> **RC4 Merchant Economy RELEASED — 2026-09-29**
>
> Production main: `1ff2907c4e69af8adc336792105894542576433b`.
> Accepted integration candidate: PR #194 head `582cc72668071e4c3e1f55fd43c143841a3bbaad`.
> Exact-main Verify #2160 and Pages #113 are SUCCESS.
> Public Merchant Economy is released: Producer → Merchant → Customer, Home Market, BuyOffer/Listing, real navigation, canonical Wallet/Rust item/Reservation/Trade/Ledger/Career, save/load and desktop/mobile browser proof.
> RC4 donor/acceptance PRs are historical/superseded by #194 and must not be merged into current main.
> Read `docs/RC4_RELEASE_CLOSEOUT.md` and `docs/RC4_RELEASE_SUCCESS_CONTRACT.md`. Earlier baseline/candidate notes below are historical where they conflict.
>
> **RC4 production candidate — 2026-09-29**
>
> PR #194 is the assembled Merchant Economy candidate on the RC3.2 production baseline.
> Read `docs/RC4_RELEASE_SUCCESS_CONTRACT.md` for authorities, user interaction, replay, corruption and exact-main/public gates.
> Implementation and old-head CI are not release evidence. Consult PR #194 and its exact current checkout, main Verify and Pages/public reports. Do not label this release complete until all mandatory exact-SHA gates succeed.
> Earlier dated baseline notes below are historical.

> **RC3.1 candidate source update — 2026-09-28**
>
> Starting released main: `c1fdbcac92f61499507ef89eec8b207da5bab7cb` (RC2, Pages #110). PR #173 adds physical Blueprint loot and personal recipe learning through existing authorities.
> Read `docs/RC3_1_BLUEPRINT_SUCCESS_CONTRACT.md` and `docs/RC3_1_BLUEPRINT_RELEASE.md`. Candidate implementation does not imply publication: exact-head Verify and exact-main Pages/native/public Blueprint + RC2 + SWA7 remain mandatory. Record actual results on #173.
> Donor #168 stays read-only/unmerged. Iron/Steel, new stations, market and upgrades are not part of RC3.1. Earlier baseline notes below are historical where they conflict.

# Autonomous Clone World — Master Game Plan

> **Repository:** Simclone  
> **Target:** Autonomous Society Simulation + Survival + Automation + Colony Sim + Emergent Narrative  
> **Core rule:** โลกต้องเดินต่อได้เมื่อผู้เล่นหยุดออกคำสั่ง

---

## 1. Vision

ผู้เล่นคือ **Original** คนแรกของโลก สร้าง Clone และเฝ้าดูสิ่งมีชีวิตที่เริ่มจากต้นแบบเดียวกันค่อย ๆ แตกต่างจากประสบการณ์ ความรู้ ความสัมพันธ์ สังคม และการตัดสินใจ จนเกิดเป็นอารยธรรมที่ดำเนินต่อได้เอง

จุดขายไม่ใช่จำนวน NPC แต่คือ **ความต่อเนื่องของเหตุและผลข้ามคนและข้าม Generation**

### Game Pillars

- **Autonomy** — Clone เลือกและลงมือเอง
- **Learning** — ประสบการณ์กลายเป็น Skill และ Knowledge
- **Emergence** — เรื่องราวเกิดจากระบบ
- **Legibility** — ผู้เล่นเห็นว่าใครทำอะไรและทำไม
- **Continuity** — ความรู้และประวัติศาสตร์ข้าม Generation
- **Scalability** — CPU ทำ known path; AI ใช้กับ novelty

---

## Independent Clone World — approved direction

Canonical design: [docs/INDEPENDENT_CLONE_WORLD.md](docs/INDEPENDENT_CLONE_WORLD.md)

After the UX V1 interaction line, the next gameplay architecture moves from settlement-first to individual-first simulation:

```text
Individual → personal survival → personal home → household → neighborhood → emergent settlement
```

A central village is not a prerequisite in the target architecture. The migration is gated and incremental: existing deterministic lifecycle, Rust possession/build authorities, modular construction, knowledge provenance and save compatibility are reused instead of rewritten. IC1 introduces only evidence-derived home ownership and personal home-site planning; birth, Camp, global stock and save schema remain unchanged until later approved slices.

