1841 articles
LangChain just launched PeopleHub, an AI platform that automates LinkedIn research through simple natural language queries. The system slashes operational costs by up to 90% while delivering structured reports in seconds
ZTE's EmbodiedBrain model introduces a unified vision-language-action system designed to improve planning and execution in physical environments, demonstrating state-of-the-art performance across complex embodied AI tasks.
ChatGPT has exploded to over 800 million weekly users in just three years, now serving about 10% of all adults globally. New usage data reveals writing leads the pack while personal, non-work conversations are taking over.
A Chinese tech company unveiled GEAIR, a humanoid robot with a synthetic womb designed to carry and deliver babies, sparking global controversy over the future of human reproduction.
Epoch AI just dropped new performance scores for the biggest models out there — Gemini 3 Pro, GPT-5.2, and Claude Opus 4.5 — and the numbers reveal some pretty interesting gaps in how long these systems can actually stay on task.
Despite explosive revenue growth forecasts, OpenAI faces persistent negative cash flow as compute costs match every dollar earned
Federal Reserve Chair Jay Powell acknowledges artificial intelligence is starting to impact American jobs, marking the first time the Fed has officially connected AI adoption to rising unemployment numbers.
A breakthrough AI agent called InternGeometry has achieved near-gold-medalist level performance on International Math Olympiad geometry challenges, using surprisingly limited training data to solve complex mathematical problems.
OpenAI's GPT-5.2 Thinking shows major accuracy improvements across 256k token contexts in new MRCRv2 benchmarks, dramatically reducing "context rot" and opening doors for more reliable enterprise AI agents.
Shanghai AI Lab's Intern-S1-MO agent just scored gold-medal level on brutal Olympiad math problems, leaving GPT-4, Grok, and other top reasoning models in the dust across multiple competition benchmarks.
Singapore researchers created an octopus-inspired swimming robot that coordinates eight soft arms using only two motors, achieving speeds up to 314 mm/s while dramatically improving energy efficiency compared to traditional multi-actuator designs.
Donald Trump signed an executive order blocking U.S. states from enforcing their own AI laws, consolidating oversight under one federal framework. The decision overrides transparency and bias-testing requirements already adopted by states like California and Colorado.
Disney has partnered with OpenAI in a three-year deal worth $1 billion, bringing over 200 beloved characters from Disney, Pixar, Marvel, and Star Wars to the Sora video platform. The agreement combines equity investment with widespread AI integration across Disney's operations.
Enterprise AI spending tripled to $37B in 2025, but open-source models are losing ground fast. Meta's Llama, despite leading the open-weight category, couldn't stop its market share from dropping to just 11% as enterprises bet bigger on commercial platforms.
Fresh SimpleBench results reveal GPT-5.2 trailing far behind competitors in common-sense reasoning tests, with Gemini 3 Pro Preview taking the crown.
New research reveals DeepCode, an autonomous coding agent that achieves 84.8% accuracy in converting scientific papers to working code—outperforming commercial tools like Cursor (58.4%) and even expert ML researchers (72.4%).
Fresh Design Arena benchmarks crown Gemini 3 Pro Preview as the top-ranked AI model, with Claude Opus 4.5 trailing closely and OpenAI's GPT-5.2 (xHigh) making a strong debut across creative categories.
The AI industry is burning through cash at an unprecedented rate, spending nearly seven times what it earns. The massive gap is being plugged with debt, creative lease arrangements, and a web of circular financing deals that tie major players together in ways that could amplify future risks.
Epoch AI has revamped its benchmark methodology to score AI models based on their best performance across all test configurations, making it easier to compare systems from Google, OpenAI, Meta, Anthropic, and xAI.
OpenAI's latest documentation reveals GPT-5.2 dramatically cuts factual errors compared to earlier versions, marking a major leap in AI reliability.
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
Alex Dudov
Usman Salis
Peter Smith
Saad Ullah
Eseandre Mordi
Victoria Bazir
Sergey Diakov
Marina Lyubimova
