How the bible becomes a game. The phases are 04-architecture.md §14’s
(Phase 0, Phase 1, then A, B, C), with content and art added. Launch
scope is Acts I–II, cap 60, with Acts III–V as later updates (D1 in 07-decisions.md). Every
milestone ends the way milestones do now: a bot or a gate stage proves
it, and Foster plays it.
docs/platform.md).docs/platform.md, lessons from MapleStory).08-scale.md §9.05-art-audio.md,
production pipeline).03-systems.md §22). What bots
cannot show (fun, retention, how strangers read the game) is the risk
this choice accepts.Goal: no open question that could invalidate the architecture, the art pipeline or the core loop.
| Track | Work | Weeks | Done when |
|---|---|---|---|
| Bible | Bible reviewed; the remaining Phase 0/1 decisions in 07-decisions.md (D4–D11)
made |
1–3 | No open decision blocks Phase 1 |
| Engineering spikes | S1 (done 2026-10-03, reopened after S1b: monster AI
as data run natively, sandbox for hooks), S1b
(GDExtension shared Body.move, ≤ 2 µs per body-tick on the
VPS, web dlink template), S3 (footholds), S2 at 300 bots after S1b |
1–14 | Each pass line in 04-architecture.md §13 met;
S1b builds the GDExtension shared core (D12), with the §5.3 ladder’s
next rungs only if it misses |
| Art S0 | Spikes A–D in 05-art-audio.md (A passed
2026-10-03: Klein 4B edit is the gear generator; on the open A-pose
mannequin with socket discs and a pelvis part, joint tears fell to 1/72
and hats, tops and capes were 10/10 usable; next a side or 3/4 view
mannequin, a tunic skirt chain and a 1x readability check), style model,
one map kit with strata kit |
1–6 | Each spike’s pass line met |
| Art S1 | Pipeline’s first deliverable: Mossway kit, 4 base-class mannequins with 3 gear sets, Puddling, Shellmite, mote VFX, lantern gauge HUD | 6–16 | Passes art-direction review and automated QA |
| Audio | Lantern Town theme, one region theme with adaptive stems, 30 core SFX | 6–12 | Plays in the build; license confirmed (05-art-audio.md sources) |
| Feel | Movement and combat numbers from 03-systems.md on the current
sim |
4–10 | Feel bot passes; Foster signs off |
| Grind loop (pulled forward 2026-10-03) | Base class skills on hotkeys with auto-target (03-systems.md §4.3, §4.7); dense
respawning spawns on Mossy Field; XP and levels 1–10 so the HUD bars are
real; a party stub (shared XP and kill credit for players near each
other); trip boons with a Warden and a lantern pool (03-systems.md §7.9) |
1–6 | A bot levels from 1 to 10 through Mossy Field, Footprint Ponds and
the Old Tollbridge in the 03-systems.md §5.1 time (about
36 min of pure grind); two players party and share XP online; Foster
says it plays like a Maple-like, not Terraria |
| State | Work |
|---|---|
| Merged | Levels 1–60 with XP and the HUD bars (#49); a dense Mossy Field
(#50); Warden skills on hotkeys with auto-target (#51); Lantern Town
rebuilt as the Rim with walkable buildings (#53, #54); the party stub
(#56); monster aggro, leash and motes homing to the killer’s lantern
(#57); loot and scaling rules (#61, 03-systems.md §5.6, §13); the
grind bot (#58, #74); trip boons (#60); brass, personal drops and the
Mosswife’s tonic shop (#64); player stats and the Hornsmith’s swords
(#71); Footprint Ponds (4–7) and the Old Tollbridge (8–10); a gentler
first field (#98); knockback prediction (#97) |
| Open | Foster’s play-through of levels 1–10 as each class; S2 (300 bots) and S3 (footholds) |
| Found by the bot | Mossy Field alone took about 81–101 min to level 10, not 36: its
monsters top out at level 4. Moving field to field by level (Mossy
Field, Footprint Ponds from 4, the Old Tollbridge from 8) and buying the
Hornsmith’s swords, the bot reaches level 10 in 35.0–40.6 min on 4
seeds, inside §5.1’s 36.2 ±15%; the nightly reports it, gated again at
Phase A (03-systems.md
§22.1). The bot exit is met |
Then Foster plays it. Phase 0’s grind track ends when he says it plays like a Maple-like. Bot times are reported, not gated, until Phase A (2026-10-04, #102): the first all-class run had the Warden at 36–41 min and the Ranger, Kindler and Shade at about 52, 43 and 47–83, mostly the new bots’ play.
