Sierra Leone RCT Links Gemini Guided Learning to Faster Progress
Google DeepMind reports that a randomized controlled trial of Gemini's Guided Learning feature in Sierra Leone improved student engagement and accelerated learning.
AI & ML
Frontier models, research results, agents, and developer tooling.
Google open-sources hydrology framework to boost flood forecasting
The move gives researchers worldwide access to Google's flood-prediction tools, aiming to improve resilience in vulnerable regions.
Hugging Face Pitches Jobs as a Home for GitHub CI Workflows
A new Hugging Face guide walks teams through moving continuous integration from GitHub Actions to Hugging Face Jobs.
Hugging Face Highlights OpenEnv for Agentic Reinforcement Learning
A Hugging Face post points to OpenEnv as a community-backed focus for agentic RL work.
Google Details Agentic RAG for Enterprise AI Dependability
Google Research describes how its Gemini Enterprise Agent Platform uses Agentic RAG to make AI responses more dependable.
BBVA brings ChatGPT Enterprise to 100,000 staff
The Spanish bank's global rollout signals generative AI shifting from pilot projects to core banking infrastructure.
OpenAI’s AI Chemist Improves Key Drug-Making Reaction
A near-autonomous AI system using GPT-5.4 improved a difficult medicinal chemistry reaction, signaling a shift toward AI-driven lab research.
OpenAI Launches LifeSciBench to Test AI in Life Science Research
The expert-reviewed benchmark aims to measure how well AI systems handle real-world life science research tasks and decisions.
UK Taps DeepMind for AI Housing Planning Prototype
A new government partnership aims to speed up Britain's sluggish planning system with AI-assisted decision tools.
GLM-5.2 Targets Long-Horizon Reasoning Tasks
Zhipu AI's new model, GLM-5.2, is built for complex tasks that require sustained planning and multi-step reasoning.
Meta's Muse Spark Open Reclaims US Lead in Open-Weight AI
Meta's release of Muse Spark Open narrows the gap between American and Chinese open-weight models, but its restrictive license may complicate adoption.
Nextdoor Engineers Use Codex with GPT-5.5 for Hard-to-Reproduce Issues
OpenAI details how Nextdoor engineers use Codex with GPT-5.5 to investigate elusive bugs, build across platforms, and stay product-focused.
Google DeepMind Unveils Gemma 4 12B, an Encoder-Free Multimodal Model
The new 12-billion-parameter model processes text and images directly, dropping the separate vision encoder used by many multimodal systems.
Fusing MLP Layers Cuts PyTorch Overhead and Boosts Speed
A new Hugging Face walkthrough shows how fusing multiple linear layers into one kernel reduces launch overhead and improves performance in PyTorch.
Google proposes audit framework for verifying machine unlearning
A new Google Research framework aims to give regulators and users a practical way to check whether models have truly forgotten deleted data.
DiffusionGemma Speeds Up Text Generation by 4x
Google DeepMind's DiffusionGemma promises fourfold faster text generation, potentially making large language models more practical for real-time use.
OpenAI Details PRC-Linked Influence Operations Targeting US AI Debates
A new report from OpenAI reveals how PRC-linked actors are using AI to steer US tech debates, data center narratives, and tariff discussions.
LSEG scales trusted AI with OpenAI across global operations
A new OpenAI case study says LSEG is using the technology to accelerate insights, shorten release cycles, and reach 4,000 employees.