Voice recordings work better than written docs when prompting AI for code. Tools like WhisperFlow and MacWhisper let you capture raw speech with natural repetition, and AI models respond better to that unpolished thinking than polished requirements.
Move over VibeCoding, BlabCode. The big secret weapon to a lot of people who are able to get a lot out of these AI system development tools is their voice. They get a recording system like WhisperFlow is good for this, SuperWhisper is good for this. I use something called MacWhisper, and there are other tools like it. Some of them have a one-time[No speech] But what they all have in common is they allow you to talk for a while. And usually we talk in a way that is not nearly as refined as the way that we write. AI likes that. AI wants the noise. AI wants you to repeat yourself and sort of circle back around to the same idea over and over. When you do that, it gets an even better sense of your intent. So don't worry about trying to create a perfect product requirements document and feed that to the AI. Blab at it. Talk for a while. Give it that raw stuff. It will figure out what you mean. It is remarkable at that job, and you get much better results much, much faster.
