Nick Milo AI OS — Preliminary Adaptation Plan

Source: The 3-File AI System That Works With ANY MODEL — Linking Your Thinking with Nick Milo (14:20)

Status: Preliminary reference only — no implementation. This document maps the model and considers how it might apply to the Baseworks KB. Additional sources will be merged here as the plan develops.

Screenshots: media.baseworks.com/kb-dev/nick-milo-ai-os/ — NAS mirror: /volume1/baseworks/media/kb-dev/nick-milo-ai-os/

---

Core Principle: File Over AI

The Obsidian CEO (Kepano) coined “File over App” — your data outlasts any tool because it lives in plain text files, not inside a proprietary platform. Nick Milo extends this: File over AI. If you build your AI workflow inside Claude, or Gemini, or any specific product, you lose that workflow the moment the tool changes. The AI OS model keeps every AI-related process in portable markdown files you own.

“You want your AI processes owned by you, built into your core files and folders… not locked away in any specific tool.”

---

The Three Layers

Layer 1 — Ideaverse (core)

Your entire personal knowledge base: all notes, ideas, connections, in plain text files. Nick organizes his using the ACE structure — Atlas (knowledge landscape), Calendar (time-based notes), Efforts (projects and tasks). The ideaverse is the center of the AI OS. The more you build it, the smarter any AI becomes when working alongside you.

The ideaverse is irreplaceable — it’s the accumulated thinking that belongs entirely to you and doesn’t exist anywhere else.

Layer 2 — Maps & Manuals (middle)

The connective tissue between your ideaverse and any external AI tool. This layer contains a small set of files that give AI a structured way to navigate your knowledge without having to scan everything.

Without a map, pointing an AI at a vault of 20,000–40,000 files is expensive in tokens, slow, and inaccurate. With a map, AI can identify which files are relevant, skip the rest, and build context quickly.

Layer 3 — Tools (outer, most replaceable)

The actual applications: Obsidian, Claude, Notebook LM, Gemini, Claude Code, local models. Tools will change — that’s expected. New models release, apps get discontinued, better options emerge. The system is designed so your ideaverse and maps stay central regardless of which tool rotates in or out.

---

The Three Core Files

These files all live in Layer 2 (Maps & Manuals). They are plain markdown — no special syntax, no tool-specific features.

---

File 1 — me.md

Question it answers: Who am I? Update frequency: Rarely after initial setup.

A 50–80 line markdown file that briefs any AI on who you are, how you think, and how you want it to work with you. It’s called me.md because it belongs to you — not to Claude, not to any specific tool. It’s your portable identity.

Lives at: vault root (e.g., /me.md) — not inside the AIOS folder, so it’s always easy to find.

Nick’s structure:

Me
  The Summary Statement — high-level who you are and what drives you
  First Principles
  Mantras / LYT principles I live by
  A Hypothesis on Meaning
  Values & Virtues
  How I Think
    The ARC Framework
    Three Thinking Styles (top-down, bottom-up, middle-out)
    Sensemaking Archetypes
  Intellectual Lineage — thinkers who shaped you
  Recurring Concepts & References
  Spark Lists
  Life Reflections
  My Roles (personal / professional)

My Working Preferences
  Preferences (address by first name, brief, honest pushback, collaborator not tool)
  Vibe (skip "great question" / "I'd be happy to help" — just help)
  Basic Rules
  Pointers to other maps

File path links at the top of me.md: At the very top, me.md includes explicit file paths to the other two maps so any AI can navigate from identity to vault structure to skills in a single session.

- me.md – /me.md – Your briefing on who I am, how I think, and how to interact with me.
- vault-map – /AIOS/Maps/vault-map – Your manual to navigate & work in my ideaverse.
- skill-map – /AIOS/Maps/skill-map – All the defined skills you can utilize on my behalf.

The CLAUDE.md bridge: Nick keeps a CLAUDE.md at his vault root — but it contains only one line: “Go immediately to me.md and you will find everything there.” Claude reads it first and immediately pivots to me.md. This is the future-proof bridge: the Claude-specific entry point stays stable as a one-liner, while all actual content lives in the tool-agnostic me.md.

---

File 2 — vault-map

Question it answers: What is my ideaverse and how should I move through it? Update frequency: More often than me.md — whenever folder structure changes or priorities shift.

Lives at: AIOS/Maps/vault-map.md

The vault-map is the AI’s manual for navigating and working inside your ideaverse. It answers three questions:

1. How to navigate the folder structure

Brief overview + expanded overview of every top-level folder: what it contains, how it’s organized, what to look for where.

