I stopped typing and started talking — voice memos plus AI gave me my commute back

The setup that changed my week costs nothing: I walk, I talk into my phone, an AI transcribes it, and a second pass turns the ramble into prose. Two hours of my week were already spent walking. They're now the two hours where most of my writing gets done. The keyboard didn't die — it just stopped being where first drafts happen.

The first surprise was quality. I expected dictation to produce sloppier thinking than typing. The opposite happened, and I think it's the backspace: typing lets you edit mid-sentence forever, which is how an hour produces three perfect paragraphs. Talking is committed — you finish the thought because your mouth is already halfway through it. The transcript reads looser and lands more often. Editing finished thoughts beats composing perfect ones.

The AI's role is specific. Transcription is solved, so that's table stakes. The real step is the cleanup prompt: 'keep my structure and examples, remove the verbal tics, don't add anything I didn't say.' Without that last sentence, the model polishes my voice into a podcast host's — smoother and not mine. With it, I get my argument in my words with the 'um's removed. It's a cheap prompt and it's the whole trick.

The failures are worth knowing. Names and numbers come out garbled — I now say 'insert product name here' instead of mangling a client's brand into the recording. Anything technical with precise terms gets typed later, because correcting a mis-transcribed acronym costs more than typing the paragraph. And walking while dictating in public earns you a reputation in your neighborhood, which is a cost I've simply accepted.

Ten years ago the advice was 'keep a notebook by the bed.' The modern version is 'your voice recorder is a first-draft machine that fits in a pocket.' The thoughts were always there on the walk. The difference is that now they survive it.