Short version: keep the worldbuilding in the same place as the draft, title every note with the words you will actually type into a search box later, write each fact exactly once, and record the chapter it first appeared in. The test is crude and it works. Look up one small thing and time yourself. Still hunting after two minutes? The filing is wrong, and your memory has nothing to do with it.
Deep into a long serial, notes rarely go missing. They go unfindable. This piece is about filing. Inventing the world is a different craft.
Why the lookups start failing around chapter 40
Three causes cover nearly all of it.
The worldbuilding lives in three places. Some of it sits in Obsidian. Some is a parenthetical you typed into the manuscript at 2am. Some is in a Discord thread with your writing group from last March. Each pile made sense on its own. Together, nobody can say which one is current.
Titles got named after categories. You made a file called “Factions and Politics.” What you need at chapter 41 is the name of that duke up north. Category names come out of the writing mood. Search terms come out of the panicking mood. They almost never match.
The same fact got written down twice, and the two copies disagree. This is worse than losing it, because you do find something. You find the stale copy, and it goes into chapter 28.
Rule one: notes and manuscript share an address
The cost of splitting them is not an extra window. The cost is that you stop checking. Drafting is a state you do not want to leave, so the moment a lookup means tabbing out to another app, you start gambling: “chapter seven, probably, I’ll catch it in revision.” That gamble is where continuity breaks. Forgetting comes much later, if ever.
Same place means one file tree and one search box. Swap tools freely: Scrivener, a plain folder of text files, Slima. The condition is what stays fixed.
Rule two: name the file what you’ll search for
While typing a filename, ask yourself one question. Three months from now, what words will I type?
- ❌ Factions and Politics → ✅ Northern noble houses, who’s who
- ❌ Magic system → ✅ What casting costs you
- ❌ Character sheet: J. Doe → ✅ Why J. Doe won’t go home
That third one is deliberately clumsy. When you go looking, the clumsy phrase is what’s in your head. Titles don’t have to be tidy. They have to match your own panic vocabulary.
Rule three: one fact, one home
Exactly one place gets to be authoritative. Everywhere else links to it. Nobody copies.
Copying feels like insurance for about a week. Then the copies drift, each gets edited on its own, and now you own two versions and zero answers. Easy diagnostic: the second you catch yourself wondering which of two files is newer, you have already split.
Rule four: write down the chapter it first appeared in
Most skipped. Biggest payoff.
Notes usually record content. What the sword is called, who forged it, what it does. But the question that stops you cold at chapter 34 is positional: which chapter did the sword first show up in? What you are really checking is how much the reader knows by now.
One line at the bottom of each note, “First appears: ch. 7,” costs nothing to maintain and turns a twenty-minute scroll-back into a two-second glance. Foreshadowing needs it most. A planted detail with no chapter recorded is a planted detail you will eventually fail to pay off.
When not to make a note at all
Worldbuilding is addictive, and it resembles writing closely enough to be an excellent way of avoiding writing.
A threshold that holds up: file it on the second lookup. The first time you go hunting for something is chance. The second time means it will keep happening. Until then, leave it in the manuscript.
A complete world and a good story are separate achievements. Filing keeps you from stalling. It will not make the book better.
Steal this table
Five columns per note. Everything past them is optional.
| Field | What goes in it | Why it earns its place |
|---|---|---|
| Search term | The words you’ll type three months from now | The title is your search interface |
| One-line definition | Keep it to a single line | You’re confirming one thing, not reading an essay |
| First appears | Chapter number | Tells you what the reader already knows |
| Hard rule | The constraint this thing can never break | Late chapters break these first |
| Links | Point at related notes; never paste them | Keeps a single version alive |
A forty-field character questionnaire never gets filled in. A five-field one does.
What this doesn’t fix
It solves findability. Memorability is a separate problem, and so is whether your world is interesting in the first place. Flat worldbuilding survives any folder structure you throw at it.
It is also close to pointless on short work. Under 50,000 words you probably do remember, and the upkeep costs more than it returns. This is for the people still going at chapter 20 who intend to reach chapter 60, and for anyone posting weekly on Royal Road with a KDP box set somewhere down the line.
Slima keeps notes, chapters and versions in one file tree, so a lookup is one search instead of a trip out to another app. We asked seven paying users which feature they would least want taken away. Six of them named the structural side, not the AI. Try it.