The Brief
One of my first experiments, testing two things at once: whether what I'd learned about working with AI lands better as a story than as dry technical writing, and whether a loop of agents could make the writing better than I'd manage alone.
I wrote none of the prose. I supplied the raw material, defined what finished meant, checked every claim, and approved each pass. Agents drafted, critiqued and rewrote; a model from a different family graded each pass before it counted. Whether any of that helped is in how this was built.