Google’s JAX-Privacy Aims to Scale Differentially Private Machine Learning
Google Research’s new JAX-Privacy library is designed to make differential privacy practical for large-scale machine learning.
AI & ML
Frontier models, research results, agents, and developer tooling.
MiniMax M2 Asks What Agent Alignment Should Target
A new post from MiniMax AI argues that aligning agents to static preferences may not be enough to generalize across tasks.
DeepMind Launches AI for Math Initiative with Top Research Institutions
Google DeepMind's new initiative partners with leading institutions to apply AI to mathematical research.
Hugging Face Promotes AMD Open Robotics Hackathon
Hugging Face is calling developers to join an AMD-branded robotics hackathon, highlighting open robotics as the focus.
Global AI Compute Shifts Toward New Regions
Hugging Face analysis outlines how energy, hardware, and policy are redrawing the map of where AI models get trained and run.
AI Moves From Counting Forest Loss to Predicting Forest Risk
Google Research describes how machine learning is shifting forest monitoring from after-the-fact loss accounting toward forward-looking risk prediction.
Hugging Face Hub Library Reaches v1.0 After Five Years
The release marks a stable foundation for the Python client that powers open machine learning workflows.
DeepMind's WeatherNext 2 Delivers More Accurate Forecasts
Google DeepMind's WeatherNext 2 model improves global weather prediction efficiency, accuracy, and resolution.
Google DeepMind Unveils Gemma 3n Developer Guide
Gemma 3n targets the developer community, building on input that shaped earlier Gemma models.
Google Research targets provably private analytics for AI usage
New work from Google Research aims to let developers measure how generative AI is used while offering mathematical privacy guarantees.
Google Research Ties AI Breakthroughs to Real-World Climate Gains
Google Research argues that feeding lessons from deployed climate tools back into AI development can speed both innovation and impact.
Healthcare Robots Move From Simulation to Deployment via NVIDIA Isaac
Two Hugging Face guides describe a simulation-to-deployment workflow for healthcare robots using NVIDIA Isaac, with consistent emphasis on the same toolchain.
Google Research Explores Space-Based AI Infrastructure Design
Google Research has published an early design exploration for a space-based, scalable AI infrastructure system, pointing to orbital computing as a possible direction for AI workloads.
Google DeepMind Announces Google Antigravity
The research lab's blog post introduces a new initiative called Google Antigravity, but the source provides no further details.
Google DeepMind releases MedGemma, its most capable open health AI models
The release gives developers open multimodal models designed for health AI development.
Gemini 2.5 Flash-Lite reaches general availability
Google DeepMind's low-cost, compact model is now stable and ready for large-scale production workloads.
IBM's Granite 4.0 Nano: Pushing the Limits of Small Language Models
A Hugging Face blog post examines how far model compression can go without sacrificing practical utility.
Google Research Details Verifiable Quantum Advantage
Google's latest quantum computing claim comes with a verification protocol, making the result easier to check than earlier demonstrations.