Eternal Haven

Home · Essays · Writing a saga without an algorithm

Craft

Writing a saga without an algorithm

A five-book fantasy series is the least optimisable object on the internet. That is not a complaint. It is the reason the thing exists.

Justin Helmer
6 November 2026 · about a ten-minute read

There is a kind of writing advice that arrives as a graph. Post at this hour. Cut the first two hundred words. Deliver a hook in the first sentence, a turn at forty percent, a cliffhanger at the end, and the next book in ninety days. It works, in the sense that a machine works. It produces books that behave like content: sized to the attention span, shaped to the thumbnail, and forgotten at the same rate they are consumed.

I wrote five novels that do not do that, and I want to explain what it costs, because most people who ask me about the series are really asking whether the trade is worth it.

A saga is a promise with a long delivery date

The first book opens with a promise made in front of witnesses. That is also what a series is: a promise to a reader that there will be a later volume, made before any of it exists. In an attention economy, that is an absurd instrument. It asks someone to invest eight hundred pages on the strength of a cover and a first chapter.

Which means the first chapter has an impossible job. It has to be good enough to earn the investment while refusing the tricks that make a first chapter "good" in the algorithmic sense. No mystery box. No shock death on page three. The Moonlit Slumber opens with elders talking about a dawn before the first dawn. A hook-writing coach would cut it. I kept it, because the reader I want is the one who reads the word elders and leans in.

What patience buys

Consequence. This is the whole return on the investment, and it is invisible until book three. When a reader has lived through a festival and a trial and a fall, a siege is not a set piece; it is a bill. You cannot get that effect with a cold open and a recap. You get it by spending chapters on things that do not pay off immediately.

The second thing patience buys is a cast that can surprise you. Corvath was introduced in the first book as a fall. Four books later he is seated among the Twelve, and the reason that lands is not a twist; it is that the reader has been carrying him for a thousand pages. A character who appears for one act cannot be redeemed, only replaced.

The costs, honestly

Momentum. A saga is slow to attract readers and slower to convert them, because the value is back-loaded. The first book has to be given away in patience before the second can be sold at all. Most independent writers run out of money or nerve before that curve turns. I am not immune; I just have other work that pays the hosting.

Correction. Once a book is published, its errors become canon. A saga compounds its own mistakes: a weak decision in volume one is load-bearing by volume four. When a standalone novel has a flaw, you fix it in the next draft. When a series has a flaw, you write a chapter that makes it look intentional.

Language drift. Four years of writing changes a prose style. Book II has a chapter title that breaks off mid-thought, deliberately. Book IV would never risk that trick, because by then the saga has a settled voice. Readers notice. A few of them prefer the earlier, looser voice, and they are right to; it was a different person writing.

What the algorithm would have me do

Split the first book into three. End each volume on a cliff. Add a romantic subplot with a guaranteed audience. Cut the choir chapters, which are the ones nobody quotes and everybody rereads. Make the dragon a recurring villain instead of a reconciled one, because redemption arcs do not screenshot.

Every one of those changes would raise the numbers and lower the ceiling. A book that ends on a cliff is a book you finish out of irritation. A book that ends on a settlement is a book you think about in the shower. I know which one I want strangers to have.

The one place I do use the machines

I use tools for the work around the work: hashing files, checking that a page loads, generating a background image, finding a broken link. I do not use them to draft the prose, and not for the reason people assume. It is not purity. It is that the prose is the only part where the specific person matters. A plot can be outlined by anyone. The sentence where Serenya's silence is described as louder than her speeches can only have come from one desk.

That is the trade in one line: let the machines own the scaffolding, and keep the sentences. If you are writing something long, the scaffolding will tempt you, because a saga looks like a product from the outside. From the inside it is a stack of decisions, and the only thing that makes it worth a stranger's evening is that a particular person made them.

If you want to start: Book I, and the reason the order matters is in Reading Eternal Haven in order.

Justin Helmer

Writes the Eternal Haven saga and the essays on this domain, publishes music as Excavationpro, and keeps the lantern on at /start.html.