Nick’s brief overview structure:

+ → Inbox (temporary holding for new items)
AIOS → AI Operating System folder
  History/ → All time-stamped AI interaction logs (YYYY-MM-DD Description)
  Maps/    → vault-map, skill-map, any other navigation files
  Skills/  → Individual skill documents
Atlas/ → Bases, LYT, Maps, PKM, People, Quotes, Sources, Statements, Things
Calendar/ → Days, Meetings, Research, Reviews
Efforts/ → Areas, Projects (Active/Simmering/Sleeping), Tasks, Works
x/ → Templates, visuals, extras

2. How to create notes properly

Note naming conventions, required frontmatter/YAML fields, template locations, and note types for different areas. The AIOS/History folder captures every AI interaction with a date-stamped filename (YYYY-MM-DD Description) so there’s a searchable log of AI activity over time.

3. How to interact with other ideaverses

The vault-map can also describe how this ideaverse connects to other ideaverses — a team vault, a family vault, a shared knowledge base. The Maps & Manuals layer acts as the translation layer, keeping core knowledge private while exposing what’s relevant for shared work.

Multi-ideaverse architecture (future direction): Nick previews connecting multiple ideaverses — his personal one, the LYT team ideaverse, and a family ideaverse — all with their own Maps & Manuals layer. Private knowledge stays private; the translation layer surfaces only what’s needed.

---

File 3 — skill-map

Question it answers: What processes does AI know how to run for me? Update frequency: Created and updated regularly as new workflows are documented.

Lives at: AIOS/Maps/skill-map.md (index) + AIOS/Skills/ (individual skill files)

Each skill is a markdown file — plain documentation for a repeatable process. Because they’re markdown, they travel to any AI tool that reads text. Skills are not tool-specific features; they’re mini maps of content for a process.

A skill file includes:

• name — machine-readable identifier
• triggers — keywords that invoke this skill
• description — what it does in one sentence
• Step-by-step instructions
• Output format

Example: holistic-briefing A morning context briefing. Steps: (1) review active projects by priority, (2) assess book/reading momentum, (3) scan intellectual life, (4) identify what you might be forgetting (stalled items, sneaking deadlines, overlooked opportunities), (5) propose 2–4 concrete next actions. Output is a formatted briefing dated to the session.

Nick’s skill-map structure:

Producer Agents:

Personal Assistant
  Briefings: holistic-briefing, daily-briefing, shutdown-routine
  Email: email-review, drafts-inbox
  Files & Maintenance: file-organizer, image-grabber
  Meetings: meeting-notes
  Reviews: weekly-review

Chief of Staff
  chief-of-staff-scan, task-management

Synthesizer Agents:

Mapmaker
  maps-of-content-mocs

---

Full System View

---

Setup Workflow — Step by Step

Step 1 — Create the base folder structure (ACE)

Create three top-level folders inside your vault:

Ideaverse/
  Atlas/
  Calendar/
  Efforts/

Optional additions: + (inbox) and x (templates/extras).

Step 2 — Create the AIOS folder

Inside the vault, create:

AIOS/
  History/
  Maps/
  Skills/

This is where all AI-related maps and logs live. It’s separate from the ideaverse content — its own dedicated system folder.

Step 3 — Create me.md at the vault root

Create /me.md. Start with:

• A summary statement (who you are, what drives you)
• First principles and key beliefs
• How you think (frameworks, styles)
• Your roles
• Working preferences for AI (tone, format, what to skip, what to do)
• File path links to vault-map and skill-map

Write 50–80 lines. This is an evolving document but should feel mostly stable once the core is down.

Step 4 — Create CLAUDE.md (or equivalent) as a redirect

If using Claude Code or Claude CoWork, create CLAUDE.md at the vault root:

# CLAUDE

Go immediately to /me.md and you will find everything there.

This keeps the tool-specific file as a stable one-liner. All actual content stays in me.md.

Step 5 — Create AIOS/Maps/vault-map.md

Document your vault structure. Start with the brief overview (folder names + one-line descriptions), then expand. Add:

• A note on how AI should create files on your behalf (naming, frontmatter, template to use)
• The AIOS/History logging convention
• Any section on how to interact with other vaults or shared resources

This file will change more often than me.md — treat it as living documentation.

Step 6 — Create AIOS/Maps/skill-map.md + first skills

Start with 1–2 skills. Suggested starting points:

• How you want your daily note structured
• How you want articles or content summarized

Create each skill as its own .md file in AIOS/Skills/, then reference it from skill-map.md by file path.

