How we use AI

What we do, and what the AI does

First, who "we" is. nüNERD is two people: a dad who went looking for real children's books about the things he loves and found dead ends, and his mother, an editor who taught school for thirty years and reads these books aloud to the grandchildren they were made for. There is no staff. Every choice that matters in every book is made by one of the two of us.

Every book starts with us. We choose the subject and go to the source: the Meditations for Marcus Aurelius, Plutarch's own text for the Plutarch books, the historical record for Cleopatra rather than the movie. We decide what the book is about, which scenes carry it, what a four-year-old's version keeps and what an older reader's version can finally say. We plan it chapter by chapter, design the look of the art, and write the instructions for every illustration, page by page.

Then the AI goes to work inside that plan. It drafts the prose we outlined and renders the pictures we described. That is the whole of its job: drafting and rendering, never deciding.

Then we take the draft back, because an AI draft is a starting point and nothing more. We rewrite hard, by hand, line by line: cutting what is padded, fixing what rings false, rebuilding sentences until the book reads right out loud. By the time one of our books ships, most of the hours in it are editing hours, not drafting ones.

What keeps it honest

Three checks, on every book, with no exceptions.

• The facts are checked against the original source. Before a book is done, every claim in it goes back to the real text: the translation itself for the Plutarch books, the Meditations for Marcus Aurelius, the earliest sources for Beowulf and Alfred the Great. This check regularly catches small errors that survived every earlier draft, a verb the source doesn't quite support, a detail in a scene that is slightly wrong. A book can read smoothly and still be wrong, and this is the step that catches it.
• An independent AI attacks every draft. A separate AI reviewer, given no context and no stake in the work, reads each draft looking for anything weak, padded, false, or machine-sounding. Its only job is to find problems, and nothing that helped write a book is ever allowed to approve it.
• A human reads every book, cover to cover, before it ships. Not a sample. Not a spot check. One of us reads the whole book, usually aloud, because reading a book aloud is the fastest way to hear what is wrong with it. If it isn't a book we would want on our own family's shelf, it does not ship.

Why we work this way

Here is the honest question underneath this page: if we care this much, why not write the books entirely by hand?

Because then almost none of them would exist. We don't publish one book on a subject. We publish the same story at several reading levels (ages 0 to 4 up through 10 to 14 for the deepest series), so a child can grow up on a subject and be ready for the real book when they open it, often years before they otherwise would. That means several full books for every single subject, and our subjects are things like Ulysses, Don Quixote, the Stoics, and Alfred the Great: wonderful, and never bestsellers. No publisher will fund a Beowulf for six-year-olds, never mind a version of it for every age from the crib up. The math has never worked, which is why you couldn't find these books anywhere, and neither could we.

AI changes the math. It lets two careful people make books that could never have paid for a team. So the real choice was never our books versus a hand-made version of the same library. Nobody was making that library, and nobody was going to. The choice was these books or nothing that carried a child all the way to the source. We made them, and we made them carefully.

Where we draw the line

The same tools that make our library possible also make it easy to flood a shelf with filler, and plenty of people are flooding. So here is our line: we only make a book when we can add something real. Either nobody has done the subject well for children (a Marcus Aurelius a three-year-old can hold, a Plutarch a six-year-old asks for again), or we honestly believe we can do it better or differently than what's out there. When a good children's book on a subject already exists, we would rather you buy that one. We build for the gaps: the subjects no one else will make for children.

A word about artists

We love illustrators, and we want to be straight about our plans for them. As our books prove out, we want to bring working illustrators in to re-illustrate them, with a share of what those editions earn. We want to give high school and college artists their first published credit, a real book with their name in it. AI is already taking work away from illustrators, whether anyone likes it or not, and we would rather be one of the places that gives them paying work in what comes next. If you're an artist and any of that interests you, get in touch.

If you would rather have zero AI, we understand

Some families draw the line at none: nothing their children read should be made with AI, full stop. We respect that, and we would rather you know exactly how our books are made and walk away than buy one without knowing. But if the worry underneath is the more common one, that AI books are soulless, careless filler, then we are on the same side of it. That is the exact book our whole process exists to prevent.

Judge for yourself

There is a simple test, and it beats anything we could tell you. Read one of our books to your child, then ask them to tell the story back. If the book is alive, they will have something to say, and how it was made will matter a good deal less than what it is. If it's hollow, no page like this one should talk you out of that.

The free Plutarch pack exists for exactly this: a working sample you can read tonight. Nothing to send, nothing to wait for, no email asked. Read it aloud, ask for it back, and decide for yourselves.