About

How this was made

The narrative of the eight volumes, their Editor's Notes and every attestation grade are the author's.

The plates, diagrams and atlas maps were generated by an image model (OpenAI) from a prompt written for each passage under the series' visual conventions. They are reconstructions to think with, not evidence; where they differ from the text, the text governs.

The podcasts were written by a language model (Google Gemini) from each volume's text and voiced by a speech model. The data tables, mind maps, infographics and slide decks were made the same way.

The companions marked NotebookLM (videos, conversations, slide decks, infographics, and the explainer slides set beside the passages they illustrate) were made with Google NotebookLM from the author's research notebooks of ancient texts. The maps marked Reconstruction were drawn for the series with an image model from the references printed on them. None of these is graded evidence; where they differ from the text, the text and its Editor's Notes govern.

The four compendia (Figures, Wars, Legions and Provinces) are the author's archivist's editions, compiled from the ancient texts and the modern references each names. Their portraits, plates and many of their maps were drawn with image models; other maps come from the open references credited in their captions. Like the Index, they are guides for orientation, not graded like the narrative.

The Crossroads, the questions at every scale, the modern parallels and the knowledge checks were written by a language model from each chapter's own text, under rules that forbid hindsight before the choice, invented characters, and invented quotations.

Your Crossroads choices, notes and reading progress are kept only in your own browser. This site sets no cookies and runs no analytics.