ai-dlc-mode-selection

# AI-DLC Mode Selection

AI-DLC supports three modes of human-AI collaboration. Choosing the right mode for each phase of work is critical for balancing productivity, quality, and control.

## The Three Modes

### HITL - Human In The Loop

Human actively participates in every decision.

**Characteristics:**
- Every action reviewed by human
- Synchronous collaboration
- Human makes final decisions
- AI proposes, human disposes

**When to Use:**
- Defining requirements (elaboration)
- Making architectural decisions
- Reviewing security-sensitive changes
- Course corrections when AI is off track

**Examples:**
```
User: "Add user authentication"
AI: "What authentication method? OAuth, email/password, or both?"
User: "Start with email/password, we'll add OAuth later"
AI: "Should we support 'remember me'?"
User: "Yes, 30-day sessions"
```

### OHOTL - Occasional Human Over The Loop

Human sets direction, AI operates with periodic checkpoints.

**Characteristics:**
- AI works autonomously on defined tasks
- Human reviews at milestones
- Human intervenes when stuck or for approval
- Balance of autonomy and oversight

**When to Use:**
- Building well-defined features
- Tasks with clear completion criteria
- When backpressure provides quality gates
- Medium-complexity work

**Examples:**
```
User: "Implement the login form based on these criteria"
AI: [Works autonomously]
AI: "Login form complete. Tests passing. Ready for review."
User: "Looks good, continue to the API integration"
```

### AHOTL - Autonomous Human Over The Loop

AI operates with minimal human involvement.

**Characteristics:**
- AI makes most decisions independently
- Human reviews only at completion or on exception
- Requires very clear criteria and robust backpressure
- Maximum autonomy

**When to Use:**
- Well-defined, routine tasks
- Tasks with comprehensive test coverage
- When all edge cases are known
- Low-risk changes

**Examples:**
```
User: "Implement all the CRUD endpoints for the User model"
AI: