AI Research news and analysis on Web Pulse: 9903 curated articles tracking the latest AI Research developments, tools, and research, updated continuously from vetted sources.
NVIDIA released Nemotron 3 Ultra, a 550-billion-parameter open model under a fully permissive license, competitive with GPT-4.5 on code and reasoning. Z.AI launched GLM-5.2 with MIT-licensed weights, excelling in long-co…
London-based robotics startup morph embeds sensing and adaptive control directly into reconfigurable deformable materials, creating soft robotic cells that can change shape and stiffness in real time. The company aims to…
Meta announced Brain2Qwerty v2, an AI system that non-invasively decodes brain activity into text, aiming to assist people with neurological injuries or speech impairments. The improved brain-computer interface builds on…
The Bank of England is warning that autonomous AI trading agents could trigger a market crisis by herding into identical decisions at machine speed, turning manageable stress into systemic collapse. Deputy Governor Sarah…
A Faros AI study of 22,000 developers across 4,000 teams found that AI coding tools boosted PR counts by 98% but tripled production incidents per PR, with code churn jumping 10x and bugs per developer up 54%. The phenome…
Researchers have demonstrated that compiling agentic workflows into the weights of small fine-tuned language models achieves near-frontier quality at two orders of magnitude less cost, addressing three perceived barriers…
A Boston University study found that managers caught 18% fewer errors when work was attributed to an AI 'employee' rather than a chatbot, raising concerns about treating AI agents as coworkers. Meanwhile, Sceye plans to …
Five strategies for managing context windows in long-running AI agents—sliding windows, recursive summarization, structured state management, ephemeral context via RAG, and dynamic context routing—are presented with thei…
Attorney Bill Savitt has defeated Elon Musk in two high-profile legal battles, first representing Twitter in Musk's attempt to back out of the acquisition and then representing Sam Altman and OpenAI in Musk's lawsuit. Sa…
Beijing-based Z.ai released GLM-5.2, an open-weight model trained on Huawei Ascend chips without Nvidia hardware, which quickly topped public leaderboards including Design Arena and Artificial Analysis's Intelligence Ind…
Aran Nayebi's new paper argues that AI consciousness may be an inevitable byproduct of building highly capable agents, using selection theorems to show that robust decision-making under uncertainty requires properties as…
Bshada released Open Schematics V2, the largest open electronic schematics dataset, containing thousands of KiCad schematic and PCB files with metadata. The dataset aims to accelerate machine learning for electronics des…
Midjourney is pivoting from AI image generation to full-body ultrasound scans, sparking debate as the medical community advocates against widespread testing. The author argues that cumulative data from pervasive testing …
The U.S. Army is opening military bases to private investment for hyperscale data centers and other infrastructure, aiming to align private capital with national security needs. These projects, involving billions in priv…
Enterprises adopting multi-agent workflows are failing because they apply the human-focused Software Development Lifecycle (SDLC) to non-human AI agents, which have different failure modes. The proposed Agentic Developme…
MongoDB has introduced a native reranking capability for its Atlas database platform, currently in public preview and powered by Voyage AI, to help enterprises improve AI retrieval quality without adding separate service…
Researchers at the University of Washington released TheoremGraph, a unified dependency graph spanning 18 million+ mathematical statements from arXiv papers and the Lean formal proof assistant. The graph enables mathemat…
DeepSeek V4 Flash, a 284-billion-parameter Mixture-of-Experts model, now runs locally on a 128GB MacBook Pro via antirez's experimental llama.cpp fork, achieving ~21 tokens/sec generation on the Metal GPU. The 2-bit quan…
The Vector Institute in Toronto released UnBias-Plus, an open-source tool that detects and rewrites biased language in text and AI training data, targeting biases related to race, gender, age, and political framing. The …
Penguin Solutions upgraded its ClusterWareAI platform on June 25, adding an AI-powered operations agent and automated GPU health monitoring to simplify management of large GPU clusters. The company, which has deployed ne…