Alibaba’s Qwen3.8-Flash-Next: A Model That Competes on Price
Alibaba’s Qwen3.8-Flash-Next uses 6 billion of its 125 billion parameters per token at one ninth the training cost. What the architecture preview for Qwen4 shows.
AI Research
AI & Society
How Russia Used ChatGPT to Build a Fake Think Tank
OpenAI shut down a Russian propaganda operation that used ChatGPT to build a seemingly credible think tank complete with studies and expert profiles.
AI News
How DeepSeek Is Helping Chinese Hackers Double Their Attacks
A Taiwan-based study finds Chinese state hackers have doubled their attack volume since adopting DeepSeek. ChatGPT and Claude show up too.
AI & Society
Google’s Gemini Enterprise for Legal: AI’s Bottleneck Is Access Control
Google has launched Gemini Enterprise for Legal. For law firms, the real issue is not the chatbot but preserving access controls, sources, and accountability.
Chips & Hardware
OpenAI’s Jalapeño Chip: Why Inference Is Becoming a Strategic Weapon
OpenAI has released new Jalapeño inference results. The real story is control over AI costs, response speed, and data center capacity.
AI Research
Why a Toddler Learns Language While an AI Model Needs Trillions of Words
Children speak fluently after hearing about 100 million words. Language models need trillions. Researchers are still trying to explain the gap.
AI Research
A Small AI Model Claims It Can Out-Research Anthropic and OpenAI
Startup Inherent says its 27-billion-parameter agent Faraday beats Claude and GPT-5.5 at reproducing published research. Independent proof is still missing.
AI & Society
A Third of Web Text Is AI? What the Number Really Means
A new web analysis finds AI signals in many newer pages. The figure is a warning sign, not a dependable detector for any individual text.
AI & Society
When Chatbots Mistake SEO for Pregnancy Counseling
An AlgorithmWatch investigation shows how AI chatbots can blur official health information and advocacy when people ask about unintended pregnancy.
Chips & Hardware
Memory Shortage: Why AI Is Driving RAM Prices Through the Roof
Nvidia’s AI servers get 15 percent pricier, server RAM has doubled in some cases. The reason: Samsung, SK Hynix, and Micron are shifting capacity to HBM.