Step 7 — Maintain as a garden

Every few weeks, review me.md and vault-map:

• Is anything outdated?
• Is it getting bloated? (Trim what’s no longer load-bearing)
• Are there new folder conventions or note types to document?

Skill files get added as new repeatable workflows emerge. There’s no need to build the full library up front.

---

Privacy and Defense

1. Keep personal content outside AI zones — journals, deeply personal notes, and sensitive material should stay in folders that AI doesn’t access. Be intentional about what you give AI access to.
2. Make backups — your ideaverse is your own; protect it with regular off-site backups.
3. Mark AI-generated content clearly — use a tag, emoji, or dedicated folder so you always know what’s yours vs. AI-assisted.

---

Mapping to the Baseworks KB — Observations

This section compares Nick’s model to the current Baseworks KB structure. These are observations, not decisions. No changes should be made until this plan is fully reviewed.

Nick’s Element	Baseworks KB — Current State	Gap / Consideration
ACE folders (Atlas / Calendar / Efforts)	00-inbox, 01-atlas, 02-areas, 03-resources, 04-archive	Different structure with numbered prefix system. Would need a mapping layer or full restructure.
me.md — portable identity	No equivalent	Identity and instructions are currently inside CLAUDE.md (Claude-specific) and ~/.claude/CLAUDE.md (global). Not portable to other tools.
vault-map.md — navigation manual	CLAUDE.md in this repo partially covers this	Currently merged with Claude-specific instructions. Not structured as a standalone navigation document.
CLAUDE.md as one-liner redirect	CLAUDE.md currently contains all instructions	Inverting this would be a significant rework — CLAUDE.md becomes a redirect, all content moves to me.md + vault-map.md.
AIOS/ folder	No equivalent	New addition — would house History, Maps, Skills.
skill-map.md + AIOS/Skills/	.claude/skills/ directory (Claude Code CLI skills)	Our skills are Claude Code-specific and not plain-text portable. skill-map would be a tool-agnostic complement.
AIOS/History/ — AI interaction logs	Partial — 00-inbox/ captures some AI output	No structured, searchable log of AI interactions.
Maps of Content (MOCs)	Scattered through 01-atlas/	MOCs exist but are not indexed for AI navigation. vault-map would give AI a structured entry point.
Multi-ideaverse translation layer	No equivalent	Shared-brain repo is a partial implementation — Asia and Patrick share the same vault. The model would formalize cross-vault interaction.

Key architectural insight: Nick’s CLAUDE.md is intentionally a one-liner. His actual instructions live in the portable, tool-agnostic me.md. The current Baseworks approach is the reverse: everything lives in CLAUDE.md. A migration to this model means:

1. Extract identity and working preferences → me.md
2. Extract vault navigation and conventions → vault-map.md
3. Reduce CLAUDE.md to a single redirect line
4. Create AIOS/ folder with History, Maps, Skills substructure
5. Optionally restructure folders toward ACE (or create a translation map that describes existing folders in ACE terms)

What this model would enable:

• Working with any AI (Gemini, local models, future tools) without reconfiguring from scratch
• Cleaner separation between “who Patrick is” and “how this vault works”
• Portable skills that could run in Claude Web, Claude CoWork, Notebook LM, or a local model
• Searchable AI interaction history in AIOS/History/
• A foundation for connecting Patrick’s vault to other vaults (Asia’s, a future team/client context)

---

Open Questions Before Any Implementation

1. ACE vs. current structure — How closely do we adapt ACE? We could either restructure toward ACE or write a vault-map that describes the current folder system in terms an AI can navigate without renaming anything.
2. Patrick-only vs. joint — Should me.md represent Patrick alone, or both Patrick and Asia? Asia has separate voice and workflow context. Likely separate me.md files make more sense.
3. Skills portability — The .claude/skills/ system works well for Claude Code CLI. A skill-map.md in AIOS would be complementary (portable), not a replacement. Do both coexist?
4. Migration risk — The current CLAUDE.md is load-bearing for every Claude Code session. Any migration should happen in a separate test branch and be validated before going to main.
5. Claude Code entry point — Claude Code reads CLAUDE.md at the project root. A one-liner redirect to me.md works exactly the same way Nick’s setup does — no tool-level changes needed.
6. Folder restructure vs. translation layer — A full ACE restructure is significant. An alternative: keep current folders, write a vault-map that maps them to ACE concepts so AI understands the structure without moving files.

---

Related

• AI OS Rework — Synthesis — cross-source comparison
• Simon Scrapes Agentic OS — companion source
• Plans Index
• Vault & Tooling — current conventions
• Source video: https://www.youtube.com/watch?v=jbHB-rzKBAs