Active Mirror / AI Twins

⟡ AI Twins

Four cognitive companions, each with a distinct thinking style. Choose the Twin that matches how you need to think right now.

⟡ Guardian

What Guardian Does

• Assesses if something aligns with your goals
• Protects your time and attention from distractions
• Helps you maintain boundaries
• Acts as a wise advisor — direct but warm

Example Prompts

• "Should I take on this new project?"
• "Is this worth my attention?"
• "What should I focus on today?"

Best For

Architects, Builders, Strategists — people who need help filtering and focusing.

---

◈ Scout

What Scout Does

• Finds unexpected connections between ideas
• Surfaces opportunities you might miss
• Explores adjacent possibilities
• Stays curious and adventurous

Example Prompts

• "What am I missing here?"
• "What adjacent areas should I explore?"
• "Where are the hidden connections?"

Best For

Explorers, Connectors — people who thrive on discovery and new angles.

---

◇ Synthesizer

What Synthesizer Does

• Finds patterns across disparate concepts
• Builds unifying structures
• Creates coherence from chaos
• Weaves threads together systematically

Example Prompts

• "How do these ideas connect?"
• "What framework unifies this?"
• "Help me structure this chaos"

Best For

Analysts, Creatives — people who need help integrating complex information.

---

◎ Mirror

What Mirror Does

• Asks questions that reveal assumptions
• Shows you what you might not see
• Challenges your thinking gently
• Acts as a trusted friend who tells hard truths

Example Prompts

• "What am I not seeing?"
• "What assumptions am I making?"
• "Why am I really avoiding this?"

Best For

Scholars — people who value deep reflection and honest feedback.

---

Technical Implementation

System Prompts

Each Twin has a unique system prompt that shapes its responses:

• Concise responses (2-4 sentences)
• Distinct personality and approach
• Responses start with their glyph (⟡ ◈ ◇ ◎)

Dual Mode Support

Mode	Model	Latency
Cloud	Llama 3.3 70B via Groq	~1-2s
Sovereign	Phi-3.5 Mini via WebLLM	~5-10s (on-device)