It's too bad there is no decent way to use speech recognition to translate voice into code.

I often find myself using transcription while driving to send long emails for example... I'll dictate stream of consciousness while driving, including the um and ahhs, then when I get in, the email will just need editing to send.