AI Research news and analysis on Web Pulse: 9863 curated articles tracking the latest AI Research developments, tools, and research, updated continuously from vetted sources.
A study in Cell reveals that bilingual brains use a shared neural map to represent word meanings across languages, based on recordings from individual brain cells in the hippocampus. The findings suggest the brain stores…
RadixArk released Miles, an open-source PyTorch-native framework for large-scale LLM reinforcement learning post-training. The framework integrates SGLang for rollout, NVIDIA Megatron-LM for training, and Ray orchestrati…
OpenAI engineers cut inference costs for guest ChatGPT users by more than half, reducing the number of Nvidia GPUs needed to just a few hundred, according to a person familiar with the discussions. The optimization appli…
OpenAI released GeneBench, a benchmark of 103 multi-stage genomics problems, revealing that top AI models like GPT-5.5 Pro achieve only a 33.2% pass rate, with 60% of problems below 20% for current models. The benchmark …
A reporter visited OpenAI, Anthropic, and Cursor in San Francisco, observing that cloud-based AI agents are becoming a major trend, with companies like Anthropic focusing on hosted services like Claude Managed Agents. No…
Anthropic launched Claude for Life Sciences, a suite of AI tools automating scientific research from hypothesis to regulatory submission. The platform scored 0.83 on Protocol QA, beating the human baseline of 0.79, and i…
Anthropic launched Claude Science, an AI workbench for scientists, integrating tools like PubMed, Jupyter, and R into a single environment. The platform generates reproducible scientific artifacts, manages computing reso…
A developer building Halo, an autonomous pentesting agent powered by a local LLM, discovered a critical bug in its failure cache that caused the agent to permanently blacklist tools across unrelated engagements. The cach…
NVIDIA launched the BioNeMo Agent Toolkit, which integrates with Anthropic's Claude Science AI workbench to give life sciences researchers access to GPU-accelerated models and workflows via natural language agents. The t…
Anthropic launched Claude Science, a workbench for scientific research, at its AI for Science event on Tuesday. The offering aims to expand revenue streams ahead of the company's planned IPO later this year, as it compet…
Anthropic launched Claude Science, an AI workbench for scientists that integrates over 60 databases and pre-built toolkits, aiming to streamline computational research without introducing a new model. The platform includ…
Etched emerged from stealth with a working chip, $800 million raised, and over $1 billion in signed customer contracts, pivoting from a transformer-specialized ASIC to a full-stack inference systems company. The startup,…
Google Cloud introduced a new skill for coding agents that automates the evaluation and improvement of agent quality, based on a three-phase flywheel methodology developed with Google DeepMind. The skill, which integrate…
A security researcher at SpecterOps has demonstrated that large language models can reverse-engineer endpoint detection and response (EDR) products, including Palo Alto Networks Cortex XDR, to extract detection rules and…
A developer reports that memory adherence is a systems problem, not a prompting problem, and that the key difference between model families is the control surface they offer. Claude provides a full ladder of control from…
Outpost VFX achieved 8x faster AI model training for visual effects face replacement by using AWS multi-GPU EC2 P5 instances, overcoming single-GPU bottlenecks that previously caused week-long delays. The collaboration w…
Rocket Lab CEO Peter Beck announced the $8 billion acquisition of Iridium Communications to challenge SpaceX's dominance in the satellite communications market, particularly in direct-to-mobile services. The deal positio…
Nvidia dominates AI training hardware, but the inference market is wide open as companies like Cerebras, Groq, d-Matrix, Etched, and Taalas challenge Nvidia's GPU-based approach by designing chips optimized for low-laten…
Purdue researchers tested 24 deepfake detection systems and found commercial tools generally performed better as AI identity fraud raises new risks for crypto KYC. The study found Incode's DeepSight achieved the lowest f…
Meta's non-surgical mind reading machine achieves 61% word accuracy, an improvement over prior projects, but still lags behind implanted brain-computer interfaces and cannot support real-time communication, as it relies …