DeepMind reimagines the mouse pointer as an AI collaborator
Google DeepMind's 'AI pointer' concept aims to shift the cursor from a passive input tool to a context-aware assistant that reduces reliance on explicit prompts.
AI & ML
Frontier models, research results, agents, and developer tooling.
Holotron-12B: A High-Throughput Agent for Computer Use
Hugging Face introduces Holotron-12B, a model built to automate computer interactions at scale.
Google's XR Blocks and Gemini Speed Up AI-Enhanced XR Prototyping
A new visual coding approach pairs block-based XR development with Gemini to cut down prototyping time for AI-powered mixed reality experiences.
Google Tests LLMs on Superconductivity Questions
A Google Research blog post explores how large language models handle specialized questions in superconductivity research.
Safetensors Format Joins PyTorch Foundation
Hugging Face's Safetensors serialization format is moving under the PyTorch Foundation, giving the widely used tensor format a neutral governance home.
DeepMind Unveils Gemma 4 Open Models for Reasoning and Agents
Google DeepMind's new Gemma 4 family targets advanced reasoning and agentic workflows, positioning itself as the most capable open model lineup to date.
Google Research adds AI agents for figures and peer review
The two new tools target specific pain points in academic publishing: visual quality and manuscript review.
Waypoint-1.5 Brings Higher-Fidelity Interactive Worlds to Everyday GPUs
Hugging Face's Waypoint-1.5 aims to make high-fidelity interactive world generation accessible on consumer-grade hardware.
Sentence Transformers Adds Multimodal Embedding and Reranking
Hugging Face's library now supports models that embed and rerank both text and images, making cross-modal retrieval more practical.
Google's ConvApparel measures how realistic AI user simulators are
A new framework from Google Research quantifies the gap between simulated and real user behavior, then uses that signal to improve the simulators.
AI Flash Flood Forecasting Helps Cities Prepare Faster
Google Research shows how machine learning models can improve the speed and accuracy of flash flood warnings for urban areas.
Hugging Face Hub Adds Storage Buckets for Large Files
The new feature gives users a dedicated space for large data and models on the Hugging Face Hub, simplifying management.
Gemini 3.1 Flash Live Targets Smoother, Lower-Latency Voice AI
DeepMind's latest voice model aims to make spoken AI interactions more fluid and precise by improving precision and reducing latency.
Machine Learning Aims to Improve Breast Cancer Screening Workflows
Google Research highlights how ML can support radiologists by streamlining breast cancer screening workflows rather than replacing human judgment.
TurboQuant targets extreme compression for efficient AI
Google Research's TurboQuant is an algorithmic push to make AI models dramatically more efficient via extreme compression.
DeepMind's Lyria 3 Pro Generates Longer, Structurally Aware Music
Google DeepMind's latest music model extends track length with structural awareness and is being integrated across more Google products.
Google Research Outlines Health AI Move to Real-World Care
Google Research's post from The Check Up focuses on translating healthcare innovations from research into practical clinical settings.
Hugging Face Hosts Gemma 4, a Multimodal Model for On-Device AI
The release brings frontier-level multimodal intelligence to edge devices, according to a Hugging Face blog post.