Ask HN: How do I examine an undergraduate programming course in the era of GPT?

I teach a course on Python programming at a UK university. The assessments are 'take-home' programming problems a little like Advent of Code. Cheating is always a concern but now it is trivially easy to generate code that gets a very high mark (I have checked). What do HN readers suggest as an strategy moving forward? In person exams are possible but not ideal.