--- title: "memory.wiki 사업계획 v7 (FINAL)" url: https://memory.wiki/xv6A2DKS updated: 2026-05-20T14:29:32.291Z source: "chrome" --- # memory.wiki 사업계획 v7 (FINAL) > **Your knowledge graph as a URL for any AI**.Markdown URL = API for any AI. 사용자가 author. AI가 graph로 정리. 어떤 AI든 URL로 사용. > **Brand**: memory.wiki (final, 도메인 확보됨) **Founder**: Hyunsang at Raymind.AI **Build started**: 2026년 3월 (부업), 2026년 4월 풀타임 **Launch deadline**: 2026년 8월 말 (16주) **Last updated**: 2026-05-20 --- # 목차 1. [Executive Summary](#1-executive-summary) 2. [v6 → v7 핵심 변경](#2-v6--v7-%ED%95%B5%EC%8B%AC-%EB%B3%80%EA%B2%BD) 3. [The One Idea](#3-the-one-idea) 4. [Strategic Thesis](#4-strategic-thesis) 5. [3-Tier Architecture (수집/소화/활용)](#5-3-tier-architecture) 6. [Tier 1: 수집 (Capture)](#6-tier-1-%EC%88%98%EC%A7%91-capture) 7. [Tier 2: 소화 (Digestion)](#7-tier-2-%EC%86%8C%ED%99%94-digestion) 8. [Tier 3: 활용 (Utilization)](#8-tier-3-%ED%99%9C%EC%9A%A9-utilization) 9. [Strategic Position](#9-strategic-position) 10. [Business Model](#10-business-model) 11. [16-Week Launch Roadmap](#11-16-week-launch-roadmap) 12. [Launch Strategy](#12-launch-strategy) 13. [12-Month Outlook](#13-12-month-outlook) 14. [Risk Register](#14-risk-register) 15. [Brand Identity (TBD)](#15-brand-identity-tbd) 16. [Decisions Log](#16-decisions-log) --- # 1. Executive Summary ## 한 줄로 > **memory.wiki — your knowledge graph as a URL for any AI.** ## The one idea (from product itself) > A memory.wiki URL is an API for any AI. Paste `memory.wiki/`, `memory.wiki/b/`, or `memory.wiki/hub/` into ChatGPT, Claude, Gemini, or Cursor. LLM이 fetch해서 markdown으로 받음. No app to install. No format to learn. Just markdown over HTTPS. ## 핵심 차별화 ### "The graph is the product" memory.wiki는 모든 surface가 같은 underlying graph를 읽음: - Concept index (lifelong ontology) - Concept relations (typed edges) - Bundle graph_data (themes, insights, edges) - Embeddings (semantic recall) → User가 떠나려면 graph를 잃음. Graph = user의 thinking shape. ### Delivery model, not retrieval - Mem0/Letta: retrieve memory from extraction - LLM Wiki: retrieve knowledge from local compilation - memory.wiki: **deliver knowledge graph in URL** (any AI inherits) 이게 진짜 moat. AI 회사들이 못 copy하는 것. ## 8월 launch까지 도달 지점 - 3단계 (수집/소화/활용) 각각 핵심 enhancement 1개씩 ship - Beta 100명 → Public 2,000+ signup - HN Show HN top 3 - $1,500-2,500 MRR - 첫 partnership conversations (Cursor, Continue 등) - Bundle Spec v1.0 RFC 공개 ## 12개월 후 - $80-150K MRR - 10,000-20,000 paying users - "AI memory" 카테고리 top 5 인지 - 첫 LLM platform integration - Bundle Spec community 활성화 --- # 2. v6 → v7 핵심 변경 이건 진짜 큰 update예요. 실제 product (mdfy.app/bmZYfZez + mdfy.app/KRKz_MD-) 기반. ## 변경 1: Brand final commit **v6**: mori.wiki 후보 단계 **v7**: **memory.wiki** final commit, 도메인 확보됨 ## 변경 2: Strategic thesis sharpen **v6**: "Personal knowledge hub for the AI era" **v7**: "**Markdown URL = API for any AI**" + "**The graph is the product**" doc의 one-idea를 plan thesis로 elevate. ## 변경 3: 3단계 framework backbone **v6**: Three Pillars (Capture / Bundle / Deploy) **v7**: **수집 / 소화 / 활용** (Capture / Digestion / Utilization) Hyunsang님이 제안한 framework. 더 자연스럽고 funnel 명확. ## 변경 4: "Graph"가 핵심 컨셉 **v6**: Bundle, Hub URL 강조 **v7**: **Graph (concept ontology + relations + embeddings)** 강조 The graph is the product. Bundle/Hub는 graph의 view. ## 변경 5: Launch timing 수정 **v6**: 2026-06-16 (8주) **v7**: **2026년 8월 말** (16주, doc 명시) Sustainable timeline. Quality 시간 확보. ## 변경 6: Already-built 자산 명시 **v6**: Multi-surface는 언급 정도 **v7**: 다음 자산 모두 명시: - Real-time collaboration (Yjs CRDT) - 26 MCP tools - /galaxy 시각화 - Concept ontology auto-extraction - 3-provider AI failover - One renderer everywhere - Compact vs Full payload ## 변경 7: Rust engine 사실 정정 **v6**: mdcore (Rust) open source 자랑 **v7**: 2026-05-16에 sunset, markdown-it (JS) only Manifesto와 marketing에서 Rust 언급 제거. ## 변경 8: 도메인 정정 **v6**: mdfy.cc + /ko/ **v7**: memory.wiki (단일 도메인, 글로벌) mdfy.app은 현재 primary, launch 시점에 memory.wiki로 rename. --- # 3. The One Idea doc에서 직접 가져온 product의 정수: > **A memory.wiki URL is an API for any AI.** 이게 전부. 