CYFD #015: The AI Learning Paradox

Why ChatGPT Makes You Dumber (And How To Fix It)

Hey friends! 👋🏻

It’s Edo from Code Your Future!

Thank you so much for subscribing to this digest!

I really appreciate your support and your time. 🙏🏻

This is the 15th edition of Code Your Future Digest, and today we’re tackling the biggest trap in modern software engineering education.

If you’re using ChatGPT like a magical answer machine, you’re not learning — you’re outsourcing your thinking.

The developers who master software engineering in the AI era won’t be the ones who use AI the most. They’ll be the ones who use AI to build deeper mental models, not replace them.

Your learning velocity depends on treating AI as a tutor, not a solution vending machine.

Most people make one critical mistake: they ask AI to solve their problems instead of helping them understand the problem.

They paste error messages and accept whatever code comes back. They ask for complete implementations instead of explanations.

They optimize for “making it work” instead of “understanding why it works.”

They fail because they’re building a dependency, not a skill.

The Learning Paradox Is Real

AI makes it easier to get code. That makes it harder to get good.

Every time you accept an AI solution without understanding it, you’re creating technical debt in your brain. You’re building on a foundation you don’t comprehend.

Six months from now, you’ll have a portfolio full of code you can’t explain.

The learners who win don’t use less AI. They use AI differently.

What AI-Assisted Learning Actually Looks Like

Let me show you the difference with a real scenario:

Wrong way (dependency building):

- Paste error: “TypeError: Cannot read property ‘map’ of undefined”

- Get solution: “Add optional chaining: `data?.map()`”

- Copy code ✅

- Move on

Right way (skill building):

- Paste error: “TypeError: Cannot read property ‘map’ of undefined”

- Ask AI: “Explain why this error happens and what JavaScript concepts I need to understand”

- Learn about: undefined vs null, optional chaining, defensive programming

- Ask follow-up: “Show me 3 different ways to handle this, and explain the tradeoffs”

- Implement yourself

- Ask AI to review your implementation

See the shift? The first approach gets you working code. The second approach gets you working knowledge.

I’m Watching This Destroy Careers

Two months ago, I interviewed a developer with an impressive GitHub. Beautiful projects. Complex features. Clean code.

I asked them to explain how their authentication system worked.

They couldn’t.

They’d built it entirely with AI prompts. They had no idea what JWT tokens were, how sessions worked, or why their approach was secure (or wasn’t).

They had the output of learning without the learning itself.

That’s a house built on sand. The first real problem they encounter will collapse their entire facade.

AI didn’t fail them. Their approach to AI did.

The AI Learning Framework

Here’s how to use AI to actually get better at software engineering:

Level 1 - Concept Understanding (Start Here)

- Ask: “Explain concept X like I’m familiar with programming but new to this domain”

- Ask: “What prerequisite knowledge do I need to understand this?”

- Ask: “Show me a simple example, then a complex real-world example”

Level 2 - Pattern Recognition (Build Your Mental Models)

- Ask: “What are 3-5 different approaches to solve this? What are the tradeoffs?”

- Ask: “When would I use X vs Y in production?”

- Ask: “Show me how this pattern appears in popular libraries”

Level 3 - Implementation (Supervised Practice)

- Implement yourself first

- Ask AI to review your code

- Ask: “What would a senior developer do differently and why?”

- Refactor based on understanding, not blind acceptance

Level 4 - Deep Dives (Build Expertise)

- Ask: “What edge cases should I consider?”

- Ask: “How does this work under the hood?”

- Ask: “What are common mistakes developers make with this?”

Track your questions. If more than 50% are “how do I build X,” you’re using AI wrong.

The Feynman + AI Method

Here’s my favorite learning technique adapted for the AI era:

Step 1: Learn a concept using AI (ask for explanations, examples, analogies)

Step 2: Close the AI. Explain the concept out loud like you’re teaching it

Step 3: Identify gaps in your explanation. What did you struggle to articulate?

Step 4: Go back to AI with targeted questions about your knowledge gaps

Step 5: Build something that uses the concept without AI assistance

Step 6: Ask AI to review what you built and teach you what you missed

This creates a feedback loop that compounds understanding instead of dependency.

The Struggle Is the Learning

Here’s a hard truth: if you never struggle, you never learn.

AI can eliminate all struggle. That’s the trap.

When you get stuck on a bug for 30 minutes, your brain is building neural pathways. You’re developing debugging intuition. You’re learning to read stack traces, trace execution, and think like the computer.

When you immediately ask AI to fix it, you’re outsourcing that growth.

New rule: Struggle for at least 15 minutes before asking AI for help.

When you do ask AI, don’t ask for the answer. Ask for a hint. Ask for similar examples. Ask for debugging strategies.

Save the “just give me the solution” prompts for when you’re building products. When you’re learning, choose the harder path.

How to Use AI for Deliberate Practice

Deliberate practice is how you get good at anything. Here’s how to use AI to supercharge it:

1. Spaced Repetition Projects

- Build the same thing 3 times over 3 weeks

- First time: Use AI heavily to understand the patterns

- Second time: Use AI only for review and validation

- Third time: Build from memory, use AI only when genuinely stuck

2. Comparative Learning

- Ask AI: “Show me how to build X in React, Vue, and vanