Advanced Patterns

Power user territory. You've mastered the basics. These patterns unlock the next level.

Welcome to Advanced

You don't need these patterns to be effective. They're here when simple approaches aren't enough.

Prerequisites:

• Comfortable with Core Concepts
• Completed Common Workflows
• Successfully customized your setup
• Ready to experiment

Philosophy: Your brain should serve you, not the other way around. Use what helps. Skip what doesn't.

Pattern 1: Parallel Research Agents

Run multiple research approaches simultaneously for comprehensive coverage.

The Pattern

"Run these three agents in parallel:
1. perplexity-researcher: [topic A]
2. claude-researcher: [topic B]
3. gemini-researcher: [topic C]"

Or same topic, different agents:

"Research '[question]' using perplexity-researcher, claude-researcher, and gemini-researcher in parallel"

Why It Works

• 3x speed: Parallel execution vs sequential
• Diverse perspectives: Different AI models, different approaches
• Better synthesis: Compare and contrast multiple angles
• Reduces bias: One model's weakness is another's strength

When to Use

• Time-sensitive decisions: Need comprehensive research fast
• Controversial topics: Want multiple perspectives
• Comparing approaches: Different ways to solve same problem
• Deep research: Exhaustive coverage of complex topic

Example

# Researching AI agent frameworks
"Use these three research agents in parallel to investigate:
What are the best AI agent frameworks for business automation?

1. perplexity-researcher: Focus on current production frameworks
2. claude-researcher: Focus on technical implementation details
3. gemini-researcher: Focus on case studies and real-world usage"

Result: Three research reports in the time of one, each with different angle.

Pattern 2: Custom Research Pipelines

Automate recurring research workflows.

Create Command for Research Flow

Create .claude/commands/my-research-flow.md:

# My Weekly Research Flow

Run this flow every Monday morning.

## Steps

1. Check newsletters
2. Generate summaries for new content
3. Update all indexes in context/ideas/
4. Run brain-advisor on standing questions:
   - "What's new in AI this week?"
   - "Any pricing or positioning insights?"
   - "Emerging tools or techniques?"
5. Compile findings into weekly-insights.md
6. Create draft Circle post with highlights

## Output

Save to: content/weekly-briefs/YYYY-MM-DD-brief.md
Format: Ready to share with community

Invoke

/my-research-flow

Level Up: Schedule It

Add to cron (Mac/Linux):

# Edit crontab
crontab -e

# Add line (runs every Monday at 9am)
0 9 * * 1 cd [path-to-your-brain-folder] && claude "/my-research-flow"

Result: Automated weekly research synthesis. No manual work.

When to Use

• Recurring research tasks
• Weekly/monthly reviews
• Content creation workflows
• Status reporting

Pattern 3: Cross-Project Memory

Share context across multiple projects.

The Setup

You have multiple projects that need access to your brain's research:

~/Projects/
  ├── your-brain/         # Your research brain
  ├── client-acme/        # Client project
  ├── personal-site/      # Personal project
  └── startup-idea/       # Side project

Make Brain Available to Other Projects

In client project's CLAUDE.md:

# Client Project - Acme Corp

## Access Brain Research

When answering strategic questions, you can search:
[path-to-your-brain-folder]/context/ideas/

**Available indexes:**
- ethan.md - AI and work transformation
- ben.md - Tech business strategy
- avinash.md - Marketing analytics
- [list your sources]

**How to use:**
When client asks strategic question, search brain indexes for relevant insights.

Example: "What do industry experts say about AI adoption?"
→ Search [path-to-your-brain-folder]/context/ideas/ for AI adoption insights
→ Cite sources from brain

Why It Works

• One brain, many projects: Research investment pays off everywhere
• Consistent expertise: Same knowledge base across all work
• Credibility: Cite authoritative sources in client work
• Efficiency: Don't re-research what you've already captured

When to Use

• Client consulting projects
• Multiple product ideas
• Personal and work projects
• Teaching or content creation

Pattern 4: Custom MCP Servers

Build your own tool integrations.

Why Build Custom MCP

Connect your brain to:

• Your CRM (pull client context)
• Your analytics (get performance data)
• Your Notion (query workspace)
• Your Airtable (sync research)
• Any API you use regularly

Basic MCP Server Template

# my_custom_server.py
from mcp import Server, Tool

server = Server("my-custom-server")

@server.tool()
async def query_my_system(query: str) -> dict:
    """
    Query my custom system.

