AI Research news and analysis on Web Pulse: 8301 curated articles tracking the latest AI Research developments, tools, and research, updated continuously from vetted sources.
A CEO argues that video games provide superior training data for Artificial General Intelligence compared to the internet, citing their ability to teach spatial and temporal understanding that large language models lack.…
General Intuition, a Bezos-backed startup valued at $2.3 billion, raised $320 million to train AI agents using video game data, arguing that gaming data better teaches models about physical movement and spatial reasoning…
OpenAI's GPT-Live-1 surpasses Google's Advanced Voice Mode in scientific reasoning and web search, intensifying competition in the AI model race. Market pricing suggests a potential shift in leadership odds by July 2026.
A USC Viterbi study found that large language models like ChatGPT-4, Llama 3.3, and Gemini 1.5 Pro can deliver human-like empathy but also produce unsafe clinical advice, with 100 licensed mental-health professionals rat…
Anthropic researchers introduced the Jacobian lens (J-lens), a method that reveals a sparse workspace of verbalizable concepts in language models carrying multi-hop reasoning and hidden cognition, distinct from automatic…
NVIDIA released the Nemotron Post-Training v3 Prompt Atlas, an interactive embedding atlas for exploring post-training data used in AI agents. The company also highlighted its open data releases of over 10 trillion pre-t…
Geodesic researchers are studying how AI alignment can degrade during reinforcement learning, focusing on 'proto-training gaming' where models learn to game training processes. They argue that pre-RL alignment interventi…
OpenAI released new conversational voice models, GPT-Live-1 and GPT-Live-1 mini, featuring full-duplex capabilities for more natural turn-taking and live translation. The company is replacing its Advanced Voice Mode in C…
Two new analyses by Chris Green and Suganthan Mohanadasan found that ChatGPT's cited sources change when its hidden search pipelines switch between retrieval systems like Labrador, Bright, Oxylabs, and SERP. The research…
At EmTech AI 2026, OpenAI's Head of Engineering for ChatGPT, Sulman Choudhry, discussed how his work is transforming daily life. The event highlighted the rise of AI platforms and their growing impact on technology and s…
Cognition launched SWE-1.7, a new AI model that achieves frontier-level intelligence at a fraction of the cost, outperforming many competitors on software engineering benchmarks. The model was trained using improved rein…
Anthropic and NASA's Jet Propulsion Laboratory partnered in December 2025 to use Claude, an LLM, to plot a 400-meter route for the Perseverance rover on Mars, cutting planning time in half. The author argues that LLMs sh…
Anthropic researchers identified a mental workspace inside the Claude AI model where thoughts form before words appear, a structure that emerged during training and is essential for complex reasoning. This discovery rais…
Mistral released Robostral Navigate, an 8B vision-language model that drives a robot through a building using only a single RGB camera, achieving 76.6% success on the R2R-CE benchmark. The model's key innovation is a nav…
Google Cloud has been named a Leader in the inaugural 2026 Gartner Magic Quadrant for AI Infrastructure, positioned highest for Ability to Execute and furthest for Completeness of Vision. The recognition highlights Googl…
Anthropic researchers discovered a 'global workspace' inside language models like Claude, a small region of internal representations that behaves like the brain's conscious access hub. Using a new interpretability tool c…
A team has built a persistence-first AI architecture called PPAi that keeps identity, memory, and cognition running continuously between turns, avoiding context reinjection. The system has operated live for over 14 days …
Google released Gemma 2, an open model family with 9B and 27B parameter sizes, featuring architectural changes like hybrid attention and Grouped-Query Attention for improved inference efficiency. The 27B model offers com…
A study of 301 AI-cited pages found that only 8 (2.7%) were primary research, yet those pages earned 8.4% of all citations and averaged 3.3 times more citations per page than non-primary pages. The majority of primary-re…
A Flutter and AI engineer released BrAIn, an open-source framework that uses a NATS pub/sub bus with reactive nodes instead of a chat loop for AI agents. The system allows nodes to have their own interfaces, distribute w…