not everything that is old is bad
k3 is a refinement of a refinement […] of APL and math notation that is both terse and composable
definitely very much worth learning today
especially if you are interested in data crunching, machine learning, and high performance computing (think GPUs, SIMD, …)
the patterns at the core of k and APL are everywhere, once you get into them you see them everywhere
the world looks different, more rich
"A language that doesn't affect the way you think about programming, is not worth knowing." — Alan Perlis
when you flip it around:
k is worth learning
in the 90s, today, and possibly 100s of years from today
IMHO the biggest thing holding K back is lack of a production-ready / batteries included open source implementation. (There are a couple nice open source implementations of K but none of them are as feature complete as this one... Plus this one has at least a start on interoperating with .NET)