다른 모든 feature는 이 URL이 return하는 markdown을 더 useful하게 만드는 것. ## 3 URL 형태 ### `memory.wiki/` — 한 문서 ``` 한 markdown 문서, frontmatter wrapper 원자 단위 ``` ### `memory.wiki/b/` — bundle (compact 기본) ``` AI graph (themes, insights, edges) + Concepts subgraph + Documents (link + annotation) = ~95% 토큰 절약 vs concat ``` ### `memory.wiki/hub/` — 전체 hub (compact 기본) ``` Top-weighted concepts + Typed relations + Bundle distribution = 사용자 thinking의 shape ``` ## 4 moving parts | Part | What | Where | | --- | --- | --- | | Markdown | Source of truth | `documents.markdown` | | AI graph | Per-bundle 분석 | `bundles.graph_data` (JSON) | | Concept index | Lifelong ontology | `concept_index` + `concept_relations` | | Embeddings | Semantic recall | pgvector (`documents`, `bundles`, `document_chunks`) | 첫 3개는 LLM이 fetch하는 markdown에 포함됨. Embeddings는 MCP tool (`memory_search`)을 통해. --- # 4. Strategic Thesis ## Thesis 1: Delivery, not retrieval 대부분의 AI memory startup은 **retrieval system**. RAG, vector DB, semantic search. memory.wiki는 **delivery system**. Graph가 URL로 ship됨. ``` Retrieval system: AI → query → memory store → return chunks Delivery system: AI → fetch URL → receive pre-organized graph ``` 이게 진짜 moat: - 어떤 AI든 fetch 가능 (no SDK) - Pre-organized (LLM이 navigation 안 해도 됨) - \~95% 토큰 효율 (Compact mode) ## Thesis 2: The graph is the product ``` 모든 mdfy surface는 같은 underlying graph의 다른 reader다. - Galaxy 시각화 = graph의 visual reader - Bundle digest = graph의 LLM reader - Hub URL = graph의 navigation reader - MCP search = graph의 query reader Graph가 product. Surface는 그래프의 representation. ``` 이게 lock-in mechanism: - 사용자가 떠나려면 graph를 잃음 - Graph = 사용자의 thinking shape (수개월 누적) - Export 가능 (open) but reconstruction은 다른 도구로 불가 ## Thesis 3: Markdown URL = universal AI interface ``` Past: AI는 SDK로 접근 (lock-in) Present: AI는 URL로 접근 (open) Future: 모든 AI memory는 URL로 deliver ``` memory.wiki는 이 future를 만드는 시도. ## Thesis 4: Two-way LLM relationship ``` Pull (URL fetch): LLM이 URL을 GET하고 markdown 받음 → 모든 AI 도구 (ChatGPT, Claude, Gemini, Perplexity, Cursor) Push + Pull (MCP): MCP host가 26 tools 호출 → Claude Desktop, Cursor, Cline, Windsurf ``` URL이 paste용, MCP가 deep integration용. ## Thesis 5: LLM platform default integration (Hyunsang님 vision) 장기 목표: > "ChatGPT, Claude, Cursor가 'memory.wiki 연결하시겠어요?' 물어보고, 사용자가 yes 하면 모든 대화에 자동 컨텍스트." 지금: 사용자가 URL paste 미래: Platform 기본 지원 이 path는: 1. Phase 2-3: 사용자가 manual paste, partnership outreach 시작 2. Phase 4: 첫 partnership 발표 3. Year 2-3: Default integration 위치 차지 --- # 5. 3-Tier Architecture Hyunsang님의 framework. Plan의 backbone. ``` ┌────────────────────────────────────────┐ │ 1. 수집 (Capture) │ │ Multi-surface로 markdown 모음 │ │ ↓ │ │ Documents at URLs │ └────────────────────────────────────────┘ ↓ ┌────────────────────────────────────────┐ │ 2. 소화 (Digestion) │ │ AI가 자동으로 graph build │ │ - Concept extraction │ │ - Relations 추출 │ │ - Bundle analysis │ │ - Embeddings │ └────────────────────────────────────────┘ ↓ ┌────────────────────────────────────────┐ │ 3. 활용 (Utilization) │ │ 어떤 AI에든 URL로 deliver │ │ - Doc URL │ │ - Bundle URL (Compact / Full) │ │ - Hub URL (Compact / Full / Since) │ │ - MCP integration │ └────────────────────────────────────────┘ ``` 각 단계가 funnel: - 수집 = entry (모두 가능) - 소화 = engagement (사용 누적시 자동) - 활용 = lock-in (deploy하면 platform 종속) 각 단계가 pricing tier 자연 매핑: - Free: 수집 unlimited, 소화 limited, 활용 basic - Pro: 모두 풀 (개인 power user) - Build: 활용 + API/MCP (AI builder) - Team: 공유 활용 - Enterprise: self-host --- # 6. Tier 1: 수집 (Capture) ## 현재 상태 (이미 강함) ✅ **Web** — mdfy.app (live, Vercel) ✅ **Chrome extension** — ChatGPT, Claude, Gemini capture ✅ **VS Code extension** — Marketplace v1.4.0 ✅ **Desktop app** — DMG v2.2.0, signed + notarized ✅ **CLI** — npm v1.3.x ✅ **MCP server** — npm v1.3.x, 26 tools ✅ **macOS QuickLook** — Bundled in DMG ✅ **GitHub import** — repo → markdown ✅ **Drag-drop, paste-as-markdown** ## 진짜 강점: 다른 도구가 따라잡기 어려운 multi-surface 대부분 memory startup은 web + API만. memory.wiki는 7+ surface가 day one부터. 이게 product의 1st moat. ## 강화 후보 (모두 keep, launch까지 일부 build) ### A. Voice capture ⭐ (Tier 1 priority) **왜 중요**: Mobile에 가장 큰 missing piece. 출퇴근, 산책, 회의 후 즉시 사고 capture. **구현**: - iOS app + Android app - 음성 → Whisper → markdown - Auto-categorize (concept extract) - Hub에 자동 추가 **Launch 전 build 가능성**: Medium (1-2개월). Mobile app은 큰 effort. **Launch 후 priority**: 매우 높음 (가장 큰 새 surface) ### B. iOS Share Sheet / Android Intent ⭐ (Quick win) **왜 중요**: 시스템 레벨 통합. 어떤 앱에서든 "Share to memory.wiki". **구현**: - iOS Share Extension - Android Share Intent - Twitter thread, blog, AI 답변 → 한 번에 hub **Launch 전 build 가능성**: High (2-3주). 빠른 win. ### C. AI conversation 자동 capture suggest ⭐ **왜 중요**: 현재는 사용자가 click. 