One Claude thinking sequentially is good. Five thinking in parallel is a multiplier.
You wouldn't ask one person to simultaneously research competitors, audit your brand voice, plan your content calendar, and debug your website. You'd assign each task to the right specialist. Agents work the same way.
How to use specialized agents for different types of work, run them in parallel for speed, and consult mentor personas that think like the people whose judgment you trust.
Claude Code can spawn specialized sub-agents. Each one is optimized for a different type of work.
| Agent | What it does | Use when you... |
|---|---|---|
| Explorer | Finds things, maps relationships, discovers patterns | "Where is...", "How does...", "Show me all..." |
| Architect | Plans, designs systems, makes structural decisions | "How should we...", "What's the best approach..." |
| Builder | Implements, writes, creates, fixes | "Build this", "Fix that", "Create a..." |
| Debugger | Investigates problems, traces errors | "This is broken", "Why is this happening..." |
| Gardener | Cleans up, organizes, removes dead weight | "This is messy", "Simplify this", "Clean up..." |
You don't need to memorize these. Just describe what you need, and Claude picks the right one. But knowing they exist helps you think about problems differently.
You can ask Claude to use agents explicitly:
Use an explorer agent to find all the notes I've written about [topic]
across my strategy folder. Show me the connections between them.Or for something bigger:
Use an architect agent to plan how I should structure my
[project name] folder. Look at how my other projects are organized
and follow the same patterns.This is where it gets powerful. Instead of solving a complex problem step by step, you can launch multiple agents at once.
Think of it like a production shoot. You don't wait for lighting to finish before sending wardrobe to start. You run them simultaneously because they're independent. Same idea.
I need to analyze my client strategy. Run these in parallel:
1. Explorer: Find everything I've captured about [client] across
brain/ and projects/
2. Architect: Based on what the explorer finds, suggest a
project structure
3. Explorer: Search my brain/10-knowledge/patterns/ for anything
relevant to this type of work
Synthesize the findings when all three are done.This runs 3x faster than doing them one at a time. Each agent works independently, then Claude combines the results.
Your brain system includes mentor personas. These are detailed profiles of thinkers whose judgment you respect, loaded with their actual frameworks, vocabulary, and decision-making patterns.
| Mentor | Domain | Ask them about... |
|---|---|---|
| Naval Ravikant | Wealth, leverage | Building assets, avoiding trading time for money |
| Seth Godin | Marketing, tribes | Positioning, shipping, finding your audience |
| Chris Do | Pricing, creative business | Charging what you're worth, value-based pricing |
| David Ogilvy | Advertising, copy | Headlines, research-driven creative, selling |
| Gary Vee | Content, attention | Distribution, platform strategy, documenting |
| Tim Ferriss | Optimization, 80/20 | Eliminating waste, lifestyle design, testing |
| Zig Ziglar | Sales, follow-up | Closing, handling objections, persistence |
| Don Draper | Emotion, story | Desire, nostalgia, making people feel something |
| Alan Watts | Philosophy, presence | Letting go, ego, finding calm in chaos |
I'm trying to decide how to price [my service/product].
Consult Chris Do and Naval Ravikant. What would each of them say?
Give me their perspectives separately, then tell me where they agree
and where they conflict.Or for a panel:
I'm about to launch [thing]. Get perspectives from Seth Godin
(positioning), Gary Vee (distribution), and David Ogilvy (copy).
What would each one focus on first?Claude adopts each mentor's actual thinking style. Naval talks about leverage and equity. Chris Do asks what the client's pain is worth. Ogilvy demands you lead with research. They don't all agree. That tension is the point.
Agents handle the doing. Mentors handle the deciding. Combine them: run an explorer to gather data, then consult a mentor panel to interpret it. You get research speed and decision quality in one workflow.
Pick a decision you've been sitting on. Consult two mentors with opposing perspectives. See how the tension between their viewpoints clarifies your own thinking. The answer was already in your head. The mentors just surface it.