Scientists around the world made huge steps forward this week in creating agentic AI systems that can work like smart assistants for research and discovery.

The biggest news came from Seattle, USA, where researchers launched a $1 million prize competition to fight Alzheimer's disease using special AI agents. This brain illness affects 152 million people and will likely affect many more by 2050. The competition asks teams to build AI systems that can study brain data all by themselves, without needing humans to guide them every step of the way.

These agentic AI systems are different from regular AI because they can make their own plans and take action. Bill Gates is helping to pay for this competition because he believes AI can help scientists discover new treatments much faster than ever before.

In China, researchers at Zhipu AI created something amazing called ComputerRL. This system teaches AI agents how to use computers just like humans do. The AI can move the mouse, click on buttons, type words, and run different programs without any help.

What makes ComputerRL special is that it combines two ways of using computers. Sometimes it uses programming codes to work very precisely. Other times it uses the regular computer screen that humans see. When scientists tested it, the AI got a 48.1% success rate on hard computer tasks. This score was better than other famous AI systems like OpenAI's tools and Claude 4.0.

Another breakthrough came with MedResearcher-R1, a special AI agent built just for medical research. This system can read thousands of research papers about health and medicine, then answer very hard questions that even human doctors might struggle with. It works by breaking down complex medical problems into smaller pieces, then finding the right information from scientific studies.

This could change how doctors and scientists do their work. Instead of spending hours reading through medical papers, they could ask the AI agent to find answers for them. This means new treatments and cures might be discovered much faster.

Robot science also got more exciting this week. Engineers created new ways for people to control robots using virtual reality headsets. These robots can now use both hands to do very tricky jobs, like putting together small machine parts or handling delicate materials.

The robots learn by watching humans do tasks first. Then they can repeat those same actions on their own. This learning from demonstration approach means robots could soon help with many jobs that need careful hand movements.

Businesses are getting very interested in these smart AI agents. New research shows that 68% of software companies now have agentic AI helping with their work. About 10.5% of companies are already testing these systems, and 27% more plan to start soon.

Companies that succeed with AI agents have something in common - they are really good at handling data. Every successful company had experience with analyzing business information, and most had used machine learning before. This shows that data skills are super important for making AI agents work well.

The business world is changing fast because of these AI agents. Research shows they can make companies 25-40% more productive and save lots of money. Some companies report saving 25-90% of their costs when they use AI agents for complex work.

What makes this week special is how many different areas are using agentic AI at the same time. From fighting brain diseases to controlling robots to helping businesses run better, these smart systems are showing up everywhere. Scientists believe this is just the beginning of a big change in how we use AI to solve hard problems and make new discoveries.