    Args:
        query: What to search for

    Returns:
        Results from my system
    """
    # Your implementation
    result = await my_system_api.search(query)
    return {"results": result}

if __name__ == "__main__":
    server.run()

Add to MCP Config

Edit ~/.claude/mcp_settings.json:

{
  "mcpServers": {
    "my-custom-server": {
      "command": "python",
      "args": ["/path/to/my_custom_server.py"]
    }
  }
}

Use in Skills

Create skill that uses your MCP server:

# My Custom Data Skill

Auto-activates when: User asks about data from my system

**Tools:** mcp__my-custom-server__query_my_system

When activated:
1. Parse user question
2. Query my system via MCP
3. Combine with brain research if relevant
4. Return synthesized answer

When to Use

• You have a system you query frequently
• Want brain to access proprietary data
• Need integration not available yet
• Building custom workflow automation

Resources

• MCP documentation: [Link to MCP docs]
• Example servers: Check context/info-and-docs/mcp-guide.md in your brain folder
• Community examples in Circle

Pattern 5: Agent Chains

Chain multiple agents for complex workflows.

The Pattern

Sequential processing through specialized agents:

"Use brain-advisor to research [topic],
then use posty to create a LinkedIn post from the insights,
then save to content/posts/"

Example Chain: Content Creation

# Step 1: Research (brain-advisor)
"Use brain-advisor: What do I know about AI automation ROI?"

# Step 2: Structure (you or another agent)
"Create 3 key points from these insights"

# Step 3: Draft (posty)
"Use posty to create LinkedIn post about AI automation ROI using these points"

# Step 4: File (inboxy or manual)
"Save this to content/posts/linkedin-YYYYMMDD-ai-roi.md"

Example Chain: Competitive Analysis

# Research phase
"Use brain-advisor: Everything I know about [competitor]"
"Use perplexity-researcher: Latest news about [competitor]"

# Analysis phase
"Compare my research insights with latest news. What's changed?"

# Report phase
"Create competitive brief combining historical research and current intel"

When to Use

• Multi-step workflows
• Need different specialist perspectives
• Combining research and creation
• Building comprehensive reports

Pro Tip: Document Your Chains

Save successful chains as commands:

# .claude/commands/competitive-brief.md

Create competitive intelligence brief:

1. brain-advisor: Historical intelligence from research
2. perplexity-researcher: Current news and updates
3. Synthesis: Compare and identify changes
4. Report: Structured competitive brief

Args: Competitor name

Pattern 6: Scheduled Automations

Set up cron jobs for recurring brain tasks.

Daily Newsletter Fetch

# Edit crontab
crontab -e

# Add: Fetch newsletters every day at 9am
0 9 * * * cd [path-to-your-brain-folder] && claude "check newsletters"

Weekly Research Summary

# Every Monday at 8am
0 8 * * 1 cd [path-to-your-brain-folder] && claude "generate weekly insights"

Daily Backup

# Every day at midnight
0 0 * * * cd [path-to-your-brain-folder] && git add . && git commit -m "Daily backup $(date +%Y-%m-%d)" && git push

Monthly Archive

# First day of month at 1am
0 1 1 * * cd [path-to-your-brain-folder] && /path/to/archive-old-research.sh

When to Use

• Regular maintenance tasks
• Content creation schedules
• Backup routines
• Archive management

Pro Tip: Test First

Before scheduling:

# Test the exact command you'll schedule
cd [path-to-your-brain-folder] && claude "check newsletters"

# Verify it works
# Then add to crontab

Pattern 7: Brain as API

Use your brain from other applications.

Python Example

# brain_api.py
import subprocess
import json

def ask_brain(question: str) -> str:
    """Query your brain from Python."""
    brain_path = "/path/to/your/brain/folder"  # Update this path
    cmd = f'cd {brain_path} && claude "{question}"'
    result = subprocess.run(cmd, shell=True, capture_output=True)
    return result.stdout.decode()

def research(topic: str) -> dict:
    """Get structured research from brain."""
    question = f"Use brain-advisor: What do I know about {topic}?"
    result = ask_brain(question)

    return {
        "topic": topic,
        "insights": result,
        "timestamp": datetime.now().isoformat()
    }

