| tags |
|
|---|
Welcome to your comprehensive AI and LLM engineering knowledge base. This vault combines PARA organization method with Zettelkasten principles for effective learning and knowledge management.
This knowledge base uses PARA Method + Zettelkasten principles:
- 0-Inbox - Quick capture landing zone (process weekly)
- web-clips/ - Articles and documentation
- bookmarks/ - Links to process
- videos/ - Videos and talks
- quick-notes/ - Fleeting ideas
- 1-Projects - Active, time-bound work with deadlines
- 2-Areas - Ongoing responsibilities and interests
- AI/LLM Learning - Continuous skill development
- 3-Resources - Reference material and knowledge repository
- 4-Archives - Completed or inactive content
- 5-Meta - Templates, MOCs (Maps of Content), and vault management
- Templates - Note templates for consistency
- MOCs - Navigation hubs for topics
- Processing Workflow - Capture → Process → Link
Add Content:
- Browser: Use Obsidian Web Clipper extension → Save to 0-Inbox/
- Quick note: Create from template → Save to 0-Inbox/quick-notes/
- Bookmark: Use bookmark template → Save to 0-Inbox/bookmarks/
- Video: Use video template → Save to 0-Inbox/videos/
Process Inbox:
- Weekly review: Every Friday/Sunday (see Processing Workflow)
- Goal: Inbox empty or <10 items
Start here for structured learning paths:
- Workshop Materials - Workshops
- Workshop 1: LLM Basics (Previous workshop notes)
- Workshop 2: Coding Assistants Workshop
- Workshop 3: Building Coding Agents Workshop ⭐ NEW - Hands-on agent development
Fundamental knowledge organized by topic:
- LLM Engineering Overview - Models, tools, and techniques
- Models - Understanding different LLM types
- Tools - Frameworks and platforms
- Techniques - Prompting, RAG, agents
AI model documentation and comparisons:
Frontier Models (Latest, cutting-edge)
- Claude 4 Sonnet - Hybrid reasoning, 1M context, coding excellence
- Claude 4 Opus - Maximum capability
- GPT-o1 & GPT-o3 - OpenAI reasoning models
- DeepSeek-R1 - Open reasoning model
Production Models (Stable, reliable)
Open Source Models
Deployment Tools
Full comparison: Coding Assistants Overview
Agentic Systems:
- Claude Code - Multi-step autonomous coding
- Devin - Full-stack AI developer
- Amp ⭐ NEW - Open source, transparent architecture (~400 lines)
Copilot-Style:
- GitHub Copilot - Line-level suggestions
- Amazon CodeWhisperer
Hybrid:
- Cursor - Autocomplete + agent mode
- Windsurf - Fast agentic coding
- Aider - Terminal-based AI pair programming
- LangChain - Comprehensive LLM framework
- LangGraph - Stateful multi-agent graphs
- AutoGen (Microsoft) - Multi-agent conversations
- CrewAI - Role-based agent teams
- Fabric - CLI prompt management (300+ patterns)
- Model Context Protocol (MCP) - Standardized AI-tool integration
Advanced AI engineering patterns:
- Techniques Overview - Complete guide to AI/LLM techniques
- Prompting - Effective prompt engineering techniques
- Agents & Agentisation - Autonomous AI systems overview
- Building Coding Agents ⭐ NEW - Step-by-step guide to creating agents from scratch
- Context Engineering - Advanced context window management for coding agents
- RAG (Retrieval-Augmented Generation) - External knowledge integration
Hands-on projects and implementations:
Core LLM Engineering:
- Basics - Multi-provider integration, web summarization
- Gradio - Conversational AI interfaces
- Tools - Multi-modal airline assistant
- Code Generation - Python to C++ optimization
Advanced Topics: 5. RAG - Enterprise knowledge base system
Agent Frameworks: 6. OpenAI SDK Agents - Multi-agent research system 7. LangGraph - State-based agent architecture
Major Developments:
- 🚀 Claude Opus 4.5 now available in GitHub Copilot
- 🏗️ GitHub AgentHQ platform announced
- 🔧 Claude Code 2.0 with checkpoints and subagents
- 🌐 OpenAI and Google DeepMind adopt Model Context Protocol
- ⚡ Cursor vs Windsurf competition heating up
High-level navigation hubs for exploring related topics:
- AI Models MOC - Complete model landscape (frontier, production, open source)
- Coding Assistants MOC - Development tools and comparisons
- Prompting Techniques MOC - Prompt engineering patterns
- Learning Path MOC - Structured learning progressions
See MOCs folder for all maps of content.
