Page 14 of 33 · 391 publications
Microsoft has officially detailed its new Nvidia RTX Spark-powered Surface hardware and a revamped Windows 11 designed to execute complex, multi-model AI agent workflows locally. Featuring advanced thermal design and native sandboxing tools, these machines target developers moving away from traditional cloud infrastructures.
OpenAI is upgrading ChatGPT with an Intelligent UI powered by GPT-6, allowing the chatbot to blend text with dynamic charts, functional forms, and interactive buttons. This major shift moves conversational AI beyond static text into rich, task-specific web applications.
OpenAI has published a massive batch of 722 manuscripts detailing solutions to hundreds of long-standing mathematical problems generated by an unreleased frontier model. The release comes amid growing debate over academic ethics, research transparency, and how AI labs communicate breakthrough discoveries.
After abandoning their initial marketing optimization tool, former Ramp engineers have successfully raised a $20 million Series A to build Melius, an agentic workspace dedicated to automated creative asset generation.
A series of coordinated aerial and sea drone strikes have sunk two commercial cargo ships and severely damaged a third within the exclusive economic zones of NATO member countries in the Black Sea, resulting in multiple casualties. This unprecedented escalation poses severe threats to maritime security and international law.
Musubi has introduced PolicyLM-1.7B, a specialized open-weights decision model that evaluates content against plain-English policies in under 50 milliseconds. This approach bypasses costly retraining cycles, bridging the gap between rigid classifiers and flexible LLMs.
Google is reportedly expanding its AI-powered 'Call for Me' feature beyond business interactions to handle communications with friends and family. APK teardowns reveal upcoming capabilities for Gemini to place quick status calls and send brief voice updates to personal contacts.
As organizations rush to deploy interconnected AI systems, new research reveals that the Model Context Protocol (MCP) creates critical trust gaps. These vulnerabilities allow malicious prompts to pivot invisibly between specialized agents across internal networks.
OpenAI is rolling out an invisible cryptographic watermarking system for ChatGPT and Codex users within the European Union. Designed to comply with strict EU AI Act transparency mandates, the technique alters token selection without degrading model performance.
Brooklyn-based startup Reflection AI has introduced Beam, a massive 501-billion-parameter open-weight model trained via intensive reinforcement learning. Designed to compete with top-tier Chinese models and Western alternatives, Beam promises high-level reasoning and agentic performance at drastically reduced inference costs.
A modern forensic re-examination of ancient Egyptian skulls has uncovered telltale linear fractures indicative of blunt force trauma. This fresh evidence strongly supports the historical hypothesis of early dynastic retainer sacrifice.
Google has halted its open-source bug bounty program until early 2027 due to an unsustainable influx of invalid, hallucinated reports driven by automated AI tools. This unprecedented move highlights the growing crisis of generative AI spam undermining software security pipelines.