while speech recognition may work (reasonably) well for 'normal' documents i see no future for it in the programming world
one reason is the nature of most programming languages (maybe except for cobol or stuff like that)
those are the reasons:
* programming languages make use of a lot of special characters
* if the language is case sensitive, good luck
* speech recognition uses quite some heuristics based on 'normal' language
thus, programming languages don't work well.
what i could imagine seeing tho would be macro-based voice recognition, which would need a special support packages for each programming language.
one would define standard blocks (for loops, class templates, switch-case, if-then-else, or more abstract templates) and name them eg. "new for-loop from 0 to 20" and it'll spit out "for(int i=0; i<20; ++i) { }"
this combined with classic keyboard input might have a future.
basically, replace everything you can't do by keyboard shortcuts (aka, you'd have to use your mouse) with a voice macro.
for everything else... no way