all systems MULHERN hire me
System · Running

Voice corpus

Every writing session captured and distilled so a model can write in my actual voice.

Runningcapture pipelineLLM2026
[ voice corpus screenshot ]

The job

Getting AI to write like a specific person fails when it has nothing real to learn from. Generic output reads as generic.

The build

A capture system saves writing across sessions into a structured corpus, distilled into a voice prompt with concrete rules and real samples that any drafting step reads before it writes a word.

What it proves

Voice is a data asset. The capture-and-distill approach personalizes any generation task to a person or a brand.

Every drafting step loads the distilled voice prompt first, so the rules and real samples constrain the output instead of a generic house style.
Voice as data
read before every draft
Build me one like this