Voice capture is not just transcription. It is intake for thinking.
The tricky part is that spoken thought is messy for a reason. The urgency, hesitation, sequence, and weird emphasis are often where the useful signal lives. Clean punctuation helps. Over-polish hurts.
I want the capture layer to make a brain dump easier to work with without sanding off the parts that made it alive. The goal is not to turn everything into a generic memo. The goal is to preserve enough intent that the real thinking can continue.