S1 took one day; S1b, started 2026-10-03, is the long pole. If the
shared core misses ≤ 2 µs, the region split (04-architecture.md §5.3)
comes next, before Phase 1 builds on a tick budget it cannot meet.
The first session of the game, at launch quality, online.
Build order (2026-10-04), each one playable before the next starts:
1. Accounts and real parties. Done. Nakama login,
character select, worldd v1 saves with item uids and the ledger, and
real parties with invites and the party finder (#75, #76, #78, #79, #83;
03-systems.md §5.5 As built).
2. Class choice. Done. Ranger, Kindler and Shade beside
the Warden, chosen at creation in the Lodge yard: New character, try
each class at level 8 on the dummies, choose, name, Begin (#88, #91,
#93–#96; 10-operations.md
§4). Skills not built yet show as “coming soon”; the 1–20 kits are
specced in 03-systems.md
§4.9. 3. The rest of the Mossway, then the Glasslands.
Dewline Hedgerows, Shellshed Slopes, the Soak and the Puddle Mayor, then
the Glasslands to the Cinder Eye, from the build sheets in 02-atlas.md §6 (#90). 4.
The prologue and the first beacon. The instanced
Pilgrim’s Stair and the Lighting. Beat sheet, instance, beacon and tech
needs: 11-prologue.md. As
built: the Stair, the Wick, B1–B7 and the Lighting, shared per channel
rather than instanced; the Low Lamp and the Sump wait (11-prologue.md §8). 5.
Appearance at creation (added 2026-10-04). 03-systems.md §1’s Create step
also picks body, face, hair and lantern glass color. Glass color first
(a tint), then hair and face as the cutout rig’s parts come out of Art
S1. As built: the glass color (03-systems.md §1, D67); body,
face and hair are still to come. 6. Quests and audio
(added 2026-10-04). The Mossway and Glasslands quest lines (02-atlas.md §6), and the Lantern
Town theme, a Mossway theme and the core SFX. 7.
Phones. The touch layout on the web build.
| State | Work |
|---|---|
| Merged | Lantern glass color at creation (#106, D67); the Mossway fields 5, 6, 7, 9 and 11 with kinds 10–14 (#108); Ranger, Kindler and Shade grind bots (#107); the Drain and the Puddle Mayor with Runlets (#109, D60–D62); a gentler Mossy Field and Ponds for the newcomer (#105); the prologue’s Stair, Wick and Lighting (#110); the explorer bot levelling again (#112) |
| Week 2 | The Sump and the Low Lamp (map 10, kinds 17–18); quests; music and core SFX; bots that beat the Mayor and keep the newcomer on its path |
| Found by the bots | The newcomer reaches level 5 in 15.5–21 min with the prologue (1.5x is 15.5), so its gate stays off; it detours through the Hedgerows portal on Mossy Field, and most of its deaths are at the Thornpig point near x 1070 on the Ponds. The explorer dies twice on the Old Tollbridge, where #108 put level-10 Tollkeepers on a level 8–10 bridge it crosses at level 1. The Mayor’s bots leave 26,274 of 55,000 in 330 s. The Ranger (about 43 min) and Shade (47–63 min) are over the 36.2 ±15% window |
| Open calls | D56–D58 (Mossway), D60–D62 and the Mayor’s 55,000 life, the Hedgerows portal on Mossy Field, the Thornpig point, Tollkeepers on the bridge, Ranger and Shade tuning (after Foster plays), D31–D34, the Glasslands palette |
| Week | Work |
|---|---|
| 1 (from 10-05) | Foster plays 1–10 as each class (Phase 0 sign-off); the rest of the Mossway (maps 5–19, its new creatures, the Puddle Mayor); the prologue and the Low Lamp; lantern glass color at creation; class bots that kite, Blink and open from stealth, and a first-field pass for the newcomer |
| 2 (from 10-12) | The Glasslands to the Cinder Eye; the first Mossway and Glasslands quests; the Lantern Town and Mossway themes and core SFX; hair and face at creation if Art S1 has the parts; the phone touch layout, or week 3 |
Alongside both: the full-colour Lampghost art, Art S1 (Puddling, mote VFX, the lantern gauge), S2 at 300 bots and S3. Foster plays at the end of each week; open calls are D31–D34 (prologue), D40–D45 (zones) and the Glasslands palette.