자동 suggest로 engagement 증가. **구현**: - Chrome extension upgrade - ChatGPT/Claude 대화 모니터링 (local) - "Interesting" 답변 detect (heuristics or AI) - "Save this?" subtle prompt **Launch 전 build 가능성**: Medium (3-4주). 가치 매우 큼. ### D. Email forward **왜 중요**: Viral mechanic. Newsletter, AI 답변 이메일 직접 hub로. **구현**: - `forward@memory.wiki` (또는 user-specific) - Email → markdown (HTML 정리) - Auto-add to hub - Newsletter sync 기반 **Launch 전 build 가능성**: Medium (2-3주). ### E. Slack/Discord integration **왜 중요**: Team tier에 deep. Slack 메시지 react → hub 저장. **구현**: - Slack app - React with custom emoji → memory.wiki에 save - Discord bot - Threads 전체 또는 single message **Launch 전 build 가능성**: Low (1개월+). Team feature이라 Phase 3에 적합. ### F. RSS/Newsletter sync **왜 중요**: 자동 input. Newsletter, blog, podcast description 자동 추가. **구현**: - RSS feed 등록 - New post → hub에 add (draft 또는 auto-publish) - Filter rules **Launch 전 build 가능성**: Medium (2-3주). 매우 niche. ## 수집 강화 우선순위 (Launch까지) | 우선순위 | Feature | Build 시간 | 효과 | | --- | --- | --- | --- | | **1** | **iOS Share Sheet** | 2-3주 | Universal capture, 즉시 ROI | | **2** | **AI conversation auto-suggest** | 3-4주 | Engagement 강화 | | 3 | Email forward | 2-3주 | Viral 가능성 | | 4 | Voice capture (POC) | 4-6주 | 새 surface, mobile 진입 | | 5 | RSS sync | 2-3주 | Niche | | 6 | Slack integration | 4주+ | Phase 3로 미루기 | **Launch 전 commit**: 1, 2 (iOS Share Sheet + AI auto-suggest). **Launch 후 즉시**: 3, 4 (Email + Voice). **Phase 3**: 5, 6 (RSS, Slack). ## 수집 KPI - Captures per user per week (engagement) - % users with 2+ surfaces (multi-surface adoption) - Daily active capture rate - Capture → first publish 시간 (UX) --- # 7. Tier 2: 소화 (Digestion) ## 현재 상태 (이미 강함, 단 underleveraged) ✅ **Concept ontology auto-extraction** (doc_ontology job, 30-min throttle) ✅ **Bundle graph analysis** (themes, insights, edges via LLM) ✅ **Embeddings** (doc + chunk via OpenAI text-embedding-3-small) ✅ **Concept relations** (typed edges between concepts) ✅ **Concept weights accumulation** (cross-bundle) ✅ **3-provider AI failover** (Anthropic → OpenAI → Gemini) ✅ **/galaxy 시각화** (concept index visual reader) ✅ **/canvas 시각화** (per-bundle graph visual) ✅ **analysis_stale tracking** (bundle freshness) ## 진짜 강점: Auto graph build 대부분 도구는 사용자가 organize. memory.wiki는 **AI가 자동**: - 매 doc save → concept 자동 추출 - Bundle 만들면 → graph 자동 분석 - Concept relations 자동 build - 모든 게 background 이게 product의 2nd moat. Lock-in mechanism. ## 강화 후보 (모두 keep) ### A. Bundle 자동 생성 ⭐ (Tier 2 priority) **왜 중요**: 현재는 manual. AI가 concept clustering으로 자동 제안. **구현**: - 사용자가 자연어 요청: "Project Acme 관련 묶어줘" - AI가 concept_index + embeddings 검색 - Bundle draft 생성 + annotations - 사용자 review/edit/save 또는 자동 suggest: - "이 5 docs가 X 주제로 cluster됩니다. Bundle 만들까요?" - 사용자 yes/no **Launch 전 build 가능성**: High (3-4주). Magic UX. ### B. Question-driven exploration ⭐ **왜 중요**: 사용자가 hub에 자연어 질문 → AI가 즉시 답변 + 관련 docs surface. **구현**: - memory.wiki/hub 페이지에 chat interface - "내가 LLM memory에 대해 뭐 알고 있어?" - AI가 embedding search + concept lookup - 답변 + cited docs **Launch 전 build 가능성**: High (2-3주). 가장 강력한 AI 활용. ### C. Auto-summarization (digest) **왜 중요**: Return user mechanic. 매일/매주/매월 review. **구현**: - Daily digest: "오늘 추가한 5 docs 요약" - Weekly digest: "이번 주 thinking" - Monthly review: "concept growth chart" - Email 또는 in-app notification **Launch 전 build 가능성**: Medium (3주). ### D. Cross-bundle auto-link **왜 중요**: Bundle A의 concept이 Bundle B에 등장 → 자동 cross-reference. **구현**: - Concept_relations 이미 있음 → activate - Bundle UI에 "Related bundles" 섹션 - 자동 link via shared concepts - Concept hub 자체에서 cross-bundle navigation **Launch 전 build 가능성**: Medium (2주). 이미 backend 80% 있음. ### E. Contradiction detection **왜 중요**: Power user feature. 두 doc이 contradictory하면 surface. **구현**: - LLM이 concept-level contradiction detection - "Doc A는 X라고 하는데 Doc B는 not X" - 사용자에게 surface - Resolve or reconcile prompt **Launch 전 build 가능성**: Low (4주+). Niche but impressive. ### F. Knowledge gap detection **왜 중요**: "이 concept group에 X 비어있다" → active learning. **구현**: - Concept cluster 분석 - Gaps surface ("X에 대한 doc 없음") - "Search/capture suggest" **Launch 전 build 가능성**: Low (4주+). Niche. ### G. Concept evolution tracking **왜 중요**: "내 thinking이 6개월 전 vs 지금 어떻게 변했나" **구현**: - Concept snapshot 매월 - Diff 비교 - Personal growth narrative - Visualization **Launch 전 build 가능성**: Low (4주+). Long-term value, not launch