Adwa — A research-led doorway into Ethiopia, 1896

Research / presentation application. The Unreal game is in development, not a finished release. Native milestone results below are preserved dated project records, not a new playtest. Historical classifications remain explicit; unresolved research is not verified history.

Source: Docs/GAME_DESIGN_BIBLE.md

# ADWA Game Design Bible

**Status:** P6 horse and mounted-traversal prototype implemented; long-term design remains draft
**Version:** 0.6
**Last reviewed:** 2026-09-27

## North star

Create a doorway into Ethiopia in 1896 through place, people, logistics, uncertainty, and culturally specific sound and visual language. Historical canon, plausible reconstruction, game fiction, and unresolved research remain explicit and separate.

## Representation principle

ADWA will not erase difference, and it will not weaponize difference. Historically relevant communities must be specific without being ranked as superior or inferior. P1 therefore uses neutral placeholders and makes no identity claims. Future character identity, language, relationships, clothing, contributions, and historical memory require evidence and appropriate review.

## P1 — The First Footstep

P1 proves only the spatial and systems skeleton of the future vertical slice. The temporary player perspective is a route-wise courier with no final identity, name, origin, language, clothing, affiliation, or historical role.

Critical path:

1. Spawn in a non-canon proxy camp.
2. Receive a fictional placeholder task.
3. Interact with one placeholder supply object.
4. Traverse camp → trail → ridge.
5. Record Observation A and Observation B as `SEEN` or `UNCERTAIN`.
6. Avoid or escape an abstract, unidentified obstacle.
7. Return to camp.
8. Report only knowledge actually recorded.
9. Trigger `PROTOTYPE_MOBILIZATION_TIMING_ONLY` with no drum audio.
10. Reach a non-historical dawn endpoint and end the prototype.

## P1 interaction model

- Walk and run; crouch only if the observation proof needs it.
- Keyboard/mouse and Xbox-style controller.
- Context interaction for task, supply, observations, report, and endpoint.
- Minimal objective types: `Interact`, `ReachArea`, `Observe`, `ReturnReport`.
- Three deterministic checkpoints: camp departure, observation arrival, camp return.
- No combat, kills, named forces, historical figures, horse, crowd simulation, final dialogue, or cultural props.

## Knowledge rule

The report UI derives options from recorded observation state. It must never expose an unobserved claim. Placeholder observations have no historical referent:

- A: “Movement seen near ridge.”
- B: “Number uncertain.”

## P1 visual language

All terrain, structures, proxy people, supplies, danger, lighting, and UI text are developer-facing greybox content. The map and milestone use the identifier `NON_CANON_PROXY_TERRAIN`. No greybox element may be cited as a reconstruction of Adwa.

## Long-term realism backlog — not authorized for P1

Research and future phase gates must precede final highlands/Adwa terrain, ecology and agriculture, rural/camp life, animals, atmosphere, cooking and craft, fabrics and dress, ammunition systems, weapons, horses/tack, shelter, food/water/medicine, spiritual life, historical languages/voices, Shelela, Negarit, communications, and regional population representation. This backlog preserves ambition without expanding P1.

## P1 implementation result

The complete critical path is playable in the neutral World Partition greybox. The real trigger/interaction functional traversal and six state/schema automation tests pass. Observation A records `SEEN`; Observation B records `UNCERTAIN`; the report cannot expose either while `UNSEEN`. Mobilization remains lighting/proxy timing only, with no drum or Negarit representation.

## P2 — Movement & Presence

P2 retains camera-relative movement and movement-oriented body facing. Walking is the precise exploration state; running increases speed while reducing turn precision, with no survival stamina system. Acceleration and braking are intentionally non-instant. The grounded spring-arm camera uses limited configurable smoothing and collision handling.

Interactions require deliberate range, facing, and active-objective eligibility. Movement/camera sensitivity, invert Y, hold/toggle run, controller dead zones, and camera smoothing are config-backed foundations. The current body and terrain remain technical proxies with no historical identity.

## P2.2 — Founder-grounded retuning

The first founder playtest rejected P2's movement/presence feel: walk, run, turning, slope traversal, and camera response were too fast; interaction feedback was unreliable; the test surface was too small; and falling through segmented proxy surfaces ended the session. P2.2 treats this as direct playtest evidence, not as a request for new traversal systems.

The design target is deliberate exploration with materially slower configured movement, restrained camera response, forgiving current-objective interaction, a wider connected technical test loop, and automatic checkpoint recovery below the world safety threshold. Jump, stamina, climbing, vaulting, final geography, historical assets, and P3 remain outside scope.

## P3 — Mission prototype

P3 proves a complete non-canon mission loop on the existing technical route. It is `GAME_FICTION`, not a claim that this route, task, observation, danger, or report occurred historically.

The mission graph supports travel, interaction, observation, return, report, and completion nodes with explicit prerequisites. Observation B is optional and can be reached before A; Observation A remains required. A player may return with B missing and report “I did not observe this.” The danger pocket remains an unidentified technical proxy with no combat, enemy, weapon, faction, or historical identity.

