$
copilot --idea "Learning Accelerator"
→
Learn any technology hands-on by building real projects with Copilot as your interactive tutor
The Problem
Tutorials are passive. You watch someone code, nod along, and forget everything by Tuesday. Documentation is comprehensive but dry. You learn best by building things, but starting from scratch in an unfamiliar technology is paralysing.
What You'll Build
A learning workflow where Copilot CLI acts as your interactive tutor:
- Generates progressively harder coding challenges
- Explains every concept with working code examples
- Reviews your attempts and suggests improvements
- Adapts to your pace and existing knowledge
Learning Paths
Path 1: Learn Rust by Building a CLI Tool
$
"Teach me Rust by building a CLI tool that counts lines of code in a directory.
Start with the absolute basics — what's a main function, how do I print output.
After each step, explain the Rust concept involved.
Build up to: file I/O, pattern matching, error handling, and concurrency."
Progress through increasingly complex features:
$
"Now add support for multiple languages (detect by file extension).
Explain ownership and borrowing as we implement this."
$
"Add a --parallel flag that counts files concurrently.
Explain Rust's concurrency model — Send, Sync, Arc, Mutex."
Path 2: Understand Distributed Systems
$
"Build a simple key-value store that works across multiple nodes.
Start with a single-node in-memory store.
Then add: TCP networking, replication, consistency guarantees.
Explain CAP theorem with our actual code as the example."
Path 3: Build a Compiler
$
"Walk me through building a compiler for a simple calculator language.
Phase 1: Lexer — tokenise '2 + 3 * 4'
Phase 2: Parser — build an AST
Phase 3: Evaluator — walk the tree and compute
Phase 4: Add variables and functions
Explain each computer science concept as we encounter it."
Path 4: Deep Dive into Kubernetes
$
"Teach me Kubernetes by building a microservices app from scratch.
Start with: what's a pod, what's a deployment, what's a service.
Build up to: ingress, config maps, secrets, health checks,
horizontal pod autoscaling, and rolling deployments.
Use a real local cluster (kind or minikube) so I can see everything working."
Path 5: Neural Network from Scratch
$
"Build a neural network in Python WITHOUT any ML libraries.
Just numpy.
Start with: what IS a neuron mathematically?
Build: forward pass, backpropagation, gradient descent.
Train it on MNIST handwritten digits.
Explain every line of math."
The Learning Loop
1. Ask Copilot to explain a concept with code
2. Modify the code yourself — break things intentionally
3. Ask "why does this fail?" — understand the error
4. Ask "how would a senior engineer write this?" — learn best practices
5. Ask "what should I learn next?" — let it guide your path
Pro Tips
• Say "explain each line" after any generated code — it becomes an interactive textbook
• Break things on purpose and ask why — you learn more from errors than successes
• Ask "what's the gotcha here?" — Copilot knows the common mistakes
• Use multiple sessions — come back tomorrow and ask Copilot to quiz you on yesterday's material
• Don't just read the output — type it yourself, modify it, experiment
What You'll Learn
• Whatever technology you choose — the framework works for anything
• How to learn effectively with AI assistance
• The habit of learning by building, not by watching
• How to ask the right questions (prompt engineering IS a learning skill)