VoiceNote · Blog
From voice to text. From text to practice.
What we learn about thinking aloud, working with AI and building VoiceNote.
The text your recording gives back isn't text yet
You record a voice note in a Google Meet meeting thinking you've solved it — depending on the length, you have to wait to receive it, then you need to click multiple times to find and open the transcript. The audio became words, and they didn't come to you.
Talking to an agent is not talking to yourself
When you record an idea to document something or to tell someone, what comes back is a draft: you will reread it and tidy it up. When you record something to hand to an agent — a coding session, an assistant that will act — what comes back is the instruction.
Dictation is not long-form transcription — I tried five tools before writing my own
Dictation is writing with your mouth: you say a sentence, look at the screen, fix it. Transcribing long-form audio is a different problem — you talk for fifteen minutes without looking at anything, and you want the whole text afterwards.
Speaking the prompt instead of typing it: why VoiceNote started as a terminal script
Anyone building with AI learns early that the output depends on the context you give — and that giving context is expensive. Typing, you summarise.
The AI mistakes that cost you don't look like mistakes
When people talk about AI errors they picture invention: a book that does not exist, a number nobody published. That error is annoying, but it is the cheapest one — it looks wrong.
The idea survives the walk and dies in the editor
I build the whole argument walking the dog. Beginning, middle, the objection someone will raise, my answer to it.