The thing that puzzles me about txtspell is that it's supposed to save time, but with if you use "predicted text" in a mobile phone it's surely liable to be slower if you don't use normal words and spellings.

And since most people read with their eyes rather than their lips it takes longer to interpret as well, especially if any, even slightly unusual, words or expressions are used. (Just turn any extended passage into txtspell and it looks like klingon, something to be decoded, rather than scanned instantly in the way standard spelling normally permits.)

I am sure there will be some txt abbreviations which will become standardised and recognisable, in the same way that "&" is recognised as an alternative to "and", or "etc" for "etcetera", but that's a different matter from having whole passages written in improvised code.