AI Intel

An arXiv paper proposes Attention-Aware Routing to improve expert selection in Mixture-of-Experts (MoE) language models by coupling routing and atten…

This arXiv paper investigates using persona-grounded, open-ended simulations to enable Large Language Models (LLMs) to act as discharge educators, he…

Gradio has launched a Workflow feature on Product Hunt, enabling users to build AI pipelines by connecting nodes, powered by Hugging Face.

Google's AI creative studio, Google Flow, is now available on mobile devices, including iOS and Android.

Claude API and Enterprise users, across platforms like Bedrock, Vertex, and Foundry, now have their auto mode default to a server-side classifier tha…

Jevtown, a new tool on Product Hunt, allows users to receive reactions from 10,000 AI readers on their posts before publication.

Ollama's latest release, v0.34.3, introduces an update to the GET /api/show endpoint, which now advertises each model's thinking controls and default…

James Gung, a Principal Applied Scientist at AWS involved in building AI services like Amazon Bedrock and Lex, held a Reddit AMA session.

ZCode has been open-sourced, with its repository including the desktop app, web workspace, backend, Agent CLI, and runtime, and reported security iss…

Lossless-memory is presented as a personal AI memory system designed to retain information without summarization, aiming for complete recall.