This weekly update shows how AI agents are becoming real and helpful tools for science and business. An agentic system is a group of AI helpers that work together to solve hard problems without needing a person to tell them every single step. This week, scientists at Duke University showed that AI agents can solve complex design problems almost as well as trained experts. The three biggest AI companies—Google, OpenAI, and Anthropic—are all racing to build better agentic AI systems, each with different strengths. OpenAI created new tools called AgentKit to help developers build agents more easily. Google announced Gemini Enterprise, which helps companies use AI agents safely and keep everything under control. Anthropic is focusing on keeping humans in charge with their Computer Use feature.

A big gathering at Michigan State University brought together over 200 people to discuss how agentic AI is changing jobs and science. They learned that AI agents work best when they help humans do better work, rather than replacing people entirely. In the medicine field, companies are already using NVIDIA-powered agentic AI to help develop new drugs faster. Scientists at Google DeepMind made progress in teaching AI to understand the world using sight, sound, and movement all at once, which helps robots and accessibility tools work better. New computer chips are making it possible to run powerful AI on smaller devices, not just in big company data centers. The week also highlighted important safety work, with OpenAI launching new safety tools to help organizations use AI responsibly. All these breakthroughs show that agentic AI is moving from science experiments into real tools that solve real problems in hospitals, factories, and research labs.