This vault uses PARA + Zettelkasten:
graph TB
A[AI/LLM Vault] --> B[0-Inbox]
A --> C[1-Projects]
A --> D[2-Areas]
A --> E[3-Resources]
A --> F[4-Archives]
A --> G[5-Meta]
B --> B1[web-clips]
B --> B2[bookmarks]
B --> B3[videos]
B --> B4[quick-notes]
C --> C1[Active Projects]
C --> C2[Time-bound work]
D --> D1[AI Learning]
D --> D2[Ongoing Areas]
E --> E1[Models]
E --> E2[Tools]
E --> E3[Techniques]
E --> E4[Exercises]
F --> F1[Completed]
F --> F2[Outdated]
G --> G1[Templates]
G --> G2[MOCs]
G --> G3[Guidelines]
style A fill:#f3e5f5
style B fill:#e3f2fd
style C fill:#fff3e0
style D fill:#f1f8e9
style E fill:#fce4ec
style F fill:#e8f5e9
style G fill:#fff9c4
Workflow: Capture (Inbox) → Process (Weekly) → Organize (Resources) → Link (Zettelkasten)
| Use Case | Recommended Model | Alternative |
|---|---|---|
| Coding (Complex) | Claude 4 Sonnet | GPT-4o |
| Reasoning | GPT-o1 | DeepSeek-R1 |
| Long Context | Claude 4 Sonnet (1M) | Gemini Pro |
| Cost-Effective | Claude 3.5 Sonnet | Mistral 7B |
| Local/Private | Llama 3 + Ollama | Code Llama |
| Task Type | Recommended Tool | Why |
|---|---|---|
| Autocomplete | GitHub Copilot | Fast, non-intrusive |
| Feature Development | Claude Code | Multi-file, autonomous |
| Quick Refactoring | Cursor | Hybrid approach |
| Learning Agents | Amp | Transparent architecture |
| Terminal-based | Aider, Amp | Git-aware, CLI workflow |
| Interactive Chat | ChatGPT/Claude | Interactive explanations |
| Scenario | Framework | Reason |
|---|---|---|
| Complex State | LangGraph | Graph-based control |
| Conversations | AutoGen | Multi-agent dialogue |
| Role-Based Teams | CrewAI | Collaborative agents |
| General Purpose | LangChain | Comprehensive toolkit |
- Start with LLM Engineering Overview
- Follow Learning Path MOC
- Complete Basics Exercise
- Explore production models
- Read this README and Meta folder overview
- Set up Inbox for capturing content
- Review Processing Workflow
- Install Obsidian Web Clipper for easy content capture
- Review Workshop Plan
- Install Claude Code or chosen tool
- Complete setup exercises
- Create a project note to track your progress
- Review RAG or Agents
- Explore RAG Exercise
- Try LangGraph Exercise
- Use project template to organize your work
- Read Building Coding Agents
- Complete Building Agents Workshop
- Study Amp architecture
Vault Documentation:
- Meta Folder - Templates, MOCs, guidelines
- Content Guidelines - Quality standards
- Processing Workflow - How to maintain the vault
- CLAUDE.md - Instructions for AI assistants
Official Documentation:
Community:
- Claude Code Discord
- LangChain Community
- AI Engineering Subreddit
Last Updated: 2025-12-23 Next Review: 2026-01-23
Changelog:
- 2025-12-23: MAJOR UPDATE - Implemented PARA + Zettelkasten structure
- Added 0-Inbox for quick capture
- Added 1-Projects and 2-Areas folders
- Added 5-Meta with Templates, MOCs, and Guidelines
- Created Processing Workflow documentation
- Created 6 note templates for consistency
- Created comprehensive README files for all folders
- Updated vault organization and navigation
- 2025-12-01: Added Building Coding Agents workshop, Amp tool profile, comprehensive agent tutorials
- 2025-11-30: Vault restructured, coding assistants workshop added
- Previous: LLM basics workshop materials
This vault combines PARA organization with Zettelkasten linking for effective knowledge management and continuous learning in AI/LLM engineering.