AI Lab

한국어

Research workspace for analyzing hyped LLM tools and extracting actionable insights.

Purpose

Study trending LLM tools, frameworks, and patterns to:

1. Research - Deep-dive into concepts and implementations
2. Analyze - Compare approaches across different AI models (Claude, GPT, Gemini)
3. Extract - Create reusable artifacts (skills, prompts, patterns)

Workflow

flowchart LR
    subgraph Research["1. Research"]
        R1[docs/NN-topic/]
        R2[README.md]
        R3[meta-prompt.md]
    end

    subgraph Analyze["2. Analyze"]
        A1[01-claude.md]
        A2[02-gpt.md]
        A3[03-gemini.md]
    end

    subgraph Extract["3. Extract"]
        E1[.claude/skills/]
        E2[.claude/commands/]
        E3[prompts/]
    end

    Research --> Analyze --> Extract

Loading

Example: Structure Organizer

Recent research on "when to use Command vs Skill vs Agent":

Phase	Output
Research	docs/01-structure-organizer/
Multi-model analysis	Claude, GPT-4, Gemini perspectives
Extracted artifact	.claude/skills/meta-llm-type/

Directory Structure

ai-lab/
├── docs/                     # Research documents
│   └── 01-structure-organizer/
│       ├── README.md         # Topic overview
│       ├── meta-prompt.md    # Reusable prompt
│       └── research/         # Multi-model analysis
├── .claude/                  # Claude Code artifacts
│   ├── skills/               # Extracted skills
│   └── commands/             # Custom commands
├── refs/                     # Reference repositories (gitignored)
├── AGENTS.md                 # AI agent instructions
└── README.md

Research Topics

#	Topic	Status	Artifact
01	Structure Organizer	Done	Skill
02	Naming Convention	Done	Reference doc

Adding New Research

1. Create docs/NN-topic-name/
2. Add README.md with topic overview
3. Include meta-prompt.md if applicable
4. Run against multiple models, save to research/
5. Extract reusable artifacts to .claude/skills/ or elsewhere

For AI Agents

See AGENTS.md for workspace context and instructions.

License

MIT