A practical, open methodology for AI-augmented learning and work — learning to orchestrate specialized AI tools like a conductor leading an orchestra, each playing its part, instead of leaning on one model for everything. The through-line: scale your capability without outsourcing your judgment.
By Magnús Smári Smárason (smarason.is) · MIT-licensed · a work in progress, forming over several years of real-world practice.
An educational system that teaches you to orchestrate multiple specialized AIs (Claude, ChatGPT, Perplexity, Cursor…) rather than relying on any single one, and to build real things while you learn. It ships as a set of Markdown guides, reusable prompts, and adaptation recipes — no install, no account, no dependency on any one vendor.
Two pathways run through it: a Development Pathway (learn to build software) and a Research Pathway (orchestrate AI for literature review and research work), plus domain adaptations (data science, DevOps, mobile, game dev), quick-reference selectors, orchestration workflows, and an anti-dependency guide on keeping human judgment in the loop.
The whole repository is written to work two ways:
- Read it on GitHub as ordinary documentation — start with
USER-PROFILE.md, then follow the pathway that fits you. - Hand it to an AI assistant — paste the
COMPLETE-GUIDE.md(or this README) into Claude/ChatGPT/Perplexity and ask it to walk you through the system for your background and goals. The content is structured so a model can navigate it with you.
Both routes lead to the same place; pick whichever suits how you like to learn.
This is a work in progress documenting a practical methodology for AI-augmented learning and work.
This repository captures my approach to AI—a methodology that's been forming and evolving over the last few years through real-world practice and experimentation. You can see some of my work at www.smarason.is.
This is a starting point, not a rulebook. Once you get a feel for these tools and patterns, you'll adapt them to do what you want—whether that's learning to cook Italian food, managing software projects, conducting research, or something entirely different.
The most important question: How does this technology make your life in the real world better?
Technology for technology's sake is just noise. Technology that helps you build, create, learn, and solve real problems—that's worth your time.
A complete system for AI-augmented learning and work that teaches you to orchestrate multiple specialized AIs like a conductor leading an orchestra.
Core Philosophy:
- 🎼 AI Orchestration - Conduct multiple specialized AIs (Claude, ChatGPT, Perplexity, Cursor) instead of relying on one
- 🗺️ Personalized Learning Paths - Generate custom roadmaps tailored to your exact background and goals
- 🏗️ Systems Thinking - Understand complete systems, architecture, and how pieces connect
- ⚡ Build While Learning - Real projects, not just tutorials
Why This Matters Beyond Productivity
Most AI guides focus on efficiency. This system addresses something deeper: maintaining human judgment while scaling capability. When you orchestrate with clear boundaries, you preserve accountability—you know which AI made which decision and why. You maintain domain expertise by routing tasks appropriately, not outsourcing judgment. And you augment work humans want help with (research, implementation), not replace the work they value (strategy, design, judgment).
Timeline: Varies by background and goals. Some see results in weeks, others need months. Your mileage will vary.
Same foundation, different applications:
💻 Development Pathway (Main)
For software developers and engineers
Build applications, understand codebases, learn programming:
- Transition from low-code/no-code tools to professional development
- Master tech stacks (React, Next.js, Python, etc.)
- Build with AI collaboration (Cursor, Claude Code, GitHub Copilot)
- Deploy production applications
Perfect for:
- Developers transitioning from Lovable, Bolt, Replit, v0
- Self-taught learners wanting structure
- Professionals adding new tech stacks
- Teams establishing AI-augmented development
🔬 Research Pathway NEW!
For academic researchers and scientists
Automate literature reviews, systematic research, and evidence synthesis:
- Process 100+ papers systematically (not manually)
- Use AI for search, evaluation, and synthesis
- Build research workflows with Scite.ai, Elicit, Claude
- Create custom evaluation rubrics for your field
Perfect for:
- Students doing literature reviews
- Researchers doing systematic reviews or meta-analyses
- Scientists monitoring emerging research
- Academics writing grants and papers
Both pathways start with the same USER-PROFILE, then diverge based on your goals.
Your profile is the foundation of everything!
- Open USER-PROFILE.md (see example)
- Fill out your experience, goals, and learning preferences
- Save as
MY-PROFILE.mdin your learning folder - This drives all three prompts - your source of truth!