---

## 2. Core Game Loop

```text
PLAY
→ OBSERVE
→ TEACH / INFLUENCE
→ CLONE
→ LEARN
→ AUTOMATE
→ SPECIALIZE
→ FORM SOCIETY
→ EVOLVE
→ HISTORY
→ CONTINUE
```

### Progression

**Early**
- เก็บ Food / Wood / Stone
- สร้าง Shelter
- สร้าง Clone
- ดู Needs และ Skill แรก

**Mid**
- อาชีพ
- Mentor / Student
- Relationship
- Knowledge transfer
- Faction

**Late**
- หลาย Generation
- เมือง
- Economy
- Conflict
- Culture
- Technology

**Endless**
- โลกสร้างประวัติศาสตร์ต่อเอง
- ผู้เล่นกลับมาแล้วพบการเปลี่ยนแปลงที่ไม่ได้ script ไว้

---

# 3. Development Roadmap

## V0.1 — Simulation Skeleton

### Goal
ทำให้มีโลกที่ tick ได้อย่าง deterministic

### Systems
- Tick scheduler
- World state
- Agent entity
- Resource node
- Basic renderer
- Pause / Speed / Reset
- Seed

### Definition of Done
- รัน 10,000 ticks โดย state ไม่เสีย
- Reset ด้วย seed เดิมได้ผลเหมือนเดิม
- UI ไม่มี simulation rule

---

## V0.2 — Survival Loop

### Systems
- Food / Wood / Stone
- Hunger
- Energy
- Health
- Forage
- Rest
- Inventory
- Shelter

### Rules
- Hunger ต่ำ → EAT / FORAGE priority สูง
- Energy ต่ำ → REST
- starvation ลด Health
- Shelter มีผลต่อ population growth

### Definition of Done
- Agent หาอาหารและพักได้โดยไม่ต้องสั่งทุก tick
- Resource ไม่ติดลบ
- ไม่มี infinite resource loop

---

## V0.3 — Clone + Lifecycle

Current knowledge contract: [V0.5.0](docs/KNOWLEDGE_MEMORY_0.5.0.md). Current next steps: [NEXT_STEPS](docs/NEXT_STEPS.md). Lifecycle/history and skill provenance remain retained foundations; earlier release descriptions below are historical.

Concrete lifecycle/save contract: [docs/LIFECYCLE_0.3.0.md](docs/LIFECYCLE_0.3.0.md)

Current clock contract: **360 ticks = 1 simulated day = 1 biological year**; CHILD 0–15, ADULT 16–54, ELDER 55+, DEAD overrides age. V0.3.1 stage effects, V0.3.2 autonomous birth, V0.3.3 deterministic lifespan/age death and V0.3.4 generation continuity are retained. V0.3.5 adds evidence-based stable death history; retained-history limits and bounded performance remain follow-up gates before skill provenance.

### Agent Identity
- id
- name
- generation
- parent
- appearance

### Life
- age
- child / adult / elder
- hp
- hunger
- energy
- death

### Rules
- Child → Adult → Elder → Death
- Elder ผลิตงานลดลง แต่มีค่าด้าน teaching / experience
- Character appearance เป็น identity ถาวร
- Role icon แยกจาก identity

### Definition of Done
- เกิด Generation ใหม่จากระบบ
- parent link ถูกต้อง
- aging/death reproduce ได้ด้วย seed

---

## V0.4 — Skill System

### Initial Skill Families
- FORAGE
- WOODCUT
- MINE
- BUILD
- EXPLORE
- SOCIAL

ภายหลัง:
- COMBAT
- RESEARCH

### Skill Model
- XP
- Level
- V1 → V2 → V3
- Source
- Preconditions
- Effects

### Requirement
Skill level ต้องมีผลต่อ gameplay จริง เช่น:
- เวลา
- output
- failure rate
- resource efficiency
- risk

---

## V0.5 — Memory + Cultural Knowledge

Current implementation starts with personal evidence-backed resource knowledge and explicit person-to-person transfer. Cultural Archive remains a later V0.5 slice; personal knowledge must not read hidden global world truth.

