What happens to your book after you hit upload.

When you upload a book to Modulus, it looks like almost nothing happens. A progress bar, a couple of minutes, then it says ready. That quiet little bar is doing the most important work in the whole product. Here is what is going on underneath it.

I am building Modulus for people whose words are the product. Pastors, authors, the small publisher with a backlog of recordings. For them, an invented quote is not a small bug, it is a machine putting words in their mouth. So the engine is built backwards from most AI tools, and that wait is the reason it can be trusted.

It reads the whole book first

Most AI tools start the moment you type a prompt. They lean on whatever they absorbed from the internet and hand you something that sounds about right. Modulus does the opposite. Before it writes a single word, it reads your book end to end. It pulls clean text out of the file, even when that file is a rough PDF export or an audio recording, and breaks the book into small passages it can work with precisely.

Then it builds a map of what is inside

This is the part that matters. Modulus goes through your book and pulls out its real building blocks: the themes, the claims you make, the quotes worth keeping, and every scripture reference, exactly as written. Then it pins each one to the precise spot it came from. We call these atoms, and the map of them is what the whole product is built on.

Two things happen at the same time here. One pass turns your passages into a search map, so any idea in the book can be found in an instant. The other pass builds the meaning map, the atoms. When both finish, your book is fully understood, and all of it stays on your machine.

Under the hood

1Indexingruns once, the moment you upload

Read & extract

Pull clean text from your file, even a messy PDF or an audio recording.

Break into passages

Split the book into bite-size pieces it can handle precisely.

two things at once

Build a search mapso any idea can be found instantly

Map the meaningevery theme, claim, quote and verse, pinned to where it appears

Indexed and ready

Your book is now fully understood, and stored on your machine.

the indexed book feeds every piece you make

2Generationruns each time you create a piece

Find the relevant pieces

Pull the exact parts of your book that fit what you are making.

Draft from them

Write grounded in those pieces, with every line traceable to the source.

Check it against your book

Re-read the draft and cut anything the book cannot back up.

Deliver, with receipts

You get the piece, each claim cited back to your own words.

Nothing comes from outside your book. Every piece traces back to a passage you wrote.

An atom is one theme, claim, quote or verse from your book, pinned to its exact source. When a passage trips a model's content filter, that section quietly re-runs on a backup model so nothing gets dropped.

Nothing gets written that was not already yours

Once your book is indexed, making a piece is a different kind of process than prompting a chatbot. When you ask for an article, a Q&A, or a set of quotes, Modulus first finds the exact parts of your book that fit, drafts from those parts with every line traceable, then re-reads its own draft against your book and cuts anything the book cannot back up. What you get is cited back to your own words. I wrote more about that checking step in why we check every draft against your source.

Why we do the slow part up front

All of that reading and mapping is why the wait exists. A short book takes a couple of minutes, a longer one a little more. We could skip it and start guessing like everyone else, instantly. We do not, because the upfront work is the entire point. Do the careful reading once, and everything you make afterward is both fast and faithful.

The engine is quiet on purpose. You should be able to hand it your life's work and spend your attention on your message, not on whether the machine made something up.