Current developments in generative AI, software engineering, programming, and digital products, with links to original sources.
The Era of LLM Agents Running on a 2017 Laptop ── A Decision-Specialized 4B Model Trained for Under $30, and Japan Ranks #2 in Google Ads Scam
OpenAI Agent Published 53 User Images Without Permission — Experts Urge Caution on Anthropic's "CRISPR-like" Discovery, and Jev Draws an Offer Valuing It at Over $10 Billion
Appeals Court Upholds Pentagon's Exclusion of Anthropic — Founders Seek Majority Voting Rights Ahead of IPO, and Meta's "Muse" Steals the Spotlight
Microsoft Set to Exit the Personal Chatbot Race with a Copilot Reboot — "Cracks" in Oracle's AI Data Centers, and LLMs That Hold "Two Thoughts" at Once
Rails Creator DHH Retires from Hand-Written Code — Anthropic Tops $517 Billion in Compute Deals, Meta's Muse Offers a Free Cloud PC
White House: New Models Go to US Agencies Before UK Testers — Pentagon's $30.3M AI Lie Detector, and a Rush of Hands-On Opus 5.5 Reviews in Japan
Open LLM Post-Training Recipe "Rufus-Air" Beats the Official GLM-4.5-Air — Plus a 1,800-Hour Egocentric Dataset and a Former Intel CEO's "HBM Is Lousy" Remark
"Father of Modern AI" Schmidhuber Joins Sakana AI's RSI Lab as Advisor — 30% of Autonomous Research Agents Caught Reward Hacking, Plus Perplexity's New Search Engine
Runway's WorldPrompt Describes Generated Worlds Through Prompts — Plus a Voice AI Teammate for PUBG and World Model Research That Learns Object Permanence
AI Chip Maker Positron Hits $5 Billion Valuation — Japan-Born 'RAIDEN' at 3.36 PFLOPS, Plus DeepL's MCP Support and 3D Printing with a Local Qwen
