Wonder Years? Meta launches Muse Wonder and Muse Wonder Full, claiming its closed models rival OpenAI's and Anthropic's best at coding, law and finance - The New York Times
not quite frontier per Artificial Analysis, on a par with Sol possibly but much more expensive
New Active Memory and Active Focus features let agents carry context across conversations and recover from errors on long tasks - Meta AI Research
Mark Zuckerberg says Meta was ready to spend $100–200 billion on AI before knowing whether its new models would work - Business Insider
Hacker group CyberXero used Claude Code with 51 agents to steal over 628,000 Ukrainian records, hitting a Kharkiv heating provider - The Hacker News
Anthropic says Z.ai's freely downloadable GLM-5.3 model poses a major cyberattack risk, reigniting the open-versus-closed debate with Nvidia's Jensen Huang - MIT Technology Review
Cold War, Hot Chips: Goldman Sachs sees a US-China "AI duopoly" emerging as Chinese AI stocks add nearly $5 trillion in market value since early 2025 - Business Insider
Michigan Senate nominee Abdul El-Sayed proposes banning superintelligent AI and creating a cabinet-level AI agency, calling it "the existential crisis of our time" - The New York Times
npm Trusted Publishing flaw left over 1,500 packages open to hijacking, a risk amplified by AI agents that install dependencies unreviewed - TNW | Contributed
NYT, The Economist, AFP and The Atlantic join a publisher coalition of 80+ members building standards to track and license AI use of content - The Straits Times
Drop bots: Ukraine's Third Army Corps says it is first to air-drop armed ground robots from drones in combat, six miles into Russia - Business Insider
Russia's defense ministry shares drone combat footage with more than 40 organizations training military AI models - Sputnik Globe
Ondas and Palantir launch Regna AI to run fine-tuned open-weight models on drones and other military edge hardware - Ondas Inc.
Rust upstart reveals 'reimplementation' of Microsoft Office – with Claude on coding duty - theregister
Mayo Clinic's REDMOD AI spots pancreatic cancer on CT scans about 16 months earlier and nearly twice as often as specialist radiologists - Intelligencer
AI pipeline rebuilds 1,328 finance studies: 75% replicate on original data, but only 42% hold up on later data, suggesting data snooping - Alpha Architect
Flash Crash Bandicoot: Robinhood expands AI agents that trade around the clock, optionally without human approval, raising herding and flash-crash worries - Advisor Perspectives
R teams partner with security specialists as fraudulent AI use in job applications rises - Financial Times
AI automates entry-level work at Wall Street banks' Indian tech hubs; Citigroup's net India hiring halves to 2,000 this year - Seattle Times
AI-assisted software development could have liberated programmers. Instead, it has become the latest way for Big Tech to confine us. - The Atlantic
It’s not business, it’s personal: AI godfather Geoffrey Hinton: Trump is ‘corrupt, senile felon’ who knows nothing about AI - the Guardian
Smok’em if you got’em: AI researchers are taking up smoking because they expect AI to cure cancer, says Palantir CTO - Business Insider
The Muskening - X (formerly Twitter)
Elon Musk says Waymo "runs over a lot of pets"; NHTSA data shows nine Waymo animal crashes in a year, one for Tesla - Business Insider
Elon Musk tweets about Ilhan Omar’s immigration status investingation, but he was violating his visa when he started his first business here and very likely obtained his U.S. citizenship through fraud. Where is that investigation?
Follow the latest AI headlines via druce.ai on Bluesky
