The seminar “AI for Math and Optimization” explored whether students with limited prior background in optimization could engage meaningfully with open research problems when supported by modern AI tools. Most participants were bachelor students (18/35) or master students (17/35) without formal optimization training. In a traditional setting, it would be unusual for such students to independently learn enough optimization theory in one semester to make research-level progress on open problems. The seminar tested whether AI could lower this barrier by acting as a tutor, literature assistant, coding assistant, proof critic, formalization assistant, and exploration partner.