### Memory Types
- Episodic Memory
- Social Memory
- Knowledge provenance
- Belief

### Knowledge Transfer
- Parent inheritance
- Mentor teaching
- Faction knowledge
- Cultural Archive
- Personal discovery

### Critical Rule
คนตายได้ แต่ความรู้ที่บันทึกใน Cultural Archive แล้วต้องไม่หายตามคน

---

## V0.6.4 — Stabilized Baseline

ล็อก baseline ของ:
- population
- lifecycle
- resources
- buildings
- skills
- inheritance
- archive
- save/load
- pause/speed/reset
- inspector

### Regression Rule
Feature ใหม่เพิ่มเป็น module  
ห้าม rewrite ของเดิมโดยไม่จำเป็น

---

# 4. V0.7 — Decision Brain

เปลี่ยนแกนการตัดสินใจจาก random เป็น **Candidate Scoring ที่อธิบายได้**

```text
World State
→ Perception
→ Needs
→ Goals
→ Memory
→ Candidate Actions
→ Score
→ Validate
→ Execute
→ Outcome
→ Learn
```

## Score Factors

- Need
- Goal
- Skill
- Social
- Expected Value
- Distance
- Risk
- Cost

ตัวอย่าง:

```text
Score(action)
= Need
+ Goal
+ Skill
+ Social
+ Expected Value
- Distance
- Risk
- Cost
```

## Hard Validation
ถ้า precondition ไม่ผ่าน action ใช้ไม่ได้ ไม่ว่าคะแนนสูงแค่ไหน

## Decision Trace
ทุกการตัดสินใจสำคัญต้องบันทึก:
- candidates
- score breakdown
- selected action
- rejected actions
- validation result

---

# 5. V0.8 — Perception + Local Knowledge

Agent ต้องไม่รู้ทุกอย่างในโลก

### Sources of Knowledge
- เห็นเอง
- ได้ยิน
- ถูกบอก
- เรียนจาก mentor
- faction
- archive

### Requirements
- แต่ละ Agent มี knowledge map ต่างกัน
- ข่าวสารเดินทางผ่าน social interaction
- belief อาจไม่ตรงกับ truth
- provenance ต้องรู้ว่า “รู้จากไหน”

---

# 6. V0.9 — Spatial World + Pathfinding

### World Objects
- Agent
- Resource
- Shelter
- Building
- Enemy
- Event

ทุกอย่างมีตำแหน่งจริง

### Action Contract
```text
Choose target
→ Plan path
→ Move
→ Reach target
→ Validate again
→ Execute
```

### Distance
ระยะทางต้องมีผลต่อ decision cost

---

# 7. V1.0 — 100-Day Autonomous Proof

นี่คือ milestone สำคัญที่สุดก่อนขยายระบบสังคมใหญ่

## Test Setup
- Original + Clone จำนวนน้อย
- ปล่อย simulation 100 วัน
- ห้ามผู้เล่นแก้ปัญหาด้วย micromanagement

## PASS Conditions
- population ไม่ล่มจาก bug
- เกิดอย่างน้อย 2 Generation
- Agent มี Skill / Memory / อาชีพแตกต่างกัน
- knowledge ส่งต่อข้ามรุ่นจริง
- มี social relationship
- เกิด emergent group
- วิกฤตทำให้ priority เปลี่ยนอย่างมีเหตุผล
- Inspector อธิบาย action ได้
- Chronicle สรุป 100 วันได้
- seed + event log replay bug ได้

ถ้า V1.0 ยังไม่ผ่าน **ห้ามแก้ด้วยการเพิ่มระบบใหญ่ขึ้น**

---

# 8. Social Brain — V1.1

### Relationship State
- Trust
- Affinity
- Respect
- Fear
- Debt

### Relationship Types
- Friend
- Rival
- Family
- Mentor
- Student

### Rule
ความสัมพันธ์เปลี่ยนจากเหตุการณ์จริง และต้องมีผลต่อ:
- teaching
- trade
- faction
- conflict
- leadership

Leader ควรเกิดจาก skill + respect + support ไม่ใช่ random ล้วน

---

# 9. Faction — V1.2

- สร้าง / ยุบ / แตก / รวม
- Leader succession
- member join / leave / exile
- faction goals
- rules
- resources
- knowledge
- settlement
- territory
- alliance
- trade
- rivalry
- conflict

