Something that the article doesn’t seem to pick up on is how we read and dark modes effect on that depends on what we’re reading. (The musician example doesn’t quite do it justice).
AKAIK We see the shapes of words first, then some of the letters, then we understand the word. Hence why if y-u blank -ut s-me /etters, y-u can sti// read a sentence.
I find light on dark easier to navigate with code because the “shapes” are easier to determine.
Code is highly structured - even different variable names have different shapes.
Having said that...
Reading a full on research paper or something has to be dark on light for me. I’m reading something with less structure, where the words are one after the other.
So... yeah. It’s context and task dependent. So it’s a good thing there’s a choice.