I'm proud of my typos. They are like the proof that I can type way too fast, probably faster than I can think.
I love the auto-correct feature, once I got used to not going "AH! Typo! furious backspacing Retype correctly!"
I have one beef though. It seems that the typos I make are nonstandard enough that about 50% of the time, Auto-Correct chooses the wrong word. This irritates me even more than making a typo in the first place. It seems to happen often with double letter endings where I miss one, but that's certainly not the only case.
Example: When typing the word 'will' I often type 'wil', I have no idea why. Auto-Correct zaps this to 'ail'. So then, instead of backspace, type another L, space, keep going, I have to delete the entire wrong word.
The reason I am asking here is that I don't want to add misspelled words to the dictionary, I want to train the behavior of the Auto-Correct engine itself so that it's more accurate to my typos as opposed to the 'standard' ones.
Is this possible? I can't imagine that there's no capability for learning in that algorithm. I dream of a time when I can just mash away at the keyboard, and the computer just knows what I mean and records my rhoughts (There's another one: it gets instantly set to 'roughest' instead of 'thoughts') accordingly. If auto-correct can be trained, that takes us a lot closer.