The Ultimate Learning Guide: How to Actually Learn (Not Just Read)
"Reading without application is like eating without digestion."
Why This Guide Exists
Most people "study" by:
- Highlighting everything
- Re-reading notes
- Watching videos at 2x speed
- Feeling productive without actually learning
This guide will teach you how to actually learn - techniques backed by cognitive science that will help you remember concepts for interviews and beyond.
The 5-Step Learning Loop
For every concept you study, follow this loop:
1. READ → Understand the concept
2. EXPLAIN → Say it in your own words (Feynman Technique)
3. TEST → Quiz yourself without looking
4. APPLY → Use it in a problem or project
5. CONNECT → Link it to other concepts you know
Example: Learning "Gradient Descent"
| Step | Action |
|---|---|
| READ | Read the textbook/article about gradient descent |
| EXPLAIN | "It's like a ball rolling downhill to find the lowest point, where we adjust parameters based on the slope" |
| TEST | Close the book: What's the update rule? Why do we need a learning rate? |
| APPLY | Implement it from scratch in Python, train a simple model |
| CONNECT | How does it relate to SGD? Adam? Backpropagation? |
The Feynman Technique (Your Secret Weapon)
Named after physicist Richard Feynman, this technique exposes gaps in your understanding.
How to Use It:
- Choose a concept (e.g., "Batch Normalization")
- Explain it to a 10-year-old (no jargon)
- Identify gaps where you struggle to explain simply
- Go back to source and re-learn those gaps
- Simplify and use analogies
Example: Explaining Batch Normalization
Bad explanation:
"Batch normalization normalizes the activations of each layer by adjusting and scaling the outputs."
Good explanation (Feynman style):
"Imagine you're baking cookies. Each batch of dough might be slightly different - some wetter, some drier. Batch norm is like standardizing every batch of dough before baking, so your cookies come out consistent every time. In neural networks, it makes sure the numbers flowing through each layer stay in a 'nice' range, so learning is faster and more stable."
Active Recall: The #1 Learning Technique
What is it? Forcing yourself to retrieve information from memory WITHOUT looking at notes.
Why it works:
- Your brain strengthens memory pathways when you struggle to recall
- Passive re-reading creates an "illusion of competence"
- Testing yourself is 3x more effective than re-reading (proven by research)
How to Practice:
| Method | How to Do It |
|---|---|
| Flashcards | Use Anki with questions like "What problem does dropout solve?" |
| Blank Page | After studying, write everything you remember on a blank page |
| Teach Someone | Explain the concept to a friend, rubber duck, or yourself |
| Practice Problems | Solve problems without looking at solutions first |
Sample Active Recall Questions:
After learning about CNNs, quiz yourself:
- What is the purpose of a convolutional layer?
- How does max pooling reduce dimensions?
- Why do we use ReLU instead of sigmoid?
- What's the receptive field and why does it matter?
Spaced Repetition: Remember Forever
The Problem: You learn something today, forget 80% by next week.
The Solution: Review at increasing intervals.
Optimal Review Schedule:
| Review # | When | Why |
|---|---|---|
| 1st | Same day | Fresh memory consolidation |
| 2nd | 1 day later | Before you forget |
| 3rd | 3 days later | Strengthening the pathway |
| 4th | 1 week later | Moving to long-term memory |
| 5th | 2 weeks later | Solidifying for good |
| 6th | 1 month later | Almost permanent now |
Tools:
- Anki - Free, customizable flashcard app with built-in spaced repetition
- RemNote - Note-taking with automatic flashcard generation
- Notion + manual scheduling - Track review dates yourself
The Interview Preparation Framework
Phase 1: Foundation Building (First 2-4 weeks)
Goal: Build solid understanding of core concepts
| Day | Focus Area | Study Method |
|---|---|---|
| Mon-Tue | ML Fundamentals | Read + Feynman |
| Wed-Thu | Coding/Algorithms | Practice + Explain |
| Fri | System Design Basics | Case studies |
| Weekend | Review + Projects | Spaced repetition |
Phase 2: Deep Practice (Weeks 3-6)
Goal: Interview-ready problem solving
Daily routine:
Morning (2 hrs):
- 1 ML concept deep-dive
- Explain it out loud (record yourself!)
Afternoon (2 hrs):
- 2-3 LeetCode problems
- Write explanations for each
Evening (1 hr):
- Review flashcards (Anki)
- Read 1 system design case
Phase 3: Mock Interview Mode (Final 1-2 weeks)
Goal: Simulate real interview pressure
- Schedule mock interviews with friends/peers
- Practice with a timer (45 min per problem)
- Record yourself and review
- Focus on communication, not just answers
The "Explain Before You Code" Rule
Before writing any code in practice or interviews:
- State the problem in your own words
- Discuss approach and tradeoffs
- Walk through an example by hand
- Identify edge cases
- THEN start coding
This habit will:
- Catch bugs early
- Show interviewers your thought process
- Reduce anxiety (you have a plan!)
Building Mental Models
What is a Mental Model?
A simplified representation of how something works that you can apply to new situations.
Example Mental Models for ML:
| Concept | Mental Model |
|---|---|
| Overfitting | "Memorizing the answers vs. understanding the subject" |
| Regularization | "Adding a penalty for complexity - keeping it simple, stupid" |
| Ensemble Methods | "Wisdom of the crowd - many weak learners become strong together" |
| Attention | "A spotlight that focuses on what's important in a sequence" |
| Embeddings | "Translating words/items into coordinates in a meaningful space" |
How to Build Mental Models:
- Learn the concept deeply
- Find or create an analogy
- Identify what the model explains well (and its limitations)
- Test it on new problems
- Refine based on feedback
Common Learning Mistakes to Avoid
| Mistake | Why It's Bad | What to Do Instead |
|---|---|---|
| Passive reading | Creates false sense of understanding | Use active recall after every section |
| Skipping basics | Advanced concepts won't stick | Master fundamentals first |
| Tutorial hell | Following along != understanding | Build something from scratch |
| No testing | You don't know what you don't know | Quiz yourself constantly |
| Cramming | Poor long-term retention | Use spaced repetition |
| Avoiding hard problems | Growth comes from struggle | Embrace difficulty |
| Not explaining out loud | Writing/thinking isn't enough | Verbalize explanations |
The Daily Learning Routine
Morning Routine (High Energy)
[ ] 5 min: Review yesterday's concepts (active recall)
[ ] 25 min: Deep learning of one new concept
[ ] 5 min: Summarize in your own words
[ ] 25 min: Apply - solve a related problem
[ ] 5 min: Connect to other concepts
Evening Routine (Review)
[ ] 10 min: Anki flashcard review
[ ] 10 min: Explain today's concept to yourself
[ ] 5 min: Plan tomorrow's learning
How to Use This Website Effectively
For Each Question on This Site:
- Read the question - Don't peek at the answer!
- Try to answer it yourself first (even if wrong)
- Write down your answer - Force active recall
- Compare with provided answer - Identify gaps
- Explain WHY the answer is correct
- Add to your flashcards for spaced review
For Each Project:
- Don't just follow the tutorial - pause and predict
- Build a variation of your own
- Be able to explain every line of code
- Prepare to discuss it in interviews
Quick Reference: Learning Techniques Summary
| Technique | When to Use | Effectiveness |
|---|---|---|
| Active Recall | Every study session | Extremely High |
| Spaced Repetition | Daily reviews | Extremely High |
| Feynman Technique | New/complex concepts | Very High |
| Practice Problems | After understanding theory | Very High |
| Teaching Others | To solidify understanding | Very High |
| Mind Mapping | Connecting concepts | High |
| Re-reading | Quick reference only | Low |
| Highlighting | Never (seriously) | Very Low |
Your Learning Checklist
Before you can say you "know" a concept:
- Can explain it without notes
- Can explain it to a non-technical person
- Can apply it to solve a problem
- Can identify when to use it (and when NOT to)
- Can connect it to related concepts
- Can answer interview questions about it
- Have used it in a project or real code
Final Advice
"The person who says they understand everything they've read has read nothing." - Unknown
Learning is uncomfortable. If it feels easy, you're probably not learning.
Embrace the struggle:
- That confused feeling? Your brain is building new connections.
- That frustration? You've found a gap worth filling.
- That "aha!" moment? You've just leveled up.
Now go apply these techniques. Start with one concept from this site, and run it through the 5-Step Learning Loop.
Resources
Books:
- "Make It Stick" - Brown, Roediger, McDaniel
- "A Mind for Numbers" - Barbara Oakley
- "Ultralearning" - Scott Young
Tools:
- Anki - Free spaced repetition flashcards
- Excalidraw - Draw diagrams and mental models
- Notion - Organize your notes and learning
Research:
- "Retrieval Practice Produces More Learning" - Karpicke & Blunt (2011)
- "The Critical Importance of Retrieval for Learning" - Karpicke & Roediger (2008)
Remember: The goal isn't to read more - it's to understand deeper.
Good luck with your preparation!