AFAIK speech recognition is non-trivial processing (though not an expert on it at all). Just a few points: I've read somewhat recently that the reason why Google's speech recognition is good is because it does a lot of the processing on their servers, also because it has a huge volume of data to work with (from worldwide users). All I know is that there are various factors at play (even though I'm aware phones are very powerful nowadays). It was an exploratory question, not meant to suggest that it cannot be done or will necessarily be too slow. And I did say "marginally to somewhat", etc.

>Are you sitting with your hands on your keyboard at all times while programming?

Obviously not. As a programmer, I'm well aware that there is a lot of think time involved.

>Do you need to move the cursor to a variable, function, etc to refactor? You are assuming that we can only say commands that are bound to a key? "Set breakpoint on line 20" -- You had to reach for the mouse in your world?

Nope, no assumptions. It was an open-ended question.

But anyway, that's not really important, I agree that good "intelligent" speech recognition would be very useful for uncommon tasks. But building a project, searching and replacing are tasks that are so common that I can't imagine having to reach for the mouse to do them, much less pronounce them out loud.

And imagine the mess if you're coding in an open space and everybody chitchats with their computers...

Yeah, someone always mentions the noise coding in a group. Someone else usually brings up subvocal speech, or something else, but I'll leave that for you to imagine. We need a few more people to imagine what is possible rather than rehash why something can't be done.