Programming exams in computer science or software engineering studies are usually far from the practical reality: Individuals solve programming problems in a tight time box and complete isolation. But do exams have to be like that? In this session, I first present some of my recent efforts to make programming exams more practically relevant, including the latest experiment, where students were allowed to ask experts for help (like they would on the job). The results of this experiment raise interesting questions regarding the competency of talking about code – with humans or AI. Based on these results and the experience of the participants, the second part of the session will be a brainstorming about the most relevant programming-related competencies in the age of generative AI and novel ways of teaching and assessing these.