Chapter 10. Phases 0–3 — The First Half, from Agreement to Pilot Part

Picture the first writers' room meeting for a film. A printed PDF lies on the table, its cover marked NDA and Show Bible v1.0. The first hour is spent locating what the document contains; only then does work on the season begin. The first month of a LOREBOOK project is similar. It is not a time for drafting body prose, but for gathering every decision about an IP into one document.

The deliverable of Phase 0 is not a book, but a packet of agreements.

The first two to four weeks consolidate decisions about how to handle 1,500 entries into a ten- to twenty-page document. This document becomes the SSOT for the next nine to fourteen months. Only after it is settled can the skeleton, pilot Part, and expansion proceed in order.

The term Phase comes from game development, where one stage's output becomes the next stage's input. A LOREBOOK has seven stages, Phases 0–6, and each stage's acceptance criteria form the gate into the next. This chapter covers the first four: agreement (0), skeleton (1), pilot Part (2), and expansion (3)—roughly the first four to six months.

The Timeline of Seven Phases

Phase 0 spends two to four weeks producing the agreement. Phase 1 spends one to two weeks building 1,500 empty LOREBOOK files and converting frontmatter automatically. Phase 2 spends three to four weeks producing and reviewing one pilot Part. Phase 3 is the longest: four to six months of prose for the remaining eleven Parts. Phase 4 is one month of consistency verification, Phase 5 one month of appendix work, and Phase 6 two months of print and web conversion.

The total is nine to fourteen months. Manual work alone takes eighteen to twenty-four months; an AI work partner can reduce it to nearly half.

Each entry is tracked through six marks: ☐ unwritten, ◐ metadata only, ◑ first prose draft, ◕ prose complete, ● complete with cross-references, and ✓ proofread and approved. A single _Progress.md updates all 1,500 entries weekly, and Phase gates are judged from these marks.

This chapter covers 0–3; the next covers 4–6.

Phase 0 — A Month for the Agreement

Phase 0 receives two to four weeks because the project's decisions have dependencies and unequal weight. Some cannot begin until earlier decisions are complete.

First comes input verification: confirm the 1,500 Vault entries, thirteen-column frontmatter, nine-category and world-axis arrangement, and working Obsidian graph view. If inputs fall short, Phase 0 returns to Librarying. Missing frontmatter goes back to second-pass integration; unresolved categories or world axes go back to the first cycle. Only clean inputs can support Part structure and a tone guide.

The two heaviest decisions are Part structure and the tone guide. The structure applies Chapter 4's procedure to choose ten, twelve, or fourteen Parts, following five checks: IP scale, distribution across world axes, genre actors, dimensional layering, and short-story assets. The tone guide maps Chapter 7's seven voices to Parts and adjusts their ratios to the IP. It is kept as a separate two- to three-page Markdown file and consulted throughout. Together these decisions define the book's identity and the acceptance criteria for Phase 2.

Next come three policy decisions. The SSOT hierarchy resolves conflicts; the standard order is character sheet, master-approved story, LOREBOOK, Vault prose, and rulebook, though a rulebook may rise higher or a foreign-language original may rank first. The short-story separation policy chooses among a recommended companion volume, integration into Part 11, or quotation boxes, following Chapter 9. The LOREBOOK type chooses one of Chapter 2's five variants, A–E. Most IPs use standard encyclopedic Type A; deviations are stated in the agreement. These three define the book's policy.

The last group contains two practical decisions. Publishing specifications establish the broad picture—standard B5 format, about 450 pages, illustration budget, binding, cover finishing, and deluxe options—to be refined during publishing. Schedule and budget agreement assigns time within the nine- to fourteen-month cycle. Even when an IP changes the proportions, the shared document must preserve the overall plan for later review.

The eight decisions form a ten- to twenty-page Markdown document called the Bible. Every action in Phases 1–6 follows it, and problems return to it. If pilot review exposes a tonal mismatch, revise its tone guide; if Phase 3 reveals strain in the Part structure, revise that decision there.

The master is the final decision-maker for the LOREBOOK, one person per IP. The master owns the Part structure, acceptance decisions, publishing specifications, and every GO/NO-GO gate. In a solo project, the creator is also the master. The role resembles a TV showrunner or game-studio director.

