I'm an nerd. So my "fun learning project" to start the year is: working through the exercise section of Scott Wlaschin's (really excellent) Domain Modeling Made Functional… in four different languages I'm interested in:
These all have something of a family resemblance: all are descended more or less directly from Standard ML. So it's not like it's a super crazy thing. They're also all languages I have active interest in, and some degree of prior familiarity with, so it's not crazy in that regard. But they're also all just enough different from each other (from where they run to how they're built) to make the project interesting.