This weekly update highlights how intelligent AI agents are changing the way scientists work and make discoveries. An AI agent is like a robot brain that can understand what you ask it to do, figure out a plan, and then complete the task without you having to give it every single step.

One major story this week involves OpenAI and a technology called Sora. Sora was a program that could create videos by itself using artificial intelligence. It seemed like an amazing tool at first, but on March 24th, OpenAI announced they were shutting down Sora for public use. Why? The company explained that it cost way too much money to run. Every minute of video created by Sora needed so much computer power that the costs were not realistic for a business. This teaches us an important lesson: even the smartest AI companies have to make tough choices about which AI tools they can actually keep running in the real world.

Despite that setback, this week also brought exciting progress. On March 25th, developers announced that a special technology called the Model Context Protocol (MCP) had reached 97 million downloads and installations. To understand why this matters, imagine if all the tools scientists use—databases, calculators, research libraries, and measurement instruments—suddenly spoke the same language. That is what MCP does for AI agents. It lets them talk to almost any tool or database in the world. Every major AI company, including OpenAI, Google, and others, now supports MCP. This means AI agents can now access scientific data, research papers, and experiments more easily than ever before.

On March 24th, a company called Oracle announced powerful new AI tools designed specifically for science and business. These tools help AI agents understand company data and research data better. Instead of scientists having to move their data from one computer to another (which takes time and can lose information), the new Oracle tools let AI agents look at the data where it lives. The company also created AI agents that are specialists—one is good at understanding scientific facts, another at analyzing data, and another at doing deep research. This is like having a team of robot assistants, each with their own special talent.

One of the most exciting developments for medical science came from researchers who created something called OpenScientist. This is an AI agent designed specifically to help doctors and scientists do research. OpenScientist can read through medical information, come up with new ideas about diseases, run tests on data, and discover patterns that humans might miss. It works like having a brilliant research assistant that never sleeps and can think through millions of possibilities. This could speed up the discovery of new medicines and ways to treat diseases.

Researchers also published important findings about how AI thinking differs from human thinking. AI is amazingly fast—it can look at information millions of times faster than a person. AI never gets tired or distracted. AI can find tiny patterns hidden in enormous piles of information, like finding specific grains of sand on a beach. But here is what AI cannot do well: it cannot truly understand why things matter, it struggles with situations it has never seen before, and it cannot explain complex reasons the way humans can. This research shows that the future of science is not AI replacing humans, but AI and humans working together. AI handles the fast, boring, repetitive work, while humans make the important decisions and understand what the results actually mean.

All of these developments point to the same direction: AI agents are moving from being interesting experiments to becoming tools that scientists actually use every day. The combination of better AI tools, universal languages like MCP that let agents communicate, and new tools designed specifically for science means that scientists around the world can now work faster and smarter than ever before. The work being done this week shows that in the coming years, AI agents will become as normal in laboratories and research centers as computers and microscopes are today.