For the little data sciences like statistics and mildly complex ML, R is really good. At least for data exploration. The REPL/emacs combo I use is much better than python/jupyter. By better I mean it's far more productive to test ideas, compute statistics, charts, etc. Unfortunately R programs I write are not as structured as python programs. I wouldn't like to maintain that kind of code for a long time, nor would I build big applications in R.