---

# 10. Economy — V1.3

### Layer 1
Shared Food / Wood / Stone

### Layer 2
Specialization
- Forager
- Builder
- Miner
- Explorer
- Researcher
- Guard

### Layer 3
Production
- tools
- buildings
- processing

### Layer 4
Exchange
- barter
- faction trade

### Advanced
scarcity + demand + logistics → value

---

# 11. Conflict — V1.4

เริ่มจาก:
- environmental threat
- animal threat

ก่อน:
- war

### Conflict Causes
- scarcity
- territory
- relationship
- faction goal

### Behaviors
- defend
- guard
- retreat
- negotiate
- alliance

ผล combat ต้องเขียนลง:
- Memory
- Relationship
- Chronicle

---

# 12. Culture + Discovery — V1.5–V1.6

```text
Experience
→ Hypothesis
→ Experiment
→ Outcome
→ Verification
→ Cultural Knowledge
```

### Properties
- ความรู้ผิดสามารถแพร่ได้
- ความรู้ผิดสามารถถูกพิสูจน์ผิด
- เมืองต่างกันมี technology ต่างกัน
- วัฒนธรรมพบกัน → knowledge transfer
- รุ่นหลังสร้างสิ่งที่ Original ไม่เคยรู้

---

# 13. Evolution — V1.7

เป้าหมายไม่ใช่แค่ stat mutation แต่เป็นการเปลี่ยนแปลงระยะยาวจาก:
- culture
- environment
- specialization
- technology
- social pressure
- historical accidents

---

# 14. AI Brain — V1.8

LLM ไม่ควรอยู่ในทุก tick

ใช้สำหรับ:
- novelty
- ambiguity
- reflection
- new hypothesis
- proposing new behavior/skill
- summarizing major history

ต้องมี:
- budget
- cooldown
- trigger condition
- deterministic fallback
- verification

---

# 15. AI CPU / Reasoning Compilation — V1.9

## System 0
Deterministic execution
- movement
- inventory
- skills
- state transitions

## System 1
CPU decision
- scoring
- planning
- search
- candidate selection

## System 2
LLM
- novelty
- ambiguity
- reflection

## Learning Flow

```text
LLM Proposal
→ Validate Preconditions
→ Execute in controlled path
→ Verify Outcome
→ Store Evidence
→ Promote to Knowledge
→ Compile into deterministic Skill/Rule
```

### Core Principle

> อย่า generate ถ้าเลือกได้  
> อย่า reasoning ถ้าคำนวณได้  
> สิ่งที่เรียนรู้แล้วควรไหลจาก AI → verified knowledge → deterministic skill

---

# 16. UX/UI — World Observation UI

## Primary UX

```text
WORLD
→ EVENT / DECISION
→ CLONE / PLACE
→ WHY
→ INFLUENCE
```

The UI must make autonomous causality legible without turning the game into an admin dashboard.

## Main Screen

### Center
- World remains the dominant surface.
- Persistent HUD stays low-chrome: world time, Food/Wood/Stone, population/housing and one compact AI Autonomy entry.
- No large permanent dashboard may cover the central or lower-middle playfield.

### Primary navigation
- World
- People
- Systems
- Items / Craft
- Chronicle

Manual Clone creation is an Inspector influence action, not a primary-navigation verb.

### Context surfaces
- Selected Clone → Inspector / bottom sheet.
- AI Autonomy → live Decision Feed.
- Systems → seven primary gameplay-system cards.
- Full runtime catalog → Advanced disclosure only.
- Chronicle → event drill-down and later event-to-place / event-to-agent navigation.

---

## Clone Inspector

### Identity
- Portrait
- Name
- Generation
- Parent
- Role

### State
- Hunger
- Energy
- Health

### Mind
- Goal
- Thought
- Memory

### Capability
- Skills
- knowledge source
- personal inventory
- equipped hand item

### Social
- mentor/student now
- faction/relationships when authoritative

### Decision Trace
- candidates
- scores
- selected action
- why
- deterministic factor breakdown

The shortcut **Why?** remains first-class even if tabs are later consolidated.

