I largely agree with this, though I think:

1) there's some value in internally generated text (automatic documentation, autosummary) being less aggravating

2) there are assistive applications — speech generated from sign language and touch panel communication etc. — and those users deserve a tool that can communicate for them in a way that has controllable or even personalisable outcomes.

3) a process of de-sloppifying text might also help train thinking clarity.