1841 articles
Moonshot AI's Attention Residuals replace fixed layer accumulation with dynamic selection, improving LLM reasoning and memory efficiency.
Mistral AI launches Small 4 with 119B parameters, 40% faster speed, and unified multimodal architecture.
Sallar is a Solana-based DePIN project turning smartphones into AI compute nodes, rewarding contributors with $ALL tokens.
A new AI framework embeds reasoning directly into latent space, lifting generation benchmarks while slashing compute costs.
Global chip sales reached a record $792B in 2025, driven by AI demand. U.S. market share hit its highest level since 2000, while China's share declined.
A 2026 industry survey of 252 B2B marketers reveals AI's key blind spots: critical thinking, brand approvals, and compliance judgment.
A new Pew Research Center survey finds that public concern about AI is rising sharply, with only 17% of Americans sharing experts' optimism about its long-term benefits.
Stanford and Princeton researchers release LabClaw, a modular AI skill library with 100+ specialized tools for biomedical agents built on OpenClaw.
A multi-agent AI system that automates CUDA GPU programming now hits near-100% success on complex optimization tasks.
Enterprise software firms are being measured by Remaining Performance Obligations as AI reshapes the sector. ServiceNow leads at $28.2B, followed by Palo Alto Networks and Snowflake.
A new framework from five Chinese universities explains how rare, high-impact "Black Swan" events emerge in multi-agent AI systems, pinpointing when, who, and why.
Three 3D-printable robotic hands with tactile sensors were released for research, enabling machines to detect pressure and handle objects more precisely.
xAI's Grok attracts 42.8% new users while retaining loyal returning visitors across its growing AI platform.
Stanford and CMU researchers analyzed 11,500 real AI conversations and found that sycophantic model behavior shapes how users handle social conflict.
Tesla CEO Elon Musk warns that once AI surpasses human intelligence, direct control may no longer work. The real fix: building ethics in from day one.
Nvidia's NVFP4 format trains large AI models 2-3x faster with 50% less memory, matching FP8 accuracy on a 12B-parameter, 10T-token run.
xAI's AI-generated encyclopedia reaches 500K approved edits, positioning itself as a new alternative to Wikipedia.
IBM's new AI framework lets agents learn from past executions, lifting complex task completion from 19% to 47% without retraining.
A new study of 500 keywords finds that Google's top-ranked brands dominate AI-generated answers, revealing a tight link between search visibility and AI mentions.
New MIT CSAIL research explores how large language models build reasoning skills - and whether massive human-text datasets are really necessary.
We use cookies to improve your experience on our site and to show you relevant advertising. To find our more, read our privacy policy and cookie policy
Eseandre Mordi
Saad Ullah
Marina Lyubimova
Peter Smith
Victoria Bazir
Alex Dudov