Phase 1 — Building the Empty Skeleton

Phase 1 is almost entirely automated. An empty LOREBOOK skeleton is built in one to two weeks; people mainly review it.

The seven-step flow is mostly mechanical. Create the root Obsidian Vault [Project LOREBOOK]/, then summarize the Bible's key decisions in _Spec.md. Create twelve Part directories, from 00 Preface through 12 Appendices, each with an empty README.md. Create chapter directories and their empty README.md files.

The core automation creates 1,500 empty entry files. Frontmatter is converted from Librarying's thirteen columns while bodies remain empty. Given the thirteen-column CSV, an AI work partner can generate all files in about five minutes. Finally, create the root README.md and initialize _Progress.md with 1,500 ☐ marks.

Human review samples fifty files at random for about thirty minutes to verify the frontmatter conversion.

The deliverable is an empty LOREBOOK skeleton. Obsidian's sidebar lists 1,500 Markdown files, and graph view shows wikilinked nodes whose interiors remain empty. Phases 2–3 fill those bodies one by one.

Phase 2 — One Part Determines the Book's Fate

Phase 2 is the project's largest branching point. One pilot Part is written and reviewed by the master. A GO enters Phase 3; a NO-GO updates the Bible and rewrites the pilot.

If defects survive into Phase 3, they spread across all twelve Parts and may not appear until Phase 4 consistency verification, when recovery is expensive. A sound pilot lets all later Parts inherit the same tone, SSOT citation pattern, and avoidance of absolute prohibitions. One Part determines the entire format.

The pilot is usually Part 2, World Laws, or Part 4, Geography—whichever expresses the IP most clearly. For the fictional IP “○○,” Geography best reveals its mixture of shamanism, food, and daily life through cities and neighborhoods. For an SF or fantasy IP defined by magic mechanics, World Laws is better. Large settings such as D&D's Forgotten Realms likewise may pilot a school-of-magic section because the mechanic defines the IP.

The workflow has six steps: select the pilot; draft its Part and chapter READMEs; turn three to five highly connected hub entries into prose; verify tone, structure, and SSOT citation patterns against Chapters 7–9 and reflect the result in the tone guide; run Part-level checks for dead links and the five absolute prohibitions; and let the master decide GO or NO-GO.

The six steps take about three to four weeks, with hub prose and master review consuming the most time.

Five checks decide acceptance: the Part README works as a guide; every chapter README exists; three to five hub entries reach at least ◕; tone, structure, and SSOT citations follow Chapters 7–9; and the five absolute prohibitions occur zero times. All five ✓ marks yield GO; any shortfall yields NO-GO.

For a NO-GO, identify whether the cause lies in the Bible or the execution. Revise the Bible and redraft in the first case; partially or fully rewrite in the second. The decision cannot be treated lightly because it controls the rest of the project.

Writing all twelve Parts simultaneously without a pilot is possible but not recommended: one defect can spread everywhere before Phase 4 exposes it.

A film pilot episode serves the same role. One episode establishes a season's tone, characters, and direction. This is why the Lost pilot cost about $14 million—nearly five times an average episode—and why the Game of Thrones pilot was reshot. One episode can determine six to eight seasons. A LOREBOOK pilot likewise tests the IP's tone, structure, and consistency before the remaining eleven Parts inherit them.

Phase 3 — Four to Six Months of Full-Scale Writing

Phase 3 writes the eleven Parts outside the pilot and takes four to six months, about two to three weeks per Part.

The order stabilizes the book's identity rather than proceeding randomly. First write Parts 1, 2, and 3—world tone, world laws, and history/chronology. Once tone, laws, and history are fixed, other Parts gain direction. Next write Parts 5, 6, and 4—society, daily life, and geography—the largest, most stable everyday framework. Then write Parts 7, 9, and 10—actors and catalogs—after character and faction SSOTs have stabilized; their lighter card entries are easier to recover if earlier material changes. Part 8, the dimensions and spaces catalog, comes last because it is the most challenging and must absorb consistency from all other Parts.

The sequence is 1 → 2 → 3 → 5 → 6 → 4 → 7 → 9 → 10 → 8. It may vary by IP, but the governing movement is identity → everyday framework → cards → integration.

