Breaking news, AGI updates, and analysis of the intelligence revolution.
📰 News LG's new robot data factory in Seoul aims to generate 100,000 hours of training data by year-end, using Nvidia's Cosmos platform for synthetic data generation. Madison Huang, Nvidia's senior director and Jensen Huang's daughter, toured the facility this week.
📰 News The $250 billion guarantee from July is now $105 billion in credit plus a $1.5 billion energy investment. Nvidia says $600 billion in compute through 2030. The chipmaker-as-banker era is here.
📰 News Serve's sidewalk delivery fleet just got a Grubhub integration, two new markets, and a hospital cousin with 15x faster perception. The company is quietly building a multi-robot ecosystem.
📰 News Unitree's Superman humanoid runs faster than Bolt's peak sprint speed and clears a 2-metre standing jump. The timing is no accident — the company goes public Wednesday.
📰 News One broker advertised $100,000 a day in API spend. Buyers get cheap inference; their data travels through an intermediary whose security practices, contractual terms, and storage arrangements are entirely unknown.
📰 News A $3.48m age-assurance report cited fake academic papers. The authors admitted using ChatGPT. The ban went into effect anyway.
📰 News Dyna Robotics trained Dyna-2 on one million hours of first-person human video and proved a scaling law for robot manipulation that transfers zero-shot to hardware the model never saw.
📰 News Wynd Kaufmyn, 69, is believed to be the first person jailed for AI protest. Supporters call her the Rosa Parks of AI risk.
📰 News Infiforce's $150M raise funds an AtomBrain world model and DataGrid infrastructure that already operates across 30+ Chinese cities, betting that first-person video can train robots the way text trained LLMs.
📰 News The team evaluating whether OpenAI's models could pose catastrophic risks is gone. Its work has been absorbed into existing teams, and the former lead has pivoted to recursive self-improvement safety — the very thing the team was meant to guard against.
📰 News From India's Cockroach Janta protests to Mexico's mass retake to Portugal's marking crisis, the 2026 exam season has produced unrest on three continents. AI cheating, digitisation failures, and economic pressure on students are the common thread.
📰 News McAfee tested 21,000 travel images with two free AI models. One pinpointed locations 91 per cent of the time. Criminals are using that data to send messages that mention your exact holiday spot — making phishing texts nearly impossible to spot.