I've been experimenting with Wispr Flow ("effortless voice dictation"), and woah, it's good.
I can talk quickly, quietly, with background noise, in my Australian accent, in an American accent, with pauses, "frequent 'umms'", and it works.
Working remotely often means working from one of two extremes: somewhere very quiet, or somewhere very public. In both settings, hearing myself trying to think and give instructions at the same time was confronting. It takes some getting used to, and I can imagine plenty of work environments where speaking freely into your computer simply would not be practical.
I started to look for uses for Flow that go beyond dictating something I could (and would) otherwise type. Talking is particularly useful for prompting AI. When I explain a problem aloud, I naturally include more context: what has already happened, what feels important, the trade-offs and what I want the result to be. Wispr's prompting guide makes a similar point. Spoken prompts tend to describe the desired outcome more naturally than tightly typed instructions.
One workflow I want to try next is immediately after a meeting. Claude already has my meeting transcript and generated summary. I can then use Flow to talk through everything the notes didn't capture, because maybe I didn't say them out loud, while it is still fresh: the context, the subtext, what needs clarification and what I think should happen next. The meeting notes are then accurate, and personal.
The accessibility side is just as important. For people managing RSI, arthritis, mobility limitations, visual fatigue or learning disabilities, voice can reduce the physical and cognitive effort involved in getting thoughts onto a screen. Flow also cleans up and structures natural speech, rather than leaving the user with a transcript that needs extensive editing.
Through all this experimenting, the question became less about whether Flow worked and more about where it actually fitted into my day. Speaking gets more out of my head, faster, but the real value is in what happens next: letting AI sort through it, find the shape and turn it into something useful. Prompting AI and downloading my thoughts after a meeting are the two best use cases I've found so far, and I'm sure there'll be more.