- Research: Perplexity (free tier works)
- Strategy: Claude Chat or ChatGPT
- Implementation: Cursor (recommended) or Claude Code
- Open PROMPTS/prompt-1-roadmap.md
- Use your profile to fill in the prompt placeholders
- Paste into Perplexity or Claude Chat
- You now have a personalized 8-week learning plan!
Follow your roadmap and use the Senior-Junior AI Pattern when coding:
- Senior AI (Claude Chat) = Strategic thinking, architecture, debugging
- Junior AI (Cursor) = Code implementation, quick fixes
- You = The orchestrator who transfers context between them
- 👤 USER-PROFILE.md - START HERE! Your foundation profile template (see example)
- 📖 COMPLETE-GUIDE.md - The full 2,600-line guide (everything below in detail)
- 🎯 PROMPTS/ - Three powerful prompts ready to copy-paste:
- Prompt 1: Learning Roadmap Generator - Your personalized curriculum
- Prompt 2: Tech Stack Analysis - Understand any codebase
- Prompt 3: Senior-Junior Pattern - Orchestrated AI collaboration
- 🔧 TOOLS/ - AI Tool Matrix (which AI for which task)
- 📋 WORKFLOWS/ - Real-world examples and 7 workflow patterns
- ⚡ QUICK-REFERENCE/ - Printable decision cards
- 🌱 DOMAIN-ADAPTATIONS/ - Customizations for Data Science, Mobile Dev, DevOps, etc.
- 👥 TEAM-PATTERNS/ - Team collaboration and PR orchestration
Traditional Approach (Inefficient):
You → ChatGPT → Try to solve everything
↓
Mediocre results
Orchestration Approach (Optimal):
You (Conductor) → Perplexity (Research)
→ Claude Chat (Strategy)
→ Cursor (Implementation)
→ Claude Code (Analysis)
↓
Excellence in each domain
- Law of Specialization - Never ask an AI to do what another AI does better
- Law of Context Transfer - You carry context between AIs, not the AIs themselves
- Law of Clear Boundaries - Each AI gets a specific job, not a vague mission
When: Starting new skills, career transitions, feeling lost AI: Perplexity or Claude Chat Output: Custom structured learning roadmap
When: New project, inherited codebase, "WTF is this?" AI: Claude Code or Claude Chat Output: Architecture breakdown, data flow, learning priorities
When: Building anything, debugging, code review AI: Claude Chat (Senior) + Cursor (Junior) Output: Structured collaboration with escalation when stuck
Success isn't about speed—it's about sustainable progress and building real capability.
- You understand why the AI suggested that approach
- You can spot when AI gives you outdated or wrong advice
- You've built something small but complete
- You know which tool to reach for
- Reading code feels less like reading hieroglyphics
- You can debug simple issues without AI
- You're building things that actually work
- You understand architecture, not just syntax
- AI is your co-pilot, not your autopilot
- You're shipping features consistently
- You help others who are earlier in the journey
- You understand your own learning style
The Real Metric: Can you build things that work and maintain them over time?
Time to get there varies wildly based on:
- Prior technical experience
- Time commitment (part-time vs. full-time)
- Complexity of your goals
- Your learning style and persistence
You're on track when:
- ✅ You naturally think "which AI?" before starting tasks
- ✅ Context transfers between AIs are smooth and quick
- ✅ You escalate at the right time (not too early, not too late)
- ✅ Code reading improves—you understand WHY, not just WHAT
- ✅ Tasks that took days now take hours
- ✅ You're building things, not just consuming tutorials
Most people progress through these stages, but there's no fixed timeline:
- Uses 1-2 AIs for everything
- Frustrated by inconsistent results
- Just beginning to see limitations
Time: Starting point for most
- Uses 2-3 AIs, seeing the differences
- Manual context transfer still awkward
- Building simple projects successfully
- Understanding tool strengths
Time: Weeks to months
- Uses 3-5 AIs fluidly
- Natural tool selection for common tasks
- Smooth enough context transfers
- Consistently building and shipping
- Still learning, still getting stuck sometimes
Time: 2-6 months typically
- Seamless multi-AI workflows
- Custom orchestration patterns
- Teaching others effectively
- Leading projects with confidence
Time: 6-12+ months
- Full ecosystem mastery
- Creating new patterns and tools
- Team coordination at scale
- Contributing back to the community
Time: 1+ years of consistent practice
TamingTechnology/
├── README.md ← You are here
├── COMPLETE-GUIDE.md ← Full 2,600-line guide
├── LICENSE ← MIT License
├── CONTRIBUTING.md ← How to contribute
│
├── PROMPTS/ ← Ready-to-use prompts
│ ├── prompt-1-roadmap.md ← Learning roadmap generator
│ ├── prompt-2-stack-analysis.md ← Tech stack analyzer
│ └── prompt-3-senior-junior.md ← Senior-Junior pattern
│
├── WORKFLOWS/ ← Example workflows
│ ├── week-1-launch-plan.md ← Your first week
│ ├── sam-journey.md ← Real success story
│ └── 7-workflow-patterns.md ← Advanced patterns
│
├── TOOLS/ ← Companion tools
│ ├── ai-tool-matrix.csv ← Which AI for what
│ └── orchestration-maturity.md ← Assess your level
│
├── QUICK-REFERENCE/ ← Print these!