critical. ## 소화 강화 우선순위 (Launch까지) | 우선순위 | Feature | Build 시간 | 효과 | | --- | --- | --- | --- | | **1** | **Bundle 자동 생성** | 3-4주 | Magic UX, conversion driver | | **2** | **Question-driven exploration** | 2-3주 | 가장 강력한 AI 활용 | | 3 | Cross-bundle auto-link | 2주 | Backend 80% 준비됨 | | 4 | Auto-summarization digest | 3주 | Return user | | 5 | Contradiction detection | 4주+ | Phase 3 | | 6 | Knowledge gap detection | 4주+ | Phase 3 | | 7 | Concept evolution | 4주+ | Phase 3 | **Launch 전 commit**: 1, 2 (Bundle auto-gen + Question exploration). **Launch 후 즉시**: 3, 4 (Cross-link + Digest). **Phase 3**: 5, 6, 7 (Contradiction, Gap, Evolution). ## 소화 KPI - Bundles created per user (manual + AI) - AI bundle generation usage rate - Concept index size per user - Hub query frequency - Return user rate (digest engagement) --- # 8. Tier 3: 활용 (Utilization) ## 현재 상태 (이미 매우 강함) ✅ **Doc URL** — `mdfy.app/` (markdown + frontmatter) ✅ **Bundle URL Compact** — `mdfy.app/b/` (\~95% 토큰 절약) ✅ **Bundle URL Full** — `?full=1` (모든 body inline) ✅ **Hub URL Compact** — `mdfy.app/hub/` (concept digest) ✅ **Hub URL Since** — `?since=date` (incremental) ✅ **/raw payload** — for AIs / scrapers ✅ **/embed** — iframe-friendly ✅ **/d/** — reader-only viewer ✅ **MCP server (26 tools)** — push + pull integration ✅ **Visitor "Ask AI"** — 공개 doc에 누구나 AI 질문 ✅ **Real-time collaboration** — Yjs CRDT ✅ **Three permission roles** — owner / editor / readonly ✅ **Edit mode controls** — owner / account / token / view / public ## 진짜 강점: 3 scope URL + Compact/Full + MCP 이게 product의 3rd moat. AI 회사들이 "delivery model"을 못 copy. ## 강화 후보 (모두 keep) ### A. Direct LLM platform integration ⭐ (Tier 3 priority — Hyunsang님 vision) **왜 중요**: "LLM의 default memory layer" vision의 직접 구현. **구현**: - "Add memory.wiki to ChatGPT" 버튼 → Custom GPT 자동 설치 - "Add memory.wiki to Claude" → MCP 자동 연결 - "Add memory.wiki to Cursor" → cursor rules 자동 설정 - 한 번 click → 모든 대화에 자동 컨텍스트 **Launch 전 build 가능성**: Medium (3-4주). Custom GPT부터 시작 가능. **전략적 가치**: 매우 큼. Partnership 시작점. ### B. Embedded chat with hub context ⭐ **왜 중요**: memory.wiki에서 직접 chat, hub 전체 자동 context. ChatGPT 안 가도 됨. **구현**: - mdfy.app에 chat UI - 사용자 hub 자동 context (Compact) - 3-provider failover - Conversation history → 새 docs 생성 가능 **Launch 전 build 가능성**: High (2-3주). 기술 자산 이미 있음. **전략적 가치**: Sticky engagement. ChatGPT/Claude lock-out도 가능. ### C. Public hub sharing ⭐ **왜 중요**: 가장 viral mechanic. Substack-like "knowledge feed". **구현**: - "내 hub 공개" 옵션 - Public hub URL - 다른 사람이 따라가기 - Concept-level subscribe - "Following" feed **Launch 전 build 가능성**: Medium (3주). 가장 viral. ### D. AI-specific format adapters **왜 중요**: Claude/ChatGPT/Cursor 각각 최적 format. **구현**: - Claude: XML tags - ChatGPT: verbose markdown - Cursor: code-context 강조 - Gemini: structured URL parameter로 지정: - `?format=claude` - `?format=cursor` **Launch 전 build 가능성**: Medium (2-3주). Quality moat. ### E. Smart Hub digest **왜 중요**: Context-aware digest. "지금 작업 중인 것" 인식. **구현**: - Recent 활동 분석 - Active concept cluster surface - "Working context" vs "background" 구분 **Launch 전 build 가능성**: Low (4주+). UX polish. ### F. Hub-to-hub merge **왜 중요**: Team의 collective concept index. **구현**: - 두 hub 합치기 - Conflict resolution - Permission inheritance - Team tier 핵심 **Launch 전 build 가능성**: Low (4주+). Team feature, Phase 3. ### G. Time travel **왜 중요**: "3개월 전 내 hub vs 지금". Personal growth narrative. **구현**: - Hub snapshot 자동 - Diff 비교 - Timeline view **Launch 전 build 가능성**: Low (4주+). Power user. ## 활용 강화 우선순위 (Launch까지) | 우선순위 | Feature | Build 시간 | 효과 | | --- | --- | --- | --- | | **1** | **Direct LLM platform integration (Custom GPT)** | 3-4주 | Vision 직접 | | **2** | **Embedded chat with hub context** | 2-3주 | Sticky engagement | | **3** | **Public hub sharing** | 3주 | Viral mechanic | | 4 | AI-specific format adapters | 2-3주 | Quality moat | | 5 | Smart Hub digest | 4주+ | Phase 3 | | 6 | Hub-to-hub merge | 4주+ | Team tier | | 7 | Time travel | 4주+ | Power user | **Launch 전 commit**: 1, 2, 3 (Platform integration + Embedded chat + Public hub). **Launch 후 즉시**: 4 (Format adapters). **Phase 3**: 5, 6, 7. ## 활용 KPI - Hub URL deploys per user per week (활용 빈도) - LLM platform integrations active (partnership) - Public hub follows (viral signal) - MCP tool usage rate - Embedded chat engagement --- # 9. Strategic Position ## 시장 자리 ``` Retrieval-focused Delivery-focused (query systems) (URL-based) ↓ ↓ Auto-extracted Mem0, Letta [없음 — gap] OpenAI Memory Google Memory Bank User-authored Notion (closed) [memory.wiki] ⭐ Obsidian (local) LLM-compiled LLM Wiki (Karpathy) [memory.wiki Phase 3] (local Obsidian) (cloud variant) ``` memory.wiki의 자리: **User-authored + Delivery-focused + URL-based**. 