World truth, player observation, player certainty, and reported information are distinct states. Valid observation conditions are required before knowledge is recorded. Certainty is limited to `CONFIRMED`, `UNCERTAIN`, and `NOT_OBSERVED`. Report content is generated from player knowledge and cannot affirm an unobserved fact. No morality, reputation, or RPG scoring is attached.

Mission completion is an explicit graph state reached only after all required nodes, a knowledge-consistent report, and the neutral prototype endpoint. Checkpoints retain objective, observation, certainty, report, route, and mission state across recovery. The temporary founder UI exposes only current/optional objectives, captured observation results, report availability, and completion; internal truth and branch reasoning remain developer-only.

FOUNDER EXPERIENCE REVIEW: DEFERRED UNTIL REPRESENTATIVE GAMEPLAY BUILD

## P4 — Combat prototype

P4 establishes dangerous, deliberate, terrain-aware combat without presenting final Battle of Adwa content. The single-shot ranged prototype, generic melee, teams, damage, AI, arena, ammunition, timings, and behavior are `GAME_FICTION` technical values.

The player must aim before firing, moves more slowly while aiming, reloads after the configured single shot, and respects recovery windows. Solid terrain blocks sight and attacks. Melee has range and recovery rather than unlimited button spam. Serious hits incapacitate quickly; failure restores mission and combat checkpoint state without restarting play.

AI knowledge is bounded by sight, recent location, and combat events. A target may be known, suspected, or unknown; occlusion prevents initial detection, lost targets degrade over time, and allies are not valid targets. Near misses create temporary technical pressure without claiming historical battlefield psychology.

Combat is an optional branch of the P3 graph. The observation/report route remains fully playable without combat, and combat cannot reveal mission knowledge. No final weapon, uniform, animation, sound, force identity, tactic, formation, casualty behavior, or geography is approved by this prototype.

## P5 — NPC camp and world life

P5 makes the technical camp visibly inhabited by generic `GAME_FICTION` people who have roles, destinations, routines, rest periods, brief interactions, local danger reactions, and resumable activity. They are not quest dispensers and do not stand for researched Ethiopian individuals, occupations, camp organization, social roles, or behavior.

The default camp uses fourteen ambient full actors across seven non-canon gameplay roles and six activity areas. Simple carrying, supply checking, maintenance, gathering, rest, lookout, walking, and conversation loops demonstrate reusable systems rather than historical practice. Temporary body/head forms, role colors, zone materials, and labels improve readability beyond cubes without introducing costumes or culturally specific assets.

NPCs perceive only explicit nearby danger or mission overrides. Workers and travelers seek safety; guards and lookouts pause and observe. All resume routine after the override clears. Ambient conversations use explicitly labeled placeholder acknowledgement text and cannot reveal hidden mission knowledge.

The representative-build gate is closer because movement, mission flow, combat, humanoid-shaped population, activity, and a readable camp now coexist. Terrain, atmosphere, skeletal animation, sound, environmental effects, historically researched presentation, and founder hands-on review remain outstanding.

FOUNDER EXPERIENCE REVIEW: DEFERRED UNTIL REPRESENTATIVE GAMEPLAY BUILD

## P6 — Mounted traversal prototype

The player can approach a clearly temporary horse proxy, mount with the interaction input, select walk/trot/canter with the run input, steer normally, and request a safe dismount. The mount has deliberate acceleration, braking, and broader high-speed turning. It respects terrain and solid obstacles and does not jump.

Mounted travel can advance travel objectives but never grants observation knowledge by proximity alone. Unsupported mounted attacks are rejected; combat resumes after dismount. Camp people provide restrained local acknowledgement while preserving crowd flow. Horse breed, tack, posture, technique, logistics, doctrine, terminology, animation, sound, and mounted combat are neither designed nor approved by P6.

Population, camp life, mission flow, combat, horse traversal, and basic readability now coexist. Founder review remains deferred until terrain, lighting, atmosphere, sound, animation, and researched presentation are sufficiently representative.

FOUNDER EXPERIENCE REVIEW: DEFERRED UNTIL REPRESENTATIVE GAMEPLAY BUILD

## P7 — Audio, atmosphere, and Negarit foundation

The representative ugly build now has quiet neutral procedural wind, spatial zone state, daypart hooks, movement and surface events, horse gait/alert events, camp-life presence events, combat/danger events, and bounded listener propagation. These systems are technical scaffolding and intentionally avoid final music, speech, instruments, regional fauna, material acoustics, or culturally specific performance.

Negarit has a dedicated silent event framework for future mission, world, cinematic, and NPC response use. `ነጋሪት ይጎሰማል` / `Negarit Yegosemal` describes the Negarit sounding, rolling, or resounding; martial, proclamation, authority, or mobilization meanings require documented context. It is not generic mission-complete feedback and is not represented by a stock war drum.

Founder review remains deferred. Mission flow, combat, camp life, mounted travel, environmental audio, zones, and daypart now coexist, but representative terrain, lighting, visible atmosphere, animation, and researched presentation are still needed.

FOUNDER EXPERIENCE REVIEW: DEFERRED UNTIL REPRESENTATIVE GAMEPLAY BUILD