Every Foundation essay has been rewritten in David’s voice. All sixteen. Four sprint batches across two days.
The essays started as what you’d expect from AI-assisted writing — competent, organized, lifeless. They had the right arguments in the right order and read like a policy brief written by committee. They didn’t sound like anyone. They certainly didn’t sound like me.
The document that changed this is the Voice Reference. Thirteen registers, 372 lines. It catalogs how I actually write — the Oklahoma directness, the controlled profanity, the way I build arguments through concrete images before I abstract. The rhetorical patterns I reach for. The ones I avoid. The specific rhythm of short declarative sentences broken by longer ones that unpack what the short sentence compressed.
Here’s what surprised me: a fresh Claude instance reads the Voice Reference cold and writes in my voice within paragraphs. Not an approximation. Not “David-flavored.” Actual voice match — the kind where I read the output and forget I didn’t draft it myself.
That’s a new capability. Not the AI’s capability — mine. I can now produce writing at my quality level, in my voice, at a pace that wasn’t previously possible. The bottleneck moved from “can I write this?” to “do I know what I want to say?”
Sixteen essays. Two days. One voice. Every word sounds like it came from the guy who means it.
— David