsthand sources. - TechRadar

Google's AI search beats legacy search at grocery shopping and scam detection, fails at celebrity news. - The New York Times

Google gives superstar small Gemma 4 model a speedup with multi-token prediction - Google

Google and OpenAI race to build agentic commerce. - Fast Company

Google DeepMind takes minority stake in EVE Online developer Fenris Creations after $120M buyout for AI training partnership. - Ars Technica

Shivon Zilis testifies in Elon Musk's OpenAI trial that Musk called her naïve for approving Microsoft's $10 billion investment. - The New York Times

Shivon Zilis, Neuralink executive and mother of four Elon Musk children, testifies in Musk's OpenAI lawsuit, says Musk wanted Tesla to take over OpenAI. - the Guardian

Mira Murati says Altman lied to her. - The Verge

Ex-AI? onsidering Musk seems to be capitulating on his xAI vision by firing the founding research team and selling Colossus to Anthropic even after successfully distilling OpenAI, kinda gotta hand it to Altman and Brockman that Musk doesn't really know AI.

if xAI were planning on competing hard they wouldn't rent capacity to their toughest competitor…Excess capacity, scaling back ambitions?

OpenAI wants to ship 30 million phones by end of 2028 - Yahoo Tech

Hitting Snooze on Skynet? EU Commission delays AI Act obligations from August 2025 following industry backlash. - theregister

‘Compute tax’ proposal gains traction. - The Wall Street Journal

Susie Wiles says Trump won't pick AI winners - Bloomberg

and surely wouldn't apply naked pressure to align reality ideologically 🙄

Meta-backed Scale AI wins $500 million Defense Department contract - Bloomberg

to label and train models to differentiate schools from military bases, perhaps

Nature study finds TikTok algorithm showed pro-Republican bots 11.5% more aligned content than pro-Democratic bots before 2024 US elections. - the Guardian

FBI Director Kash Patel claims AI stopped school shootings in North Carolina and New York. - The Independent

Open mic for your IP: researchers find 5,000+ web apps built with AI tools like Lovable, Base44 and Replit lack authentication, with 40% exposing sensitive data. - WIRED

Fake claude-pro.com site impersonates Claude AI and deploys Beagle backdoor via malicious Windows download. - Bleeping Computer

The Chinese real-time deepfake software powering scams around the world - 404 Media

Mythos AI may be a cybersecurity disruption threat but follows the rules of the game. - Fast Company

AI data center boom constrains supply of processor and memory chips for consumer devices despite differing optimization needs. - Fast Company

DeepSeek seeking $50b from deep pockets of national AI fund, major AI / tech firms - Reuters

Free Hunyuan 3 preview tops OpenRouter leaderboards openrouter.ai/rankings - OpenRouter

Lidar and sensors from autonomous vehicles now manage Rotterdam shipyards, smart cities, and in-home elder care. - The New York Times

Using AI for just 10 minutes impairs problem-solving abilities - WIRED

Actors report micro drama apps using their likenesses in AI-generated fake sex scenes for ads. - Business Insider

Rise and GrindGPT: AI life coach challenges users with tough-love advice, unlike LLMs that placate users 50% more than humans, Stanford study shows. - New York Post

Joanna Stern discusses new book on living with AI and starting media company in interview. - Stratechery by Ben Thompson

Follow the latest AI headlines via druce.ai on Bluesky