---

## Visual Language — Icon / Image First

Player-facing information follows this disclosure order:

```text
ICON / PORTRAIT / WORLD IMAGE
→ SHORT NUMBER / STATUS
→ TAP / EXPAND
→ EXPLANATION / PROVENANCE
```

Default rules:
- Prefer icons, portraits, item/building pictograms, progress bars/rings, status colour and world-space markers over prose.
- Persistent HUD labels stay short; mobile may hide visual labels when an accessible `aria-label` remains.
- Chronicle uses event icon + related portrait + one-line event summary before any evidence text.
- Systems use large system icons + status badge + primary value; implementation detail stays collapsed.
- Inventory uses visual slots and item pictograms; item-instance IDs/details are secondary.
- Inspector tabs are icon-first. **Why?** remains a first-class visual shortcut.
- Long explanation, provenance and technical truth remain available on demand; visual-first never means deleting evidence.
- Images/icons are presentation only and must not imply authority that runtime does not have.
- Colour is never the only carrier of meaning; status also has shape/icon/text for accessibility.

Avoid:
- paragraph-first HUD cards
- repeated explanatory copy on every visible surface
- decorative imagery that invents gameplay state
- replacing `UNKNOWN` with a confident visual guess

## Menu Interior Contract

Menus are secondary evidence/detail surfaces, not dashboards that replace the world.

Default structure:
- visual hero / identity
- 3–6 primary metric tiles
- icon-first actions
- portrait/item cards for people and possessions
- progress bars for active work
- advanced/shadow/provenance content behind collapsed disclosure
- long explanatory prose behind `details`

Specific expectations:
- **Main Menu** → visual action cards and compact save status.
- **People** → portrait cards + mini Need bars; search/filter remains accessible.
- **Survival** → settlement overview only; it must not own actions that belong to a physical structure.
- **Items / Rust** → personal possessions, hand crafting and global production policy; station-specific work is not nested here.
- **Structure Context Menus** → tap the world object itself:
  - Camp owns Cultural Archive creation/automation/reading.
  - Crafting Table owns recipes requiring `CRAFTING_TABLE_LV1`.
  - Furnace owns Charcoal processing.
  - Modular House owns its derived completion/capacity/missing-piece/builder status.
- **Systems** → icon/status/value first; detailed descriptions collapsed.
- **Chronicle/Event/Decision Feed** → event/agent imagery first, evidence and deep trace on demand.
- **Inspector** → icon-first tabs on mobile with accessible labels.

Menu redesign must not remove an existing command path or hide an action required for an accepted gameplay workflow.

Structure menus are contextual surfaces opened from world-space hit targets. They must reuse existing command authorities and canonical evaluators; they do not create a separate Building Mode, station ledger, housing completion rule, or resource writer.

## Visual Analytics Contract

Interactive visualizations are projections of existing evidence, never a new data source.

- **Knowledge Graph** lives at the Camp/Cultural Archive and may connect retained authors → archive entries → the currently selected Clone only when those relationships exist in `culture.entries` and personal `knowledgeState`.
- Knowledge Graph filtering/selection is UI-local. Reading an archive entry still routes through the existing `READ_ARCHIVE` command.
- **Bar charts** may compare current stock to canonical reserve targets, current task counts to living population, or derived missing-piece composition. Values must come from current state/evaluators.
- **Trend/line graphs** require retained historical points. Chronicle may graph event counts by day because event ticks are retained; do not synthesize missing telemetry or invent historical resource curves.
- Graphs may support filter/select/focus interactions, but visual interaction must remain read-only unless it invokes an already-authoritative command.
- A graph must expose the same `LIVE / READY / SHADOW / INFRA` semantics as text surfaces where authority status matters.


---

## Diegetic World UI — Bubble First

The world itself is the primary information surface. Prefer in-world bubbles and markers before opening a menu.

Bubble semantics:
- **Thought cloud** → current selected/recent decision or an actual Need derived from state.
- **Speech bubble** → only a real communication event such as Knowledge sharing or Mentor communication. Never invent dialogue.
- **Work bubble** → current authoritative BUILD / CRAFT / PROCESS task.
- **World marker** → building progress, resource/place state, or another spatial fact owned by runtime.

