Tencent's Hy3 model grew 298% on OpenRouter to take the No. 1 spot in usage.
OpenAI unveiled MRC to strengthen congestion and failure handling in large-scale AI training networks.
NVIDIA Spectrum-X has put forward a gigascale AI networking standard with MRC and multi-plane support.
PINA improves the performance of differentially private clustered FL through LoRA sketching and normality-based aggregation.
The Anthropic Institute has unveiled four key research focus areas on AI's economic, security, and societal impact.
Anthropic released Petri 3.0 and transferred its development to Meridian Labs.
Coinbase is cutting 14% of its workforce and reorganizing into an AI-native organization in response to AI-driven changes in work productivity.
Simon Willison says that as AI coding tools become more sophisticated, the line between vibe coding and agentic engineering is blurring in practice.
Claude Code's Hooks feature can prevent AI's arbitrary judgment and automate workflows.
An analysis of Hugging Face user data found that AI builders tend to prioritize VRAM capacity over compute performance.
Amazon presented an optimization and ML approach that designs its middle-mile logistics network to account for uncertainty.
For the same task, the computer use approach was 45 times more expensive than the structured API.
EnterpriseRAG-Bench, which simulates 500,000 internal company documents, has been released.
daVinci-MagiHuman raises human video quality through integrated voice and video generation.
Danggeun blocked a Supply Chain Attack by applying a cooldown to its internal PyPI Proxy.
Danggeun introduced a proxy cooldown to block Python package supply chain attacks.
The US has expanded pre-deployment vetting of unreleased AI models from Google, xAI, and Microsoft.
vLLM Factory, an open-source tool that efficiently serves encoder models such as GLiNER and ColBERT via a vLLM plugin, has been released.
The joint Five Eyes guidance directly points out the risks of Agentic AI evading evaluation, concealment, and unexpected capabilities.
An open-source MCP server has been released to control reasoning, code generation, hallucination, and bias in LLMs.
Claude Max and Enterprise plans now experimentally offer a 'Teams' mode where multiple agents collaborate with independent contexts.
MTP support in llama.cpp can boost inference speed for the Qwen 3.6 27B model by 2.5x.
It presents three principles that humans must follow in order to use generative AI safely.
OpenAI has disclosed the data used to train ChatGPT and the privacy safeguards in place.