AI systems are moving beyond answering questions. They can write and run code, search for information, check results, and keep working towards a goal. Increasingly, they are being used to tackle mathematical problems, too.
What changed? What turns a language model into an agent, and how autonomous are these systems really?
This talk takes you inside the loop. Using coding agents such as Claude Code and Codex as examples, we’ll unpack how models, tools, and the software around them work together. You’ll learn what an agent harness does, where MCP fits in, and why even powerful agents still get stuck.
Finally, we’ll explore how you could build an agent tailored to your own needs, and what these systems might mean for mathematical research and your future work.
No prior knowledge of AI agents required. Just curiosity about how they work.
Speaker
Alexander Kammerer is a software engineering lead at finccam, a Munich-based quantitative asset manager, and president of the FiMaTUM alumni club.
Registration
Please register by e-mail to fimatum-alumni.ma@tum.de. This talk is organized by the FiMaTUM Alumni Club. All students and alumni are welcome.