Voice-to-Text for Developers: Why I Stopped Typing Half My Code Comments

I type fast. Probably 90-100 WPM on a good day. So when someone first suggested I try voice-to-text for development work, I laughed. Why would I dictate when my fingers are already on the keyboard? Then I timed myself writing a pull request description. Three paragraphs explaining a refactor — what changed, why, what to watch for in review. It took eight minutes. Not because I type slowly, but because I kept rewording things, deleting sentences, second-guessing phrasing. Writing prose is a different cognitive task than writing code, and the keyboard creates friction between thinking and expressing. I tried dictating the same kind of description the next day. Spoke for about 90 seconds, let the tool clean it up, made two small edits. Done in under three minutes. The output was arguably better because I'd just explained it like I was talking to a colleague, which is exactly what a good PR description should sound like. That was six months ago. Now I dictate roughly half of all the non-co