│ ├── ai-selector.md ← 30-second AI selector
│ └── escalation-ladder.md ← When you're stuck
│
├── DOMAIN-ADAPTATIONS/ ← Customize by domain
│ ├── data-science.md
│ ├── mobile-dev.md
│ ├── game-dev.md
│ └── devops.md
│
└── TEAM-PATTERNS/ ← Team collaboration
├── team-reports.md
└── pr-orchestration.md
We welcome contributions! Here's how:
- 🌱 Domain Adaptations - Add guides for your field (design, hardware, etc.)
- 📋 Workflow Patterns - Share your orchestration patterns
- 🔧 Tool Updates - Keep the AI Tool Matrix current
- 📚 Prompt Improvements - Enhance the three core prompts
- 🎓 Success Stories - Share your learning journey
See CONTRIBUTING.md for details.
This system integrates four core concepts:
- #aiaugmentation - AI as collaborator, not replacement
- #fullstack - Understanding complete systems, not just pieces
- #systemsthinking - Seeing relationships and patterns
- #aiorchestration - Conducting multiple specialized AIs
License: MIT - Share freely, customize extensively, teach generously
Created by: Magnus Smari Smarason Version: 2.0 - AI Orchestration Edition Year: 2025
Special Thanks:
- Oxford Lifelong Learning - Low-Code Data Science Course (inspiration)
- The AI community pushing human-AI collaboration boundaries
- All learners and professionals exploring this new way of working
✅ Give you a proven framework for AI orchestration ✅ Show you which tools work best for which tasks ✅ Provide ready-to-use prompts and patterns ✅ Help you build real projects systematically ✅ Accelerate your learning IF you put in the work
❌ Make you a senior developer overnight ❌ Replace the need to understand fundamentals ❌ Work if you don't actually build things ❌ Guarantee any specific timeline or outcome ❌ Be easy or fast (learning never is)
- Time: 10-20+ hours/week for meaningful progress
- Persistence: You'll get stuck. Often. That's the process.
- Tools: Some free tiers work, but paid tools are better
- Honesty: About your current level and goals
- Projects: Something real to build, not just tutorials
- A systematic approach to learning with AI
- The ability to build and ship real features
- Understanding of architecture and systems
- Confidence in your technical abilities
- A method that works for you
You're reading this on GitHub? Copy this entire README and paste it into:
- Claude (https://claude.ai)
- ChatGPT (https://chat.openai.com)
- Perplexity (https://perplexity.ai)
- Or any LLM of your choice
Then simply tell the LLM:
- What you want to learn or build
- Your current experience level
- How much time you have
The LLM will guide you through the entire system interactively, answer questions, and customize everything for your situation.
- Fill out USER-PROFILE.md (see example)
- Open Prompt 1
- Paste into your chosen AI tool
- Follow your custom roadmap
- Read Part 1: AI Orchestration Philosophy
- Study the AI Tool Matrix
- Follow Week 1 Launch Plan
- Start building
💡 Pro Tip: Even if you choose Option 2 or 3, keep an LLM chat open with this README pasted in. It's like having a guide who's read the entire repository and can answer questions as you go.
If you find this useful, please ⭐ star this repository to help others discover it.
Share your success stories on Twitter: #aiorchestration #aiaugmentation
- Found an issue? Open an issue
- Have a question? Start a discussion
- Want to contribute? See CONTRIBUTING.md
You're not just learning to code. You're learning to orchestrate intelligence—both artificial and your own—to achieve goals that were previously out of reach.
