I had it as a mind game for a while to design keyboard layouts that would be optimal for Swedish, German and English (my main languages). I found that you can only get so far with algorithmic approaches. I liked the ones I designed without computer help much better than the one i made with generic algorithms or machine learning, despite being worse on paper. Neither "combos" or distance travelled explained what made the layouts good.

Now, I never used any of the layouts for more than a month or two, but it was still a fun experiment.