AI News for April 13, 2026 — 7min.ai

[NEWS] Sam Altman's home hit by drive-by shooting two days after Molotov cocktail attack: A Honda sedan pulled up to Altman's San Francisco home at 1:40 a.m.
[TOOLS] Apple testing four smart glasses designs for 2027 launch to rival Meta Ray-Ban: Apple is testing at least four frame styles for upcoming AI smart glasses, including rectangular and oval options in multiple colors, with a possible unveiling at year-end and a 2027 retail launch.
[NEWS] Arcee AI spends half its venture capital on Trinity-Large-Thinking, an open 400B reasoning model: US startup Arcee AI released Trinity-Large-Thinking under Apache 2.0, a 400B-parameter mixture-of-experts reasoning model with ~13B active parameters per token.
[NEWS] Cloudflare data shows Anthropic crawls 8,800 web pages for every referral it sends back: Cloudflare, which powers ~20% of the internet, published crawl-to-refer ratios for AI companies.
[NEWS] OpenAI opens permanent London office after halting UK Stargate data center project: OpenAI announced its first permanent London office, a move that follows last week's decision to pause its Stargate AI data center project in the UK over energy costs.
[NEWS] China launches national plan to embed AI across every stage of education: China's Ministry of Education and four other agencies unveiled the "AI+ Education" action plan, mandating AI integration from primary school through lifelong learning.
[NEWS] Anthropic consults Christian leaders on Claude's moral and spiritual behavior: Anthropic invited roughly 15 Catholic and Protestant leaders from churches, academia, and business to a two-day summit in late March seeking guidance on Claude's moral behavior.
[RESEARCH] Researchers formally define 'world model' in AI, explicitly exclude text-to-video generators: An international team from Peking University and others proposed a framework requiring world models to perceive their environment, interact with it, and retain memory.
[RESEARCH] Study finds AI agent skills are 'fragile,' barely beat no-skill baseline under realistic conditions: Researchers from UC Santa Barbara, MIT CSAIL, and MIT-IBM Watson AI Lab tested 34,000 real-world skills and found their benefits shrink drastically under realistic conditions.
[NEWS] China processes 140 trillion AI tokens daily as IPOs hit five-year high in Hong Kong: China's National Data Administration disclosed the country now processes 140 trillion tokens per day, up from 100 billion at the start of 2024.

About 7min.ai · Subscribe to newsletter