This weekly update covers the most important scientific discoveries and research breakthroughs powered by artificial intelligence agents. Scientists and engineers around the world are discovering that agentic AI - a special type of AI that can reason, plan, and make decisions without being told every single step - is making scientific work faster and better than ever before.

## Designing the Future with AI Agents

Engineers at Stanford University in California have created something remarkable called MetaChat. This is a new computer tool that uses AI agents to help scientists design optical devices. Optical devices are things that work with light, like special lenses or sensors. The problem is that designing these devices used to take a very long time - sometimes weeks or even months of hard work.

MetaChat is different because it combines multiple AI agents that work together like a team of expert scientists. One agent acts like an optics designer, while another acts like a materials expert. These agents don't just follow a simple list of steps. Instead, they can think about what they're doing, change direction if needed, and ask humans for help when they need it. One impressive test showed an AI agent designing a special metal lens that bends blue light to one spot and red light to another spot - something that would normally be very complicated.

Why does this matter? Metasurfaces - the tiny devices these agents help design - could lead to amazing inventions like super-clear holograms for games and movies, better cameras for our phones, sensors for robots, and even augmented reality glasses that don't need to be thick and bulky. The researchers believe that MetaChat could help solve a real problem: there aren't enough optical designers in the world, but many companies need them. With AI agents helping out, scientists and companies can design these cutting-edge optical devices much faster.

## AI Scientists Helping Human Scientists

Google Research announced that they've been working on brand new kinds of agentic AI that help with scientific discovery. They created something called an AI co-scientist - basically, an AI that works like a partner scientist alongside human researchers. This AI co-scientist can help generate new ideas for experiments and help scientists test those ideas.

But that's not all. Google also made an AI-powered tool that acts like an expert computer programmer and helps scientists write code. Scientists often need to write computer programs to analyze their data, but writing code takes time and knowledge. This new tool helps by writing expert-level code, which means scientists can spend more time thinking about their actual science instead of getting stuck on computer problems.

Google has been using these tools to help with some amazing science. Researchers used these AI tools to study how the human brain works. Scientists from universities like Princeton, NYU, and universities in Israel worked with Google to compare how the human brain processes language and how artificial intelligence processes language. They discovered something incredible: the way AI language models organize information is very similar to how our brains do it. This discovery could help scientists understand how our brains really work and could help us build better AI in the future.

Another exciting application is Earth AI, which uses AI power to look at massive amounts of information about our planet. Tasks that used to take scientists years to complete - like analyzing weather patterns or understanding climate change - can now be done in just minutes. This tool is already helping cities plan better futures and helping organizations respond to disasters more quickly.

## The Tools Behind the Science

NVIDIA, one of the world's biggest computer chip companies, shared news about new open-source AI models called Nemotron 3. Open-source means anyone can use these models and improve them. These models can be used to power the kinds of AI agents that researchers use in their work. Think of them as the engine that powers the AI assistants scientists use every day.

## What This Means for the Future

All of these breakthroughs show that AI agents are revolutionizing science. Scientists are moving past the idea that AI is just a tool that answers questions. Instead, agentic AI can handle complex, multi-step scientific work. AI agents can reason about problems, work with different tools and data, and keep learning from their experiences.

The beautiful part is that this doesn't mean robots are taking over science. Instead, AI agents free up human scientists to do what they do best - ask important questions, think creatively, and make big discoveries. The AI handles the repetitive work and the heavy calculations. Humans provide the wisdom, creativity, and judgment that science really needs. Together, AI agents and human scientists are creating an exciting future where scientific discovery happens faster than ever before.