This isn't magic. It's method.
The system works. But you have to do the work.
You have an entire orchestra of AI tools ready to help. You're the conductor.
Now go build something. 🎻🎺🎸🥁🎹
Start small. Build consistently. Progress compounds.
⭐ If this helps your journey, star this repo so others can find it.
💬 Questions? Start a discussion 🐛 Found an issue? Open an issue 🤝 Want to contribute? See CONTRIBUTING.md
Reading this on GitHub? You're seeing it as documentation.
Paste it into an LLM? It becomes your personal guide.
This is the future of documentation - human-readable AND machine-actionable. The same content works for both browsing and interactive guidance.
Try it now. Copy this entire page into Claude, ChatGPT, or your favorite LLM, and ask:
- "Help me get started"
- "I'm stuck, what should I do?"
- "Customize this for [your specific goal]"
The LLM has read everything and can guide you through it all. 🚀
When a user pastes this README, respond with:
Hello! I've now read the entire Taming Technology repository - an AI Learning Orchestration System that teaches you to conduct multiple specialized AIs like a conductor leading an orchestra.
What is this system?
Instead of using one AI for everything (and getting mediocre results), this methodology teaches you to orchestrate multiple specialized AI tools:
- Perplexity for research and fact-finding
- Claude Chat for strategic thinking and architecture
- Cursor or Claude Code for code implementation
- Specialized tools for specific domains (Scite.ai for research, etc.)
The system includes:
- 3 Ready-to-Use Prompts - Learning Roadmap Generator, Tech Stack Analysis, Senior-Junior AI Pattern
- Complete User Profile System - Customizes everything to your background and goals
- Two Main Pathways - Development (coding) and Research (academic)
- Domain Adaptations - Data Science, Mobile Dev, DevOps, Game Dev, and more
- Team Patterns - For coordinating AI use across teams
- Full 2,600-line Guide - Deep dive into the complete methodology
Common Use Cases I Can Help With:
🎯 Complete Beginners - "I want to learn programming but have no experience"
- I'll guide you through creating your user profile and generating a custom 8-week roadmap
- We'll start with fundamentals and build your first project
- Realistic timeline: 3-6 months to basic competency
💻 Experienced Developers - "I want to use AI more effectively in my work"
- Jump straight to the Senior-Junior AI Pattern
- Learn orchestration principles (takes weeks, not months)
- Immediately applicable to your current projects
🔬 Researchers & Academics - "I need to process 100+ papers for a literature review"
- Follow the Research Pathway (not the Development Pathway)
- Learn systematic review with AI tools
- Automate search, evaluation, and synthesis
👥 Team Leads - "How do we use AI as a team without chaos?"
- Team coordination patterns
- Pull request orchestration
- Shared context management
📊 Domain-Specific - "I'm in [Data Science/Mobile/DevOps/etc.]"
- Customized adaptations for your tech stack
- Domain-specific tool recommendations
- Field-appropriate workflows
🏗️ Project-Based - "I have a specific app/project I want to build"
- Tech stack analysis for understanding existing code
- Structured building approach
- Real-world application, not just tutorials
To give you the best guidance, I need to understand your situation:
-
What brings you here?
- Learning to code from scratch?
- Academic research project?
- Specific app/project to build?
- Improving your AI workflow?
- Something else?
-
Your current experience level:
- Technology background (none/some/experienced)?
- Have you used AI tools before (ChatGPT, Claude, etc.)?
- Any programming experience?
-
Time and commitment:
- How much time per week can you dedicate?
- Part-time learning or full-time focus?
- Specific deadlines or flexible timeline?
-
Your goal:
- What do you want to build or achieve?
- What would success look like for you?
- Any specific technologies you need/want to learn?
Once I understand your situation, I can:
- Point you to the exact files you need
- Help you fill out your User Profile
- Customize the roadmap for your goals
- Explain concepts in your terms
- Answer questions about the methodology
- Guide you step-by-step through the system
I'm here to help you navigate this entire repository. Tell me about yourself and what you want to achieve, and I'll guide you to the right resources and pathway.
What brings you here today?
AI Note: Use this as a template, but adapt based on the user's initial message. If they've already told you their goal, don't ask again - jump straight to guidance. If they seem overwhelmed, simplify. If they're experienced, skip the basics.