It constrains what’s possible, not what’s reasonable. For example, one could typically fit more text on a screen by compressing it, but most of the time, that’s not the reasonable thing to do.
I’m saying most of the time because of the existence of English Braille (https://en.wikipedia.org/wiki/English_Braille#System) which uses a compression scheme to compress frequently used words and character sequences such as ‘and’ and ‘ing’ shows that, if there is enough pressure to keep texts short, humans are willing to learn fairly idiosyncratic text compression schemes.
colorforth (https://en.wikipedia.org/wiki/ColorForth) is another, way less popular example. It uses color to shorten program source code.
One could also argue Unix, which uses a widely inconsistent ad-hoc compression scheme, writing “move” as “mv”, “copy” as “cp” or “cpy” (as in “strcpy”), etc. also shows that, but I think that would be a weaker argument.