이 자리가 비어있음. memory.wiki가 차지. ## 5 Pillars of memory.wiki 1. **URL as API** — 어떤 AI든 fetch 가능, no SDK 2. **Graph is product** — concept ontology + relations 3. **Multi-surface capture** — 7+ surfaces from day one 4. **AI-augmented authorship** — user author + AI organize 5. **Open Bundle Spec** — standard formation ## 차별화 narrative (sharp) ### vs Mem0/Letta > "They extract memory from your conversations. memory.wiki delivers your knowledge graph to any AI." ### vs Karpathy LLM Wiki > "LLM Wiki is one local knowledge base, LLM-compiled. memory.wiki is doc + bundle + hub — scoped composition for cloud delivery." ### vs Notion / Obsidian > "Notion locks knowledge in workspace. Obsidian locks in folder. memory.wiki ships knowledge as URL — readable by any AI." ### vs ChatGPT Memory / OpenAI > "ChatGPT remembers about you, inside ChatGPT. memory.wiki is your knowledge, deployable to ChatGPT, Claude, Cursor — anywhere." ## Honest positioning > Mem0와 Letta는 자동 추출에 훌륭합니다. Karpathy LLM Wiki는 local compilation에 훌륭합니다. Notion은 closed workspace에 훌륭합니다. memory.wiki는 그 사이의 layer입니다 — user-authored, AI-organized, URL-delivered, multi-AI compatible. ## Strategic moat 3 layers of defensibility: 1. **Multi-surface infrastructure** — 7+ surfaces, day one shipped 2. **Graph engine** — concept ontology auto-build, accumulating 3. **Delivery model** — URL-based, any AI inherits 이 셋이 합쳐지면: - 사용자 떠나기 어려움 (graph lock-in) - 경쟁사 따라잡기 어려움 (multi-surface) - AI 회사들이 copy 어려움 (delivery infrastructure) --- # 10. Business Model ## Pricing (v7 — 3-tier framework 반영) ``` Free - 수집 unlimited (모든 surfaces) - 소화 limited (월 50 docs concept extract, 3 bundles AI analyze) - 활용 basic (Doc URL public only) - Real-time collab (2 users) - Community MCP server (read-only) Pro $9/mo (개인) ⭐ 수집 + 소화 full - 수집 unlimited - 소화 unlimited (concept extract, bundle analyze) - 활용 essential: * Private docs * Custom domain * Bundle Compact + Full * Hub URL (basic) * 10 AI-generated bundles/month * Embedded chat (limited) - Tags, folders - Version history - Custom domain Build $19/mo (Power user, AI builder) ⭐ 활용 full - All Pro features - 활용 unlimited: * Unlimited AI-generated bundles * AI-specific format adapters (Claude/GPT/Cursor) * Embedded chat unlimited * Public hub sharing * Hub URL custom domain - API access (read/write) - MCP server full (write enabled) - Bundle versioning + snapshots - Webhook integrations - Voice capture (mobile) Team $19/seat/mo - All Build features - Shared workspaces - Shared bundles + hubs - Hub-to-hub merge - Permissions, audit log - SSO (Google → SAML) - Slack/Discord integration Scale $499+/mo - Public bundle marketplace - Custom rate limits - SLA, priority support - Dedicated MCP infrastructure - Multi-region Enterprise (협의) - Self-host option - SAML SSO - LLM Wiki integration support - Custom audit/compliance - Direct LLM platform partnership pricing ``` ## Pricing 의도 (3-tier driven) ### Free → Pro $9 - Free에서 publishing 충분 - Pro upgrade 동기: **AI-generated bundles 10개/월** (가장 강력) **Private docs** **Custom domain** **소화 unlimited** ### Pro → Build $19 - Bundle 무제한 AI generation - **AI-specific format adapters** (quality moat) - **Public hub sharing** (viral mechanic) - **API/MCP write access** (developer) ### Build → Team - Shared workspaces - Hub-to-hub merge - Audit log ### Team → Enterprise - Self-host - Platform partnership - SOC 2 ## Unit Economics ### Pro $9 - Stripe: $0.56 - Hosting: \~$0.50 - AI base (concept extract, bundle analyze): \~$1.00 - AI generation (avg 5/month): \~$0.30 - Embeddings: \~$0.05 - **Margin: \~74%** ### Build $19 - Stripe: $0.85 - Hosting: \~$0.80 - Unlimited AI generation: \~$2.00 - Format adapters processing: \~$0.30 - API/MCP infra: \~$0.50 - **Margin: \~71%** Margin이 v6보다 낮음 (AI cost 추가). 단 conversion이 더 강력해서 LTV 증가. ## Revenue Projections (12개월) ### Conservative - 3,000 paid users - ARPU $11 - $33K MRR / $400K ARR ### Realistic - 8,000 paid users - ARPU $11 - $88K MRR / $1.0M ARR ### Optimistic - 18,000 paid users - ARPU $12 - $216K MRR / $2.5M ARR --- # 11. 