Structured Story Planning for AI: Making Your Story Legible

You already have the structure. That is worth saying before anything else, because most writing about structured story planning opens by telling working novelists they ought to plan more, which is both condescending and untrue of the people who actually read it.

You keep character sheets. You know which act a chapter belongs to and what that act is supposed to accomplish. Somewhere there is a document with the rules of the magic system, or the corporate hierarchy, or the timeline you had to draw twice because the second book broke the first version. That material is not decoration on the writing — it is load-bearing for it. You built it because you could not hold a 300,000-word series in your head, and it works.

And then, when you want help with a scene, you paste it into a chat box. Again. The character sheet, the two paragraphs of context, the reminder about who knows what. Tomorrow you will paste some of it again, because tomorrow is a new session and the session before it left no trace.

That is the odd part. The structure exists. You maintain it. It is just not legible to anything except you.

What "legible" actually means

Take one fact, at the smallest useful size. The captain lost the use of her left hand in chapter nine.

Right now that fact almost certainly lives as a sentence — a good sentence, in a paragraph, in a scene, three-quarters of the way through a file. It is completely clear to a human reader. It is also, to any tool holding that file, indistinguishable from every other sentence around it.

Now record the same fact as a typed Property on the element it belongs to: the captain. Nothing about the story changed. Three things about the artifact changed.

It became addressable. A fact on an element can be asked for by name. You do not have to know which chapter you wrote it in, or search for a word you may not have used — "crushed," maybe, or "ruined," or nothing so direct at all. There is a place the fact lives, and that place has a name.

It became bounded. The fact has edges now. It belongs to the captain rather than to the paragraph it happened to appear in. It travels with her, into every scene she walks into, in book three as much as book one. Left in prose, the fact belongs instead to a location in a file — and a location in a file is a terrible place to keep something that follows a person around for four hundred pages.

It became typed. Something reading it can tell what kind of thing it is: a persistent physical trait, not a passing action, not a metaphor, not a line of dialogue in which someone is lying about her hand. Typing is what separates "this is true of her" from "this was said once, in a room, by someone with a motive."

Here is the part that matters, and it is smaller and harder than the version usually sold. None of those three is a claim about an AI. Addressable, bounded, typed — those are properties of what you wrote down. They are true of a structured story sitting on a disk with the network unplugged and no model anywhere near it. The fact has edges or it does not, and whether anything ever comes along to read it does not change the answer.

That is the whole of what structured story planning for AI asks of you: not more planning than you already do, but planning with edges.

Why the wiki in the other tab isn't this

The obvious objection is that you have already solved it. The character sheet exists. The world rules are written down. Isn't that structure?

It is — for you. A planning tool that lives apart from the prose is legible to its reader, which is the point of it and no small thing. But the legibility stops at the tab boundary. The tool holding your manuscript cannot see the tool holding your plans, and the tool holding your plans has never read a word of the book. Neither one is wrong. They were built as separate categories, decades apart, by people solving separate problems.

The cost lands on you, in the only currency that matters: you become the integration layer. Every time the two need to meet — every session, every scene where the AI needs to know something you already wrote down — a person carries the fact across by hand. That person is you, and you are doing it inside the ninety minutes you set aside to write.

The alternative worth wanting is not a better wiki. It is planning that lives with the prose — in the same place, at the same grain, so that nothing has to be carried anywhere by hand.

The shape Sagaly will give it

Sagaly will model your story as a typed hierarchy — Universe > Epic > Series > Story > Act > Chapter > Scene — with typed Properties attachable to any element in it. The captain will be an element; her hand, a Property on her. The act she is introduced in will be an element too, and what that act has to accomplish a Property on it. It will be the same material you already keep, put somewhere with names and edges instead of somewhere with paragraphs.

Sagaly is designed so that structure enables intelligence. That is the commitment, and it is a statement about how the application is designed, not a report of what it does for you today. What it will be built on is specific enough to check.

You will be able to search the whole hierarchy by meaning rather than by string — which is exactly the thing a flat file cannot give you. You cannot search one long document by meaning across a novel and get structure-aware results back; across a typed element tree, you will be able to. And the collaborator will read the elements that bear on your question before it answers, and it will cite the ones it drew from, so you can see the scenes and notes an answer came out of instead of taking it on faith. Ask your story a question — and Sagaly will answer from your own canon.

That is the second half of the argument in why your AI forgets your story: the fix for a finite context window is not a bigger one, it is a map. This is what the map is made of. And it is what the agentic-versus-generative distinction runs on — a system whose unit of work is reading before it answers needs something addressable to read.

What structure will not do

The honest section, because the overclaim here is easy and this audience can smell it.

Structure will not check your continuity for you. Recording the captain's hand as a Property does not mean anything is watching for the chapter where she catches a rope with both of them. An automated consistency checker — something that reads a whole manuscript hunting for contradictions and brings them to you unasked — is direction, further out, and it is not what any of this is arguing. What structure will do is narrower and still useful: it will make the question askable, in seconds, in plain language, instead of costing you twenty minutes of scrolling. Asking is not checking. We are not going to blur the two.

Structure will not write in your voice. It cannot and it should not. The sentences stay yours — that is the entire point of the arrangement, not a limitation of it.

Structure will not be a tax. You will be able to work as loosely or as rigorously as your process demands; nothing here requires you to fill in a form before you are allowed to write a scene. A novelist who wants to open a blank scene and go should be able to open a blank scene and go.

And the claim being made is not "the more structure you add, the smarter your AI gets." That sentence sounds generous and is actually a promise about a machine nobody has met. The claim is smaller, and it is about your side of the desk: what has edges can be found, and what has none cannot. That was true before any of this and will be true after.

Where this goes

One paragraph, labeled plainly as direction rather than as anything shipping. Further out, the elements themselves will remember — each carrying its own private, evolving memory of what it went through, rather than sitting still and waiting to be read. You will author the destination a scene must reach, and the path there will be discovered as those elements collide instead of scripted in advance. Authored destination, emergent path. That is the destination, not release one, and we will not pretend otherwise.

You are the author

The structure is already in your head, and in five other files, and in a spreadsheet you are quietly embarrassed by. The argument is not that you should build more of it. It is that the version you already built is worth putting somewhere with edges — where the next tool that comes along can find a fact by name instead of hoping it survived the window, and where you never have to introduce your own book to a stranger again.

You are the author. The story is yours, all of it, down to the captain's hand. Structure is just what makes it possible for something else to see what you wrote.