Budgets:
- mobile: at most 3 simultaneous agent bubbles
- desktop/tablet: at most 5 simultaneous agent bubbles
- selected Clone gets priority
- communication and emergencies outrank ordinary recent-task bubbles

Menus/dialogs are secondary and should mainly provide:
- exact evidence / provenance
- advanced system detail
- inventory actions
- save/import/reset
- accessibility fallback and searchable history

Do not:
- turn a current decision into historical intent
- render speech when no communication event exists
- persist a second thought/dialogue history just for UI
- let bubbles write simulation state

Additional world-story cues:
- **Communication link** → a short-lived line may connect sender and receiver only when retained message/Mentor evidence identifies both agents.
- **Relationship link** → selecting a Clone may show bounded living Parent/Child and active Mentor links from existing `parentId` / mentorship records.
- **Task target marker** → selected/recent productive work may show a small pictogram at the authoritative `task.x/y` destination.
- **Resource pulse** → a resource node may pulse only when an exposed current task actually targets that node.
- **Dropped item marker** → physical dropped Rust items are rendered from the existing item ledger and its `location.x/y`.
- **Life burst** → recent retained Birth/Death events may create short visual bursts at the retained agent position.
- **Achievement burst** → recent Skill/Craft/Build events may create a bounded success burst.
- **Construction pulse** → incomplete modular houses may animate their already-derived progress marker.
- These cues are projections, not additional ownership or history.

There is currently **no general Emotion Authority**. Do not label a Clone sad, angry, afraid, happy, etc. unless a future authoritative system owns those states. Visual reactions may show factual Needs such as hunger, low energy or low HP, but must not reinterpret those Needs as emotion.

---

## World Feedback

World feedback is read-only presentation derived from authoritative state.

Show:
- selected Clone marker
- emergency Need marker
- persistent BUILD / CRAFT / PROCESS marker
- short-lived icon-first marker for a newly started task
- equipped tool when visually relevant
- incomplete modular-house progress

Transient task markers use simulation tick age, not wall-clock time, so presentation is reproducible for a given state.

Do not:
- write simulation state from a visual marker
- invent motives not present in the decision trace
- show every action permanently
- create a second building-progress or task ledger

---

## Thought Bubble

แสดงเฉพาะข้อมูลสำคัญ:
- newly selected action for a short bounded window
- productive BUILD / CRAFT / PROCESS
- Emergency / critical Need
- future Discovery / Social / Conflict events when their systems are authoritative

ห้าม spam ทุก action และให้ icon-first บน world view

---

## Systems View

Basic view exposes only the main gameplay layers:
- Lifecycle
- Housing
- Production / Craft
- Inventory / Equipment
- Knowledge / Mentor
- WorldSim Ecology
- Kingdom systems

Every implementation-visible subsystem may still be inspected in **Advanced Systems**, but Advanced is collapsed by default and labels each entry as:
- LIVE
- READY
- SHADOW
- INFRA

A SHADOW or INFRA card must never imply gameplay authority.

---

## Timeline / Chronicle

Track:
- Birth
- Death
- Discovery
- Build
- Craft / production milestone
- Mentor / knowledge transfer
- future Faction / Leader / Disaster / Conflict

Current event drill-down:
- event selection
- evidence-first cause panel
- jump to agent
- jump to retained authoritative place evidence when available
- explicit UNKNOWN when historical cause/provenance was not retained
- current Why may be inspected separately but is never presented as historical intent

Roadmap:
- consequence chain when durable cross-event links exist
- generation view
- replay only when an actual replay authority exists

---

## Knowledge View

แสดงเส้นทางความรู้:
- discoverer
- mentor
- parent
- faction
- cultural archive
- learner
- version

---

## Relationship View

เลือก Agent หนึ่งคน แล้วแสดง graph รอบตัว:
- family
- friend
- rival
- mentor
- faction

ห้ามวาด social graph ทั้งโลกพร้อมกันจนอ่านไม่ได้

---

## Mobile UX

- World เต็มจอเป็น default
- Inspector → Bottom Sheet
- Main nav → Bottom Bar
- Thought / task feedback → icon first
- Advanced systems → collapsed disclosure
- Timeline → horizontal scroll
- Large information surfaces must be dismissible and must pause/gate world interaction as appropriate

---

# 17. Technical Modules

```text
world/
agents/
decision/
skills/
knowledge/
society/
simulation/
ai/
verification/
ui/
persistence/
```

## Tick Pipeline

```text
Tick
→ World Update
→ Perception
→ Needs
→ Decision
→ Validate
→ Execute
→ Outcome
→ Memory
→ Learning
→ Society/Event
→ Snapshot
→ UI Render
```

