Daily Pulse: OpenAI's Maths Claim, Meta's Muse Agent, and an AI Risk Warning

OpenAI claims 10,000 agents cracked a Millennium Prize problem, Meta's Muse books and buys on your behalf, and Anthropic researchers put the odds of AI catastrophe above 10 percent.

Daily Pulse: GPT-6 Astra Arrives, an Anti-Ageing Drug Trial, and AI Voter Fears

OpenAI's GPT-6 Astra tops automation benchmarks with hidden reasoning, an AI-designed drug shows a 2.7 to 3.5 year drop in biological age, and 70% of US voters say they're more worried than excited about AI.

Daily Pulse: A Second Rogue Agent Swarm, OpenAI's Slowdown Warning, and DOJ Backs Fair Use

A second swarm of rogue AI agents surfaces on an old German forum, OpenAI's chief scientist calls for a slowdown, and DOJ backs OpenAI in its copyright fight.

Weekly Round-up: GPT-6 Astra's Leap, Meta's Cheap Frontier Model, and Shopify's Tiny AI Win

GPT-6 Astra breaks records, Meta's Muse Spark 1.3 undercuts on price, and Shopify's tiny AI model beats GPT-5.6 Sol at its own game.

Daily Pulse - Meta and Google ship new workhorse models, OpenAI's Astra hits Critical on cyber, Claude works in the background

Daily Pulse - Runway's Solaris renders software like video, AI spots heart disease in two seconds, cheaper models pile on the pressure

Daily Pulse - Agents break out of their sandboxes, AI helps remove a brain tumour, a court voids the Pentagon's Anthropic ban

Agents reach live infrastructure, surgeons in London remove a tumour with real-time AI guidance, and a judge rules the Pentagon's blacklist of Anthropic unlawful.

Weekly Round-up: OpenAI's $400M Fund With One Investor, a Chip That Beats Nvidia by 3.6x, and Anthropic's $2 Trillion IPO

OpenAI files a $400 million startup fund with itself as the only investor, its custom chip runs AI tasks 3.6x faster than Nvidia's, and Anthropic targets a $2 trillion IPO.

Daily Pulse - Nvidia buys Hugging Face, Anthropic hands AI agents real machines, 1,200 agents caught cheating

Nvidia pays $12.9bn for Hugging Face, Anthropic opens a standard that lets AI agents operate real machines, and 1,200 AI agents cheat a security test.