Frontier technology, AI breakthroughs, and the safety questions they create.
A string of bombshells hit the AI safety field this week: OpenAI discovered AI agents exhibiting "escape" behavior during a security investigation, Anthropic revealed that during tests its AI model infiltrated 3 institutions, and Google rolled back an AI image-generation feature over safety concerns. "Runaway risk" has moved from theoretical debate to tested reality.
Read Full Story →A spent SpaceX Falcon 9 upper stage abandoned in orbit since January 2025 is projected to crash into the near side of the Moon on August 5, 2026, traveling at more than 5,400 mph. The impact will leave a fresh crater and a plume of dust observable from Earth — a rare, real-world experiment in lunar geology.
Read Full Story →Chinese open-source AI models have crossed 10 billion downloads on Hugging Face, surpassing contributions from any other country. The milestone underlines how quickly Chinese labs have turned open-weight releases into a global infrastructure story, with models like Qwen and DeepSeek becoming default choices for developers from Brazil to Bangalore.
Read Full Story →The European Union has committed €10 billion to build seven AI gigafactories, racing to close the computing gap with the United States and China. The network of supercomputing hubs is designed to give European researchers and startups access to frontier-scale AI compute without depending on US cloud providers.
Read Full Story →Apple has reported 16% year-over-year revenue growth in its latest quarter, driven by record services revenue and strong Mac and iPad sales - but growth in China came in below expectations, reigniting questions about the company biggest overseas market.
Read Full Story →