# Use it
insight = research("AI pricing models")
print(insight)

JavaScript Example

// brain-api.js
const { exec } = require('child_process');
const util = require('util');
const execPromise = util.promisify(exec);

async function askBrain(question) {
  const brainPath = '/path/to/your/brain/folder';  // Update this path
  const cmd = `cd ${brainPath} && claude "${question}"`;
  const { stdout } = await execPromise(cmd);
  return stdout;
}

async function research(topic) {
  const question = `Use brain-advisor: What do I know about ${topic}?`;
  const result = await askBrain(question);

  return {
    topic,
    insights: result,
    timestamp: new Date().toISOString()
  };
}

// Use it
research('AI pricing models').then(console.log);

Use Cases

Dashboard showing research insights:

# Auto-refresh dashboard with latest insights
insights = ask_brain("What themes emerged this week?")
dashboard.update(insights)

Automated reporting:

# Weekly report generation
weekly_insights = ask_brain("Generate weekly research summary")
send_email(to=team, subject="Weekly Insights", body=weekly_insights)

Integration with other tools:

# Enrich CRM with research insights
for client in clients:
    context = ask_brain(f"What do I know about {client.industry}?")
    client.update(research_context=context)

When to Use

• Building tools that need strategic input
• Automated reporting
• Integration with existing systems
• Custom dashboards

Pattern 8: Multi-Brain Architecture

Run separate brains for different contexts.

The Setup

~/Projects/
  ├── personal-brain/     # Personal research
  ├── work-brain/         # Client work
  ├── startup-brain/      # Your company
  └── research-brain/     # Deep technical research

Each brain:

• Own research sources
• Own configuration
• Own memory
• Own agents/skills
• Own focus

Why Multiple Brains

Separation of contexts:

• Personal vs professional
• Client A vs Client B
• Teaching vs building
• Experimental vs production

Focused research:

• Personal: Life, family, hobbies
• Work: Industry, clients, strategy
• Startup: Market, competitors, product
• Research: Deep technical topics

Different configurations:

• Personal: Casual tone, life-focused
• Work: Professional, client-focused
• Startup: Speed-focused, experiment-friendly
• Research: Depth-focused, comprehensive

Switch Contexts

# Switch to work context
cd ~/Projects/work-brain

# Switch to personal
cd ~/Projects/personal-brain

# Switch to startup
cd ~/Projects/startup-brain

Each has independent:

• Research sources
• Configurations
• History
• Customizations

When to Use

• Clear separation between contexts
• Different research sources per context
• Different output requirements
• Privacy or confidentiality needs

Pro Tip: Shared Components

Some things can be shared:

# Link shared agents (adjust paths to your brain folders)
ln -s ~/Projects/personal-brain/.claude/agents/brain-advisor ~/Projects/work-brain/.claude/agents/

# Link shared skills
ln -s ~/Projects/personal-brain/.claude/skills/posty ~/Projects/work-brain/.claude/skills/

Reuse what works, customize what differs.

Pattern 9: Research Quality Scoring

Automatically filter low-value content.

Create Quality Rating Skill

.claude/skills/rate-research/skill.md:

# Research Quality Rater

**Auto-activates when:** New newsletter or video is saved to research folder

**Purpose:** Rate quality and relevance of new research content

## Scoring Criteria

Rate each item 1-5 on:

1. **Actionability** - Can I act on these insights?
2. **Signal vs Noise** - Insight density vs fluff
3. **Relevance** - Relates to my focus areas
4. **Uniqueness** - New perspective or rehashed?

**Total score:** Sum of above (max 20)

## Actions

- **Score 16-20:** Keep, high-priority index update
- **Score 11-15:** Keep, standard indexing
- **Score 6-10:** Review - might archive
- **Score 1-5:** Archive immediately

## Process

1. Read new content
2. Score using criteria
3. Add score to file metadata
4. If score < 11, suggest archiving
5. If score > 15, flag for review

## Example

Content: Newsletter about AI trends Actionability: 4 (specific tactics) Signal/Noise: 5 (dense insights) Relevance: 5 (directly relevant) Uniqueness: 4 (fresh perspective) Total: 18

Action: Keep, prioritize in next review

Result

Your brain contains only high-quality insights. No noise.

When to Use

• Drowning in content
• Quality declining
• Need to be more selective
• Building premium knowledge base

Pattern 10: Collaboration Mode

Share brain insights with team or clients.

Export for Sharing

# Generate shareable research summary
"Create a research summary about [topic] that I can share with my team.
Include key insights, source citations, and tactical recommendations.
Format for readability."

