Chapter 14. Before the Empty Vault: What This Book Has Tried to Say

Return to the first sentence of Chapter 1. It is impossible to explain Star Wars in one word. The Force, Jedi, lightsabers, Darth Vader, the Rebels, and the Galactic Empire are not isolated points but an interconnected web of relations, and that whole web is Star Wars. An IP is the network made by every word that constitutes it.

Over fourteen chapters, this book has guided the deliberate construction of that network. From the four-stage pipeline through the nine classifications, worldbuilding axes, idempotency rules, validation, and operating cycles, these were procedures for bringing an IP’s scattered mental network into the tangible material of an Obsidian Vault.

Now on the reader’s desk lie the template for a thirteen-column CSV, the nine-classification guide, and Markdown files that are still empty. No line connects those files yet. But the reader already knows that when the first row is filled, Stage 1 extraction begins; rows grow into hundreds, become headwords in Stage 2 consolidation, turn into Markdown files in Stage 3, and receive narration in Stage 4. Lines will be woven between the points, and a map of one IP’s network will unfold on the screen.

How Librarying differs from existing creative practices

IP has always been managed somehow: a writer’s private setting notes, team-shared documents, Google Sheets, and wiki sites have each held IP information in their own ways.

Yet these methods share four limits: unstructured formats, dependence on a single author, lack of version control, and difficulty of transfer.

Setting notes are written however a writer wishes. A person may understand them, but machines cannot readily read and process them; searching, consistency checking, and reuse in other tools are difficult. The real setting often remains in the writer’s mind, with documents holding only part of it, leaving much information uncertain when the writer is absent. As an IP develops, settings change, but tracking which version is current is difficult; material changed in edition three may leave edition-two settings elsewhere. When the IP passes to another team or is reopened after a long time, much of it remains only in the original creator’s memory and resists transfer.

Librarying is a methodological response to these limits. The thirteen-column CSV and Markdown structures are predetermined, so machines can read them, automation can process them, and other tools can reuse them. A Vault lives in a file system, not in one person’s head; even someone seeing the IP for the first time can open it and verify official settings. The version-status and source-list fields in frontmatter preserve a headword’s version history, while idempotency rules protect existing work during updates. Handing over one Vault folder hands over the IP’s official vocabulary and worldbuilding structure together.

This is where Librarying improves on existing practice. It is not a method for becoming a better writer; it is a method for building a system that manages an IP better.

What this book has tried to convey

Reduced to one sentence, the fourteen chapters say this: Librarying is not making a wiki with Obsidian; it is building the one original from which a producer operates the whole IP.

The difference is not the container but the way of cutting. Even in the same Obsidian, dividing through the two coordinates of nine classifications and worldbuilding axes turns this Vault from a searchable dictionary into an operational original. New short fiction, new characters, foreign-language translation, and encyclopedias all pass through this one volume.

That is why the Vault does not end when it is completed. New vocabulary arising from those tasks returns through backflow, and idempotency rules protect existing narration as the next cycle is built upon it. So long as backflow does not stop, the IP remains alive.

The producer’s role: the hand that decides

One point has been emphasized throughout this book: a producer is not merely a writer.

If a writer writes, a producer writes while also managing, delegating, reviewing, and rewriting. Writing is only one of several tasks.

Librarying is producer’s work. The producer designs the four-stage pipeline, decides what and how to delegate to the AI work partner, reviews results, and operates the Vault. It includes stages written directly (direct hub writing), delegated stages (leaf narration), and confirming stages (validation).

Chapter 2 said, “The hand that decides is human; the hand that spins out is AI.” Having read fourteen chapters of Librarying, the reader should be able to feel how this sentence works.

Humans decide which of the nine classifications applies and design the worldbuilding axes. Humans write hubs directly, fix P0 issues immediately, and keep the Vault growing through backflow. What AI produces but humans review and pass fills the Vault; what AI gets wrong but humans correct makes its quality. When this division of work functions well, a scale difficult for one person to carry alone becomes achievable in realistic time.

To the reader beginning a first cycle

Start with a small IP. There is no need to target 1,500 headwords in the first cycle. Choose only five to ten core documents and first make a Vault of 300 to 500 headwords. The aim is to experience the whole four-stage process once. Completing a small IP to the end is far more valuable than trying to perfect a large IP from the beginning and stopping midway.

For first practice, worldbuilding axes may begin lightly with three or four. Moving to an operational Vault is different: the profiling stage should establish five to seven core axes plus “general” and “unclassified,” and Stage 2 consolidation and validation then work within those allowed values. This does not mean guessing perfect axis names from the start, but it does mean writing down which axes you will test before work begins.

For a first practice Vault, Stage 2 consolidation need not be perfect. If the file exists to learn the full flow, it may be pushed through Stages 3 and 4 at 85–90 percent. But an operational Vault handed to other work requires a different pass line: synonym bundles, canonical keywords, worldbuilding axes, and version status must reach roughly 95 percent, and automated validation and handoff checks must pass. Waiting for perfection and never finishing is a problem; mistaking practice-level looseness for an operational standard is also a problem.

In Stage 4, write several hubs directly first. Three to five are enough to establish a tone in a small practice Vault; for an operational first cycle, five to ten are recommended. Their narration becomes the benchmark for what follows. Many people stall Stage 4 by trying to write every hub alone. The point is not the number itself but creating enough clear examples of voice and perspective to hand to an AI work partner.

When the first Vault is complete, open it again the next day. After a night’s distance, the same Vault looks different: missing places become visible, and wrong connections emerge. They become the first P2 issue list and carry over unchanged into the next cycle.

What continues in companion texts

How to use a Librarying Vault to write short fiction is covered in Storytelling 100. Vault character headwords become character sheets, and worldbuilding axes become axes for everyday-scene material.

How to use a Vault in translation work is covered in Seoyeongak. Translation-note, alternate-name, and alias columns become material for translation consistency.

How to reference the Vault’s worldbuilding axes and character-relation network while shaping a character from zero is covered in New Character Build, the text introduced as this book’s companion in Chapter 1. How to gather 1,500 Vault entries into one published encyclopedia is covered in LOREBOOK.

These texts and this one connect to one another within a single pipeline.

Before the empty CSV

The moment the first row is entered into an empty CSV file is the beginning of Librarying.

IDX, classification, keyword, description, source. The first row receives its headers; the next receives its first keyword. Stage 1 extraction begins.

That keyword passes through hundreds of rows, is gathered into a canonical keyword in Stage 2 consolidation, becomes a Markdown file in Stage 3, and receives narration in Stage 4. The file becomes a node in Obsidian’s Graph View, and links connect to it. Nodes accumulate into a network; that network is the map of the IP’s world.

The map that was empty at first is now filled with 1,500 points and thousands of connecting lines. Characters live there, places exist, events occur, devices operate, and concepts flow. The IP’s world has unfolded as one map.

Just as Star Wars cannot be explained in one word, neither can the reader’s IP. That network is, for now, only in the reader’s mind.

Reading ends here. Return to the desk and write the first row of the empty CSV. The IP walks onto the screen only from that one row.