### Critical Rule
UI อ่าน state จาก Engine เท่านั้น  
UI ห้ามเป็นเจ้าของกฎ simulation

---

# 18. Minimum Data Model

## Agent
- id
- name
- generation
- parent
- appearance
- life
- needs
- goal
- faction

## Skill
- id
- level
- xp
- source
- preconditions
- effects

## Memory
- agent
- event
- time
- perception
- valence
- provenance

## Knowledge
- id
- version
- discoverer
- evidence
- visibility

## Relationship
- a
- b
- trust
- affinity
- respect
- fear
- debt

## Faction
- id
- leader
- members
- goals
- rules
- resources

## WorldEvent
- id
- type
- position
- participants
- effects
- time

## Chronicle
- event
- significance
- narrative
- provenance

---

# 19. Verification / Regression

ต้องตรวจอย่างน้อย:

- Population และ life stages
- Generation เกิดเองได้
- Parent inheritance
- Cultural Archive
- Skill V1 → V2 → V3
- Needs / resources / building
- Relationship / Faction persistence
- Aging / Death
- Pause / Speed / Reset
- Inspector ตรงกับ simulation state
- seed + event log reproduce bug

---

# 20. VIP Development Rule

ทุก milestone ใช้:

```text
Verify
→ Improve
→ Prove
```

## Before Coding
กำหนด:
- Success Contract
- Definition of Done
- Verification Plan

## After First Result
ผลแรก = candidate

## Verification State
- SAT
- VIOL
- UNKNOWN

**UNKNOWN ห้ามนับเป็น PASS**

---

# 21. Performance Strategy

- Agent ไกลใช้ simulation fidelity ต่ำลง
- Known skills ใช้ deterministic code
- Decision scoring ทำ batch / CPU
- Perception จำกัดพื้นที่
- Memory retrieval เฉพาะที่เกี่ยวข้อง
- LLM มี budget / cooldown
- Full fidelity เฉพาะพื้นที่สำคัญ

---

# 22. MVP Content Scope

- 1 main biome
- forest / ore / water / base
- 3–5 resources
- 6–8 jobs
- 5–8 buildings
- 6–10 skill families
- 3 natural disasters
- 2–3 basic animals/enemies
- 3 initial faction archetypes

---

# 23. Do Not Build Before V1.0

- ไม่ทำ 3D ใหญ่ก่อน Decision / Perception ผ่าน
- ไม่เพิ่มหลายเมืองก่อนโลกเดียวอยู่รอด 100 วัน
- ไม่เรียก LLM ทุก Agent ทุก tick
- ไม่ทำสงครามซับซ้อนก่อน Social/Faction เสถียร
- ไม่ rewrite engine เพื่อแก้ UI
- UI แสดงค่าได้ไม่ถือว่าระบบผ่าน ต้องมี test

---

# 24. Next Build Order

1. Simulation Skeleton
2. Survival loop
3. Lifecycle
4. Skill system
5. Memory / Knowledge
6. Stabilized baseline
7. V0.7 Decision Brain + Decision Trace
8. V0.8 Perception + Local Knowledge
9. V0.9 Spatial + Pathfinding
10. V1.0 100-Day Autonomous Proof
11. Social
12. Faction
13. Economy
14. Conflict
15. Culture
16. Discovery
17. Evolution
18. AI Brain
19. AI CPU / Reasoning Compilation
20. V2.0 Living World

---

# 25. Final Product Vision

ผู้เล่นสามารถกลับมาเปิดโลกหลังปล่อย simulation ไปหลายร้อยปี แล้วพบว่า Clone รุ่นหลังสร้างเมือง กลุ่ม ความรู้ เทคโนโลยี และเรื่องเล่าที่ไม่ได้ถูก script ตายตัวตั้งแต่แรก

ผู้เล่นเลือกเป็น:
- ผู้สร้าง
- ผู้สอน
- นักสำรวจ
- ผู้สังเกต
- ผู้แทรกแซง

แต่โลกต้องมีแรงขับของตัวเอง

## Heart of the Game

```text
คนหนึ่งพบสิ่งใหม่
→ เรียนรู้
→ ถ่ายทอด
→ คนรุ่นหลังดัดแปลง
→ สังคมเปลี่ยน
→ เหตุการณ์กลายเป็นประวัติศาสตร์
```