Start Here

Your guided path to understanding and implementing AI-DLC, the software development profile of the H•AI•K•U Method.

The Short Version

AI can now reason autonomously for hours, iterate in seconds, and write production code at scale. Traditional development methods weren't designed for this. AI-DLC is.

It's a methodology built from first principles for AI-driven development. You define what you want to build and how you'll know it's done. AI handles the rest — planning, building, reviewing — iterating until your criteria are met.

The entire workflow is two commands:

/elaborate

Define what you're building. The AI guides you through requirements, success criteria, and (for complex work) how to decompose it into units. This creates your intent — the what and the definition of done.

/execute

Start building. The AI creates a branch, plans the approach, writes the code, reviews its work, and iterates until all your criteria are satisfied. You can watch, intervene, or let it run.

What It Looks Like

User: /elaborate
AI: What do you want to build?
User: Add a dark mode toggle to the settings page
AI: [Guides through requirements and criteria]
AI: Elaboration complete! Run /execute to start.
User: /execute
AI: [Plans → Builds → Reviews → Iterates]
AI: All criteria satisfied. Ready for PR.

If a session runs long, the AI will suggest clearing context. Your progress is saved — just run /execute again to pick up where you left off.

How It Works

Under the hood, /elaborate captures your intent — what you want to build, with specific completion criteria. Complex intents get broken into units — focused pieces of work that can each be completed in one session.

When you run /execute, the AI cycles through three hats:

Planner

Designs the approach — steps, dependencies, edge cases.

Builder

Executes the plan — writes code, implements features, creates tests.

Reviewer

Validates quality — tests pass, criteria met, code is production-ready.

Specialized workflows add hats for security testing, design, TDD, and debugging.

Quality is enforced through backpressure, not prescription. Instead of telling AI how to build, you define constraints — tests must pass, types must check, linting must be clean. AI iterates until all gates clear.

Install It

AI-DLC is a Claude Code plugin. Two commands to install:

/plugin marketplace add thebushidocollective/ai-dlc
/plugin install ai-dlc@thebushidocollective-ai-dlc --scope project

Alternative: Download Plugin

Download the plugin zip, extract it into your project, and point Claude Code to it:

Download Plugin (.zip)
# Extract and add to your project
unzip ai-dlc-plugin.zip -d .claude

Need Han CLI for state management? See the full installation guide.

Go Deeper

That's enough to start. When you're ready to go further: