I was producing more than I ever had. Sounding less like myself than I ever had.
At one point I had fourteen AI tools open at once. Research papers, client work, MBA deadlines. The output went up. The writing stopped sounding like a person, and specifically it stopped sounding like me.
Volume without voice. That is the trade nobody warns you about.
So I took my own writing apart. The rhythms, the fragments, the dry line that lands before the sincere one, the words I never use. All of it became rules a model can actually follow, in one file.
Three passes run every time. The first strips the tells that make writing read as machine-made. The second puts the voice back in. The third hunts down every last em dash, because that one is personal.
I use it for everything public I write. It is also the standing rule for every message my own automations send, which is the real test: a generic message gets ignored, and a message in your own voice gets answered.
This product is what got me invited onto the Analog to Automation podcast, episode two, where Sarah Mitchell asked me the question most people skip: what was the first thing you automated, and what changed. It lives on Gumroad: kitaa.gumroad.com/l/find-your-voice. MIT licensed. Questions or ideas: say hi.