Category: Artificial Intelligence

• Google's Big Release: WikiSkill — Skills Can Now Evolve on Their Own!
• First Tests of OpenAI's New Flagship 'Astra' Leak — Is Front-End Development Over?
• Andrew Ng Co-Authors New Stanford Paper: Dropping Intermediate Reasoning Tokens Speeds Up Long Thinking by Up to 3x
• Terence Tao's Counterintuitive Warning: AI Is Making Humans Too Efficient—And It's Killing the Next Technological Revolution!
• Hy4 Preview Released
• OpenAI Has Lost Its "God of CUDA Kernels"
• "Some Models Just Don't Want to Learn?" Why Do Recurrent Models Get It More Wrong the More They Think?
• Alibaba Open-Sources 0.8B Document Parsing Model: End-to-End Model Surpasses Pipeline Methods for the First Time
• “At least a few old COBOL programmers understood the system. After migrating to Java with AI, no one can maintain it.”
• Kimi K3 vs. GPT-5.6 Sol: A Comprehensive Comparison Review!
• Google Research Releases Cappy: Enhancing Multi-Task Large Language Models with a 360-Million Parameter Model
• When Async Agentic RL Meets 'Amnesia for Old Policies': Rethinking Off-Policy Correction
• Z Tech | From Biological Evolution to AI Evolution, Toward AGI and Immortality: Building a Life Science Engine for the Post-AGI Era
• ICML 2026 | Teaching Multimodal Large Models to Think with Time: Peking University and Huawei Team Open-Source the TaRO Framework
• Yann LeCun's Team's Latest Research: Teaching World Models to "Adapt" and Continuously Evolve Through Action
• The Flawless 'Flaw': Qwen and Fudan University Uncover Structural Dilemmas in Coding Agent Reward Design
• Ask, Don’t Judge: Binary Questions for Interpretable LLM Evaluation and Self-Improvement
• Unlimited OCR: One-Shot Parsing of Long Documents with Reference Sliding Window Attention (R-SWA)
• HIT KaLM-Reranker-V1: An Efficient RAG Retrieval Booster, Deep Dive into High-Speed Reranking
• Beyond Prediction: The Five Realms of 'Memory' in Dynamics Learning
• Cursor's First Conference Lays It All Out: 95% of Usage Comes from Agents, Founder Unveils Three Major Updates
• ByteDance's Seedance Shakes Things Up Again: Native 30-Second Output, 4K+10-bit, 50-Source Combined Input, Stephen Chow Films Now Authorized
• The End of Code Review: Coding Agents Supersede Human Inspection
• Anthropic's First Mythic-Class Claude 5 Officially Unleashed
• Latest Discovery: AI Large Models Know When They're Being Evaluated