AI Research news and analysis on Web Pulse: 8402 curated articles tracking the latest AI Research developments, tools, and research, updated continuously from vetted sources.
Google DeepMind's mechanistic interpretability team proposed a pragmatic framework for validating interpretability tools using proxy tasks, demonstrating its effectiveness by subtracting an "eval-awareness" vector from C…
Runway opened its first office in Paris with a 10-person team and a $30 million investment, positioning the hub as a research center for world models and physical AI. The move reflects the company's strategy to recruit t…
Chinese AI startup DeepSeek has begun developing its own AI chips focused on inference workloads to reduce reliance on NVIDIA and cut costs, sources told Reuters. The project is in early stages and has been underway for …
AI systems are shifting from single-prompt engineering to self-correcting loop architectures, where outputs are iteratively evaluated and refined. Research from Carnegie Mellon shows iterative self-refinement improves LL…
Sejong University's Future GPU Research Institute, led by professor Park Woo-chan, is accelerating development of next-generation GPU and AI-semiconductor technologies, focusing on real-time path tracing for applications…
Two developers built AI agent loops last month; one ran for six hours with little output, the other never ran unattended. The common mistake was failing to define what the loop could do autonomously, a decision that dete…
Nebula Security published exploit code for GhostLock, a 15-year-old Linux kernel vulnerability (CVE-2026-43499) that allows local privilege escalation and container escape. The flaw, introduced in Linux 2.6.39 and fixed …
OpenAI's technical post on Codex CLI and Philip Zeyliger's work at Sketch.dev demonstrate that AI agents using iterative loops outperform single-shot prompts for complex tasks. The agent loop, where the model repeatedly …
SpaceXAI will reportedly release a new AI model as early as Wednesday, built jointly with AI startup Cursor, which SpaceX is acquiring for $60 billion. The model, internally compared to Anthropic's Claude Opus 4.8 and Op…
Lilian Weng, cofounder of Thinky, published a summary of 35 papers on harness engineering for recursive self-improvement (RSI), arguing that harness improvements will remain essential even as core models evolve. The post…
Meta Superintelligence Labs launched Muse Image on July 7, the first image generator that autonomously uses tools like web search and code execution at inference time, with self-refinement emerging from reinforcement lea…
AI has made execution nearly free, shifting scarcity from engineering labor to decision-making, and the next frontier is wanting—the goals that drive decisions. As AI improves at deciding, humanity's last scarce resource…
A developer evaluates CrewAI, a multi-agent orchestration framework, highlighting its strengths in decomposing complex tasks into focused subtasks with specialized agents, while noting limitations such as compounding tok…
South Korea's proposed RE100 industrial complex remains stalled without a passed special bill or selected site more than a year after its introduction, while the 800 trillion won southwestern semiconductor hub advances r…
A developer named ArshiaLogic has created a roadmap for aspiring AI engineers, focusing on hands-on projects rather than tutorials. The roadmap covers Python, data science, machine learning, deep learning with PyTorch, t…
NVIDIA open-sourced Nemotron-Labs-TwoTower, a diffusion language model that generates text 2.42x faster than its autoregressive counterpart without retraining original weights. The model achieves 98.7% quality retention …
Microsoft has begun replacing OpenAI and Anthropic models with its own MAI models in Excel and Outlook as part of a cost-cutting strategy, according to a Bloomberg report. The shift, which handles tens of thousands of pr…
JetBrains released Mellum2, a 12-billion parameter open-source Mixture-of-Experts model that activates only 2.5 billion parameters per inference, making it twice as fast as equivalent models. Designed for high-frequency …
Nvidia has spent over $27 billion on investments, acquisitions, and partnerships with former rivals including Marvell, Intel, and Groq under CEO Jensen Huang's 'AI factory' strategy. The moves aim to consolidate Nvidia's…
A new research paper introduces the concept of probabilistic 'copies' in generative AI models, exploring how these models can inadvertently reproduce training data with varying degrees of fidelity. The study has implicat…