Each Part repeats the pilot flow: draft Part and chapter READMEs; prose begins with hubs; write standard and card entries; run Part-level consistency checks; update the Part README to reflect actual lengths and states. Repeat five steps eleven times.

Illustration commissioning proceeds in parallel. Begin a Part's commissions only after its prose reaches ◕. Ordering art before the text is stable creates expensive recommissioning when prose changes. Thus the standard is commission only after body v_final.

Update _Progress.md weekly. A delayed Part becomes visible immediately, allowing small weekly checks to prevent large drift over four to six months.

A typical five-month schedule looks like this. At the end of week 1, five hubs in Part 1 reach ◑. Week 2 adds five standard entries. In week 3, card entries reach ◕ and Part 1 passes; week 4 begins Part 2 while seven to ten Part 1 illustrations are outsourced. A tonal mismatch found in Part 2 triggers rewriting immediately. Part 2 passes by week 8, Part 3 by week 12, Part 5 by week 16, Part 6 by week 20, and Part 4's integrated overview arrives in week 23. About one fixed hour of review each week keeps the long project on course.

The deliverable is a LOREBOOK body in which all Parts have reached ●–✓. All 1,500 entries have prose and wikilinks, and illustration commissioning is at least 80 percent complete. Obsidian graph view contains 1,500 nodes and 5,000–10,000 links. The LOREBOOK is not finished yet: Phases 4–6 still perform consistency verification, appendices, and web/print packaging.

Entry Gates for Every Phase

Explicit acceptance criteria for every gate give the work its structure.

To move from Phase 0 to 1, input verification and the Bible must be complete, and the master inspects the gate. From 1 to 2, automated structural checks verify the directory skeleton, Spec, root README, and _Progress.md with every entry at ☐ or ◐. The 2-to-3 gate is strictest: the pilot must reach ◕–●, the tone guide must be updated, and the master must issue GO. The 3-to-4 gate requires every entry in every Part to reach at least ◕ and illustration commissioning to have begun; automated checks and master review operate together.

Failure returns the project to the same Phase. Temporary bypasses are not recommended: carrying a shortfall forward spreads it throughout the next stage and sharply increases recovery cost.

Phases 0–3 for the Fictional IP “○○”

Phase 0 took about three weeks. Input review confirmed 1,200 Vault entries. The team chose twelve Parts because sixty characters justified separate Parts 9 and 10; mapped seven voices while increasing catalog and field voices for shamanic material; established the SSOT order character sheet → story → LOREBOOK → Vault → rulebook; separated thirty stories into a companion volume; chose standard encyclopedic Type A; specified a roughly 450-page B5 hardcover with a KRW 15 million illustration budget; and planned eleven months for Phases 0–6. The deliverable was an eighteen-page Bible.

Phase 1 took about 1.5 weeks. It created [○○_LOREBOOK]/, _Spec.md, twelve Part directories, three to six chapters per Part, 1,200 empty entry files with converted frontmatter, the root README, and _Progress.md with every entry at ◐. The deliverable was the empty skeleton.

Phase 2 took about four weeks. Part 4, Geography, became the pilot because the IP's mix of shamanism, food, and daily life was clearest in its cities and neighborhoods. After the Part README and five chapter READMEs, four hubs—overview of the six-style district, Meryton, Haewon restaurant quarter, and old mudang street—became prose. Tone, structure, and SSOT checks found zero dead links and zero absolute prohibitions. The master found a tonal mismatch in the fourth entry; a partial rewrite earned GO. The deliverable was Part 4 at ◕–●.

Phase 3 ran for about five months in the order Part 1 (three weeks) → 2 (four) → 3 (four) → 5 (three) → 6 (three) → 7 (three) → 9 (four) → 10 (three) → 8 (three). Every Part followed README → hubs → standard entries → cards → consistency check → README update. Illustration work began once Part 1 reached ◕ and was 90 percent complete after five months. _Progress.md was updated every Friday. The deliverable was every Part at ●–✓.

At this point the LOREBOOK body is fully written. Consistency verification, appendices, and conversion into print and public web editions come next.

The copyright in this work belongs to Kim Dongeun WhtDrgon. and MEJE Works Corp. No part of this work may be reproduced, distributed, transmitted, displayed, performed, broadcast, translated, adapted, or otherwise used by any means without prior written permission from the copyright holders.