This weekly update shows how AI agents are changing scientific research. These smart programs can now do tasks that used to take humans much longer, helping discoveries happen faster.

Microsoft released a new tool called Deep Research in their Azure AI Foundry. It uses AI agents to handle research in five steps: understanding the question, searching the web, doing deep analysis, checking facts, and making reports. This system is designed for businesses but could help research teams too. It costs $10 per million input tokens and $40 per million output tokens.

In Singapore, scientists are using AI to speed up materials science. Government labs and universities have AI models that simulate how chemicals behave. This lets them discover new sustainable materials much faster than old methods. For example, finding materials for clean energy or recycling might now take months instead of years. Singapore wants to be a leader in this kind of deep-tech research.

New research papers showed how AI agents can help in science. One introduced the AIRA-dojo framework, which lets AI agents automatically design, build, and improve machine learning models. This could make AI research faster and more reliable because the agents follow clear steps and can be checked by humans. It's like having a robot scientist that works alongside people.

For medical research, scientists created MedGemma, a new AI model that understands both medical images and text. Based on Google's Gemma 3 system, it helps with tasks like reading X-rays or explaining medical reports. This could give doctors better tools for spotting diseases early and planning treatments. The model works right away without special training.

Tools for building AI agents improved this week too. Clarifai launched features that let developers create custom AI agents more easily. Now researchers can upload their own tools (like data calculators) and connect them to AI agents. This helps scientists build helpers that understand their specific work, like agents that analyze lab results or manage research data.

These advances show AI agents becoming key partners in science. They help in different ways: Microsoft's tool tackles complex research questions, Singapore's AI speeds up materials discovery, and medical AI like MedGemma improves healthcare tools. As these technologies spread, they could make scientific breakthroughs happen faster across many fields.