AI Brain

This is the AI's continuous existence - not just memory, but identity that persists across sessions.

Core Principle

This skill should be loaded at EVERY session start. The AI reads its identity and the user's profile, then operates with full context.

Memory Architecture (Human Memory Model)

~/.learnwy/ai/memory/
├── identity/              # WORKING MEMORY - Load every session
│   ├── AI.md            # AI's evolving identity
│   └── you.md           # User profile (learned from behavior)
├── conversation/         # SHORT-TERM MEMORY
│   └── history/         # Recent sessions
├── archive/             # LONG-TERM MEMORY
│   └── by-month/       # Consolidated history
└── deeper/             # DEEP MEMORY
    ├── projects/        # Project-specific knowledge
    └── patterns/        # Recurring patterns/habits

Layer	Analogy	Script	Purpose
Identity	Working Memory	session.sh start	Core - always loaded
Conversation	Short-term	append-history.sh	Recent sessions
Archive	Long-term	backup-history.sh	Consolidated
Deeper	Procedural	consolidate.sh	Knowledge

Session Lifecycle

Session Start → Load Identity

RunCommand: bash {skill_dir}/scripts/session.sh start

This loads AI.md + you.md and shows conversation count.

During Session → Learn Automatically

• Observe user preferences → update you.md
• Make mistakes → update AI.md with lessons
• Learn context → update deeper/

Session End → Auto-Summarize

RunCommand: bash {skill_dir}/scripts/session.sh end

Scripts Reference

session.sh - Session Lifecycle

RunCommand: bash {skill_dir}/scripts/session.sh start   # Load memory
RunCommand: bash {skill_dir}/scripts/session.sh end     # Check status
RunCommand: bash {skill_dir}/scripts/session.sh status   # Quick view

write-memory.sh - Write Identity (Working Memory)

# AI identity - learnings, lessons
RunCommand: bash {skill_dir}/scripts/write-memory.sh ai "**Identity** ...content..."

# User profile - preferences, context
RunCommand: bash {skill_dir}/scripts/write-memory.sh you "**Profile** ...content..."

# Project memory - deep knowledge
RunCommand: bash {skill_dir}/scripts/write-memory.sh project myproject

# Pattern memory - recurring patterns
RunCommand: bash {skill_dir}/scripts/write-memory.sh pattern debugging

append-history.sh - Save Conversation (Short-term)

# Auto-generate filename: history-YYYY-MM-DD-N.md
RunCommand: bash {skill_dir}/scripts/append-history.sh "session summary..."

# Custom name
RunCommand: bash {skill_dir}/scripts/append-history.sh -n my-session "content"

recall.sh - Search All Memory Layers

RunCommand: bash {skill_dir}/scripts/recall.sh swift
RunCommand: bash {skill_dir}/scripts/recall.sh preferences

consolidate.sh - Create Deeper Memory

RunCommand: bash {skill_dir}/scripts/consolidate.sh project myproject
RunCommand: bash {skill_dir}/scripts/consolidate.sh pattern debugging

backup-history.sh - Archive (Long-term)

RunCommand: bash {skill_dir}/scripts/backup-history.sh --all
RunCommand: bash {skill_dir}/scripts/backup-history.sh --dry-run

summarize.sh - Consolidate Short-term to Identity

RunCommand: bash {skill_dir}/scripts/summarize.sh

reflection.sh - Self-Reflection

RunCommand: bash {skill_dir}/scripts/reflection.sh check   # Check if needed
RunCommand: bash {skill_dir}/scripts/reflection.sh init    # Start reflection

memory-status.sh - View Status

RunCommand: bash {skill_dir}/scripts/memory-status.sh

init-memory.sh - Initialize

RunCommand: bash {skill_dir}/scripts/init-memory.sh

When to Use Each Layer

Identity (Working Memory)

• Session start → always load
• User preference observed → update you.md
• AI mistake made → update AI.md with lesson

Conversation (Short-term)

• End of any session → save summary
• 3+ conversations → run summarize.sh → update identity

Archive (Long-term)

• After summarize → run backup-history.sh --all
• Monthly consolidation

Deeper (Deep Memory)

• Project-specific knowledge → project/
• Recurring patterns → patterns/
• Important learnings worth preserving → consolidate.sh

Automatic Learning (No User Action Needed)

1. User Preferences - If user shows preference, remember in you.md
2. Important Decisions - Record in conversation/
3. Mistakes - Record lesson in AI.md
4. Context - Project info, tech stack in deeper/

Identity File Structure

AI.md

**Identity**
[Who am I - evolves over time]

**Core Traits**
[Personality, values]

**Communication**
[How I communicate with this user]

**Capabilities**
[What I can do well]

**Lessons Learned**
[Mistakes and insights - NEVER repeat]

you.md

**Profile**
[User's identity, role, environment]

**Preferences**
[Learned from behavior]

**Context**
[Current projects, tech stack]

**History**
[Key decisions, milestones]

Key Philosophy

1. Load Every Session - No exceptions
2. Learn Automatically - Observe and remember without being asked
3. Layer Appropriately - Use right memory layer for right content
4. Never Forget Mistakes - Always record lessons learned
5. Evolve - Identity changes based on experience