Multi-Format Export

"Export this research as:
1. PDF report (formatted, professional)
2. Markdown (for Notion/GitHub)
3. Circle post (community format)
4. Slide deck outline (presentation ready)"

Protect Your Brain

Don't share:

• Raw brain files
• Your full research archive
• Personal configurations
• Custom agents/skills

Do share:

• Specific insights
• Synthesized summaries
• Curated recommendations
• Cited conclusions

Think of it as:

• Brain = Your private research library
• Exports = Public papers from that library

Team Knowledge Base

Create shared knowledge from your brain:

# Generate team knowledge base entry
"Create a knowledge base article about [topic] using my research.
Target audience: My team
Format: [Tool you use - Notion, Confluence, etc.]
Include: Context, key insights, recommendations, further reading"

Client Deliverables

Use brain for client work:

# Research phase
"Use brain-advisor: Everything I know about [client industry]"

# Generate deliverable
"Create a strategic brief for [client] about [topic] based on my research.
Include: Industry context, strategic options, recommendations
Tone: Professional, evidence-based
Citations: Include sources"

When to Use

• Working with teams
• Client deliverables
• Teaching or training
• Building public content

Power User Principles

1. Automate Repetition

If you do it weekly, make it a command. If you do it daily, schedule it. If you do it more than 3 times, script it.

Time saved compounds.

2. Chain Specialists

Combine agents for complex workflows. Each agent does what it's best at. Chain creates power beyond sum of parts.

Specialization wins.

3. Context Switching

Multiple brains for multiple contexts. Each focused and optimized. Switch contexts, don't mix them.

Clarity over convenience.

4. Quality Gates

Filter noise before it enters your brain. Rate content, archive low value. Your brain's value = quality of inputs.

Garbage in, garbage out.

5. Share Selectively

Export insights, don't expose raw brain. Your brain is competitive advantage. Share conclusions, protect process.

Insights are public, brain is private.

6. Backup Everything

Git commit daily. Push weekly minimum. Keep archives of old research.

Data loss is devastating.

7. Iterate Fast

Try patterns, keep what works. Discard what doesn't. Build for yourself, not imagined future.

Real use beats perfect design.

Going Deeper

Want to Build Custom Agents?

Read: context/info-and-docs/agents-skills-plugins-commands.md

Understand agent architecture, then build your own specialists.

Want to Create Skills?

See: .claude/skills/create-skill/skill.md

Template and guide for building auto-activating expertise.

Want MCP Servers?

Check: context/info-and-docs/mcp-guide.md

Connect your brain to external systems.

Want to Contribute?

Best patterns end up in:

• Community pattern library
• Shared agent repository
• Template collections
• Official docs (like this!)

Share what works. Learn from others.

Community Power Users

See how others push boundaries:

• Circle Showcase Posts - Members share advanced setups
• Community Patterns Repo - Shared workflows and configs
• Agent Library - Pre-built specialists to download
• Weekly Office Hours - Learn from power users

Not building alone. We're all figuring this out together.

Final Note

Advanced doesn't mean better.

Use these patterns when simple ones aren't enough.

Your brain should serve you, not the other way around.

The best pattern is the one you'll actually use.

Start simple. Add complexity only when you need it. Most productivity comes from basics done well, not advanced patterns done poorly.

Questions? Ask in Circle - power users love helping.

Share your patterns - what you discover might help others.

Keep building - the best advanced patterns haven't been invented yet.

Quick Reference

Advanced patterns:

1. Parallel research (3x speed)
2. Custom pipelines (automation)
3. Cross-project memory (share context)
4. Custom MCP servers (integrations)
5. Agent chains (complex workflows)
6. Scheduled automations (cron)
7. Brain as API (programmatic access)
8. Multi-brain (separate contexts)
9. Quality scoring (filter noise)
10. Collaboration mode (safe sharing)

Power user principles:

• Automate repetition
• Chain specialists
• Context switching
• Quality gates
• Share selectively
• Backup everything
• Iterate fast

Resources:

• Agent guide: context/info-and-docs/agents-skills-plugins-commands.md
• Skill template: .claude/skills/create-skill/skill.md
• MCP guide: context/info-and-docs/mcp-guide.md
• Community: Circle showcase and patterns repo

Remember:

• Advanced ≠ Better
• Use what helps
• Skip what doesn't
• Simple done well beats complex done poorly