16-Week Launch Roadmap ## 전체 schedule (Now → 2026년 8월 말) ``` Week 1-4 Brand + Foundation Week 5-8 Tier 1 강화 (수집) Week 9-11 Tier 2 강화 (소화) Week 12-14 Tier 3 강화 (활용) Week 15 Beta + Polish Week 16 Public Launch ``` ## Week 1-4: Brand + Foundation ### Week 1 - Brand identity 결정 (visual) - About 페이지 v7 framing 적용 - Manifesto v7 작성 - memory.wiki landing page 활성화 (SEO 시작) ### Week 2 - mdfy.app → memory.wiki migration 시작 (phased) - Hero 새 카피 push - Three pillars 새 본문 (수집/소화/활용) - Bundle Spec v1.0 draft 공개 준비 ### Week 3 - 한국어 사이트 (선택, 또는 launch 후) - Pricing 페이지 update (Build tier 활성화) - Demo 영상 시나리오 작성 (3편) - Press kit 준비 ### Week 4 - Investor 자산 (one-pager, deck draft) - Beta tester 모집 시작 (50명 target) - Manifesto post (Substack/blog) ## Week 5-8: Tier 1 강화 (수집) ### Week 5-6: iOS Share Sheet - iOS Share Extension 개발 - Universal capture from any iOS app - TestFlight 배포 - 기존 사용자 (Hyunsang님) 자체 테스트 ### Week 7-8: AI conversation auto-suggest - Chrome extension upgrade - ChatGPT/Claude 대화 monitor - "Save this?" subtle prompt - A/B test heuristics ## Week 9-11: Tier 2 강화 (소화) ### Week 9-10: Bundle 자동 생성 - AI bundle generation backend - 자연어 요청 → bundle draft - 사용자 review/edit UX - Cross-bundle auto-link activate ### Week 11: Question-driven exploration - Hub chat interface - Embeddings + concept lookup - Cited answers UI ## Week 12-14: Tier 3 강화 (활용) ### Week 12: Embedded chat with hub context - mdfy.app/chat UI - Hub auto-context - Conversation history → docs ### Week 13: Public hub sharing - "Make hub public" option - Public hub URL - Following feed (basic) ### Week 14: Direct LLM platform integration (Custom GPT first) - ChatGPT Custom GPT for memory.wiki - One-click install - Optional: Claude MCP guide, Cursor rules ## Week 15: Beta + Polish - 50명 beta tester 본격 운영 - 매일 피드백 처리 - Bug fix - Onboarding 최적화 ## Week 16: Public Launch ⭐ ### Day 1-2: Final prep - Demo 영상 3편 완성 (수집/소화/활용) - Show HN 글 5개 후보 finalize - Twitter thread 10 tweets - Product Hunt 페이지 ### Day 3: Show HN - 오전 9시 PST - "Show HN: memory.wiki — your knowledge graph as a URL for any AI" ### Day 4: Twitter + dev community ### Day 5: Product Hunt ### Day 6: Manifesto post + Bundle Spec announcement ### Day 7: AI newsletters (Latent Space, AI Engineer, etc.) --- # 12. Launch Strategy ## Big Launch v7 3 narratives 동시에: ### 1. Product narrative > "memory.wiki — your knowledge graph as a URL for any AI." Multi-surface capture. AI auto-graph. Universal delivery. ### 2. Open standard narrative > "Bundle Spec v1.0 — open standard for AI-deployable knowledge." GitHub RFC. Reference implementation. Community formation. ### 3. Founder narrative > "I built memory.wiki in 5 months. Multi-surface, AI graph, MCP-ready. Now going full-time. Betting that markdown URLs become AI's universal interface." ## Launch Day 채널 (Week 16) ### Tuesday: Show HN **제목 후보**: 1. "Show HN: memory.wiki — your knowledge graph as a URL for any AI" 2. "Show HN: I built a delivery-first AI memory layer (vs Mem0's extraction)" 3. "Show HN: Bundle Spec — open standard for AI-deployable knowledge" 4. "Show HN: memory.wiki + 26-tool MCP server for any LLM host" 5. "Show HN: A wiki for AI that any LLM can fetch as a URL" 저는 **1 또는 5** 추천. Sharp + accessible. ### Wednesday: Twitter/X - 10-tweet thread - Demo GIFs (각 단계별) - Karpathy LLM Wiki 보완재 framing ### Thursday: Product Hunt - "memory.wiki — Your knowledge graph as a URL for any AI" ### Friday: Manifesto + Bundle Spec - Manifesto post (1,800 단어) - Bundle Spec announcement post (separate) - AI newsletters ## Pre-Launch Strategic Conversations Week 12-15에 outreach: ### Tier 1 (Partnership priority) - **Anthropic**: MCP integration showcase - **Cursor**: Memory layer integration (cursor rules) - **Continue (OSS)**: Reference implementation - **Cognition / Devin**: Agent memory backbone ### Tier 2 (Awareness) - **OpenAI**: Custom GPT showcase - **Perplexity**: Research memory - **Replit Agent**: Memory infrastructure ### Tier 3 (Community) - **LLM Wiki community**: Spec feedback, complementary framing - **Karpathy**: Spec 인지 (가능하면) - **Markdown community** (Obsidian, HackMD): Awareness ## Bundle Spec 발표 ### Pre-launch (Week 14-15) - GitHub: github.com/raymindai/bundle-spec - README + full spec - Reference impl: memory.wiki ### Launch day - Spec announcement post - HN/Twitter에 spec 링크 - LLM Wiki community에 보완재 framing ### Post-launch - RFC iteration - Community contributions - Other tools implement (Obsidian plugin 가능성) ## KPI 목표 (v7) ### Week 16 (Launch) - 2,000-3,000 signup - 500+ Chrome ext install - 100-150 paid (mix Pro $9 + Build $19) - $1,500-2,500 MRR - 100+ Bundle 생성 (50%+ AI-generated) - 50+ Hub URL 활성 - HN top 3 - Bundle Spec GitHub stars 200+ - 5-10 media mentions ### Month 3 (Post-launch) - 8,000-12,000 signup - $8-15K MRR - 500+ AI-generated bundles/week - 100+ Public hubs - 첫 LLM platform partnership 발표 ### Month 6 - 30,000-50,000 signup - $25-40K MRR - Voice capture launched (mobile) - AI agent integrations 15+ - Team workspace beta ### Month 9 - 100,000+ signup - $50-80K MRR - Bundle Spec v1.0 final - 첫 enterprise pilot - LLM Wiki Obsidian integration POC ### Month 12 - $80-150K MRR - 카테고리 인지 top 5 - Strategic conversations (Anthropic, OpenAI 등) - Series A 가능 시점 --- # 13. 12-Month Outlook ## Year 1 outcome 시나리오 ### Conservative - $50-100K MRR - 8,000-15,000 paying users - Bundle Spec published, 200+ stars - Voice capture launched - 첫 partnership announcement ### Realistic - $100-200K MRR - 15,000-25,000 paying users - Bundle Spec v1.0 final - LLM Wiki integration - 첫 enterprise pilot - Series A optional ### Optimistic - $250-400K MRR - 30,000+ paying users - Bundle Spec adoption (3+ tools) - 첫 LLM platform default integration - Series A active ## Year 2 ### Conservative - $500K-1.5M ARR - Sustainable indie + spec community - 강한 카테고리 brand ### Realistic - $2-5M ARR - Bundle Spec open standard - 첫 enterprise customers - 외부 자본 검토 또는 받음 ### Optimistic - $5-15M ARR - C2PA-style consortium - Strategic acquisition 제안 ($50-200M) - Series A ($30-100M valuation) ## Year 3 ### Conservative - $1.5-5M ARR sustainable ### Realistic - $10-25M ARR - Bundle Spec industry standard - Acquisition ($100-300M) - 또는 Series B ### Optimistic - $30M+ ARR - LLM platform default integration - Acquisition $200M-1B range - Series B 가능 ## Exit Scenarios ### Strategic Acquisition Candidates - **Anthropic**: Memory layer for Claude/MCP ecosystem - **OpenAI**: Personal memory infrastructure - **GitHub/Microsoft**: Knowledge hub for Copilot - **Notion**: AI memory acquisition (counter-positioning) - **Cursor / Cognition**: Agent memory backbone - **Atlassian / Linear**: Team docs + memory ### Sustainable Indie - $5M+ MRR - Spec community - Personal asset + freedom ### Standards Body - Bundle Spec → industry standard - Long-term shaping (W3C-like) --- # 14. Risk Register ## Risk 1: 거인들 진입 (확률 60%, impact 매우 큼) ### 시나리오 - Anthropic memory layer - OpenAI Memory 확장 - Google Memory Bank ### Mitigation - 16주 launch (first-mover advantage 빠르게) - Bundle Spec 발표 (표준 차지) - Multi-LLM agnostic (거인은 자기 ecosystem만) - Delivery model (closed system 따라잡기 어려움) - LLM Wiki community 동맹 ## Risk 2: Build scope creep (확률 50%, impact 매우 큼) ### 시나리오 - 3 tier 각각 enhancement 모두 build 시도 - Launch 16주 → 24주 → 무한 연기 ### Mitigation - 각 tier launch 전 commit features 명확 (1, 2번만) - 나머지는 Phase 3로 - 매주 progress review - Week 8, 12, 14 reality checks ## Risk 3: 첫 beta tester 모집 어려움 (확률 15%, impact 큼) ### Mitigation - 채널 5개 다각화 (Discord, HN, Twitter, LLM Wiki community, Markdown community) - mdfy.app 이미 사용 가능 (try first, no waitlist) - Manifesto post로 inbound ## Risk 4: Graph 개념 사용자 이해 못 함 (확률 40%, impact 중간) ### 시나리오 - "Concept graph"가 abstract - 사용자가 가치 못 느낌 ### Mitigation - /galaxy 시각화로 즉시 see - Demo 영상에서 use case - "그래프 자동" UX (사용자가 build할 필요 없음) - Question-driven exploration로 즉시 utility ## Risk 5: Bundle Spec adoption 안 됨 (확률 50%, impact 중간) ### Mitigation - memory.wiki가 reference impl - LLM Wiki community 협업 - Spec 단순함 (CommonMark + YAML) - Format adapters로 immediate value ## Risk 6: AI cost spiral (확률 30%, impact 중간) ⭐ NEW ### 시나리오 - Free tier 사용자가 AI feature 남용 - AI generation cost가 revenue 초과 ### Mitigation - Rate limits (월 50 docs concept extract for Free) - AI generation paywall (Pro 10/month) - 3-provider failover (cost optimize) - Embedding cache (`embedding_source_hash`) - Concept extraction throttle (30 min per doc) ## Risk 7: Solo burnout (확률 50%, impact 매우 큼) ### Mitigation - 16주 timeline (8주보다 sustainable) - 매주 1일 OFF - Week 8, 12 reality check - Month 3+ 자동화 - Month 6+ 외주 (콘텐츠, 카피) ## Risk 8: LLM platform 미협력 (확률 40%, impact 큼) ⭐ NEW ### 시나리오 - "Default memory layer" vision - 단 Anthropic/OpenAI가 자기 solution 우선 - Partnership 거부 ### Mitigation - Plan B: bottom-up (사용자 → request 압박) - Custom GPT, MCP는 platform 동의 필요 없음 - 사용자 base 충분히 크면 platform이 협력 동기 - 단기에는 ChatGPT Custom GPT부터 ## Risk 9: Brand rename confusion (확률 20%, impact 중간) ⭐ NEW ### 시나리오 - mdfy.app → memory.wiki migration - 기존 사용자 confusion - SEO 손실 ### Mitigation - Phased migration (Week 2 시작) - mdfy.app → memory.wiki 자동 redirect (영구) - Email 사용자 알림 - Brand story로 narrative 만들기 --- # 15. Brand Identity (TBD) ## 결정된 것 - **Name**: memory.wiki (final, 도메인 확보) - **Tone**: 진지함 + 따뜻함 (knowledge + craft) - **Voice**: Authoritative + accessible ## TBD (Visual) 이전에 mori.wiki 시절 결정한 "90% 숲 + 10% 해골 + pixel art"는 mori (memento mori) 의미와 align한 것. memory.wiki로 변경되면서 reconsider 필요. ### Visual 옵션 **A**: Pixel art forest + skull keep (narrative 조정) > "Memory like a forest. The skull is the reminder." **B**: 새 visual identity (memory에 fit) > Brain, neural network, library 등 motif **C**: Minimalist (no specific motif) > 단순 wordmark, monogram **D**: Decide at launch -2 weeks (doc 전략) ### 추천: D (Launch 임박 시 결정) doc의 strategy 따라 launch 2주 전 finalize. 그때까지 wordmark만 사용. ### Color palette 후보 **Tone 1: 깊고 진지함** - Deep blue + cream - Inkwell knowledge feel **Tone 2: 자연 + 지식** - Forest greens + bone cream (mori legacy) - Growth + memento **Tone 3: AI 시대 modern** - Charcoal + electric blue - Tech but human Launch 2주 전 결정. --- # 16. Decisions Log (v7) ## v6 결정사항 유지 | 결정 | 답 | | --- | --- | | Manifesto core | "Own your markdown, build your knowledge graph" | | Open source | Bundle Spec OSS | | Founder commit | 풀타임 100% | | Multi-surface | Day one core moat | | 3-tier funnel | 수집 / 소화 / 활용 | ## v7 신규 결정사항 | 결정 | 답 | | --- | --- | | **Brand name** | **memory.wiki** (final, 확보) | | **Strategic thesis** | **"Markdown URL = API for any AI" + "Graph is product"** | | **Three Pillars** | **수집 / 소화 / 활용 (Capture / Digestion / Utilization)** | | **Launch deadline** | **2026년 8월 말 (16주)** | | **Visual identity** | **TBD (launch 2주 전 결정)** | | **Existing domain** | **mdfy.app → memory.wiki phased migration** | | **Rust engine narrative** | **제거 (sunset된 사실)** | | **Graph 개념** | **Plan + marketing 중심** | | **Real-time collab** | **Marketing 자산으로 추가** | | **MCP 26 tools** | **자산으로 자랑** | | **/galaxy 시각화** | **Demo 자산** | ## Tier별 launch features (commit) ### 수집 (Tier 1) - ✅ iOS Share Sheet - ✅ AI conversation auto-suggest - ⏳ Voice capture (POC if time) - 🔜 Email forward (post-launch) - 🔜 RSS sync (Phase 3) - 🔜 Slack/Discord (Phase 3) ### 소화 (Tier 2) - ✅ Bundle 자동 생성 - ✅ Question-driven exploration - ✅ Cross-bundle auto-link - 🔜 Auto-summarization digest (post-launch) - 🔜 Contradiction detection (Phase 3) - 🔜 Knowledge gap detection (Phase 3) - 🔜 Concept evolution (Phase 3) ### 활용 (Tier 3) - ✅ Direct LLM platform integration (Custom GPT first) - ✅ Embedded chat with hub context - ✅ Public hub sharing - 🔜 AI-specific format adapters (post-launch) - 🔜 Smart Hub digest (Phase 3) - 🔜 Hub-to-hub merge (Phase 3) - 🔜 Time travel (Phase 3) --- # Mantra ## 매일 self-check > **"이 feature 없으면 launch 못 하는가?**"NO → Phase 3 list로 미룸 > **"3-tier 어느 부분 강화 중인가?**"수집 / 소화 / 활용 각각 1개씩 commit > **"Graph가 작동하는가?**"Graph = product. 깨지면 안 됨. > **"URL = API thesis가 살아있는가?**"모든 결정의 anchor > **"이게 movement를 만드는가?**"Standard-setting ambition. ## 한 줄 > memory.wiki — your knowledge graph as a URL for any AI. 수집은 multi-surface로. 소화는 AI graph로. 활용은 어떤 AI든 URL로. 16주 안에 launch한다. Markdown URLs become the universal AI memory interface. --- # 부록: 관련 문서 ## 작성 완료된 문서 1. mdfy-master-business-plan.md — v4 (deprecated) 2. mdfy-direction-v5.md — v5 (deprecated) 3. mdfy-direction-v6.md — v6 (deprecated) 4. mdfy-manifesto-v6-en.md — v6 manifesto (needs v7 rewrite) 5. mdfy-manifesto-v6-ko.md — v6 한국어 manifesto 6. mdfy-three-pillars-v6.md — v6 pillars 7. mdfy-bundle-spec-v1.md — Bundle Spec (v7 update 필요) 8. mdfy-claude-code-handoff.md — interim site (deprecated) 9. **mdfy-direction-v7.md** — 이 문서 (v7 FINAL) ## 다음 작성 필요 - **Manifesto v7** (memory.wiki + 3-tier + graph thesis 반영) ← 다음 task - Bundle Spec v2 (memory_bundle field 변경) - Three pillars v7 (수집/소화/활용) - Show HN 글 5개 후보 - Demo 영상 시나리오 3편 ## Ground truth references - **OVERVIEW.md**: - **HOW-IT-WORKS.md**: --- *문서 작성: 2026-05-20Brand: memory.wiki (final)Launch: 2026년 8월 말 (16주)Built by: Hyunsang at Raymind.AIExisting: github.com/raymindai/* (mdfy → memory.wiki migration) *Contact: [hi@raymind.ai](mailto:hi@raymind.ai)*