Architecture Phase A plus the launch content: level 60, the Prologue and Acts I–II. Private (D28): Foster and the bot fleet only.
03-systems.md targets to 60; the
AI critic passes each new region’s first hour.Still private (D28). Whether the public reveal is an open beta or the full release is D29. Architecture Phase B plus the systems an MMO needs at launch. No new regions: Phase B finishes Acts I–II to the density bar.
03-systems.md bands for 4 weeks;
levels 1–60 within 15% of the time-to-cap target; Act II’s server event
run end to end on a test world with bots, then reset.Architecture Phase C. Store releases on iOS, Android and desktop;
in-app purchases; 1–5k CCU; 99.5% uptime over the first month; restore
drill green. The Prologue and Act I open at launch and Act II opens
server-wide 4–6 weeks later (D2, 00-canon.md). Each Act’s world
event happens once per world.
A: Acts I–II, cap 60: regions 1–5, 3 holds (Lantern Town, Brazenbrow, the Almanac), 2 dungeons (the Gutterworks, Hornheart), 5 world bosses, about 125 quests, curated by Foster with agents. Phase 0 to launch is about 14–16 months. Acts III–V ship after launch as major updates, each raising the cap.
Art and curation figures are from 05-art-audio.md (launch scope
roughly 55–65 of the full 106 curator-weeks). Quest throughput assumes
~25 quests per region, written and scripted by agents at about 8 quests
per week with Foster reviewing, plus a week of bot tests per region.
| Rejected | Why |
|---|---|
| B: the full game, cap 120 (2+ years) | Two years with no revenue and no players to tell us what works; first chosen, then reversed the same day |
| C: full game with 2 paid curators ($95k–160k, ~14–16 months) | Cost, and curators need the style locked first |
The risk of A is that engaged players reach 60 (about 32 hours, 3–4
weeks at 8–12 hours a week) well before Act III. The endgame at 60 has
to hold them: Deep-touched dungeons, the Deep, world bosses and the
Sunline (03-systems.md §18).
Act III is built during Phase B so it can ship about 3–4 months after
launch.
08-scale.md).docs/platform.md) opens only once
the game holds players. It opens with Halvesper’s kits as the starter
asset library, a published creator payout share (D24) and a creator
challenge with cash prizes.| When | Foster plays or reviews |
|---|---|
| Every merge | The live web client at halvesper.fdhproperty.com |
| Each spike | Its result and the decision it forces, in 04-architecture.md §13 or
05-art-audio.md |
| End of Phase 0 | Art S1 deliverable beside the key art; S1b/S2/S3 results; D4–D11 |
| End of Phase 1 | The slice, in a desktop browser and on a phone’s browser |
| Each region at the density bar | The region, played end to end |
| Each alpha/beta month | CCU, retention, time-to-level, economy, crashes (10-operations.md
metrics) |
Week counts assume agent-driven development with Foster curating and reviewing, as in M0–M6: about 16 + 10 weeks, then 6 and 4 months, so about 14–16 months from Phase 0 to launch, and roughly another year of Act updates to cap 120. They are estimates to plan against. S1b, the art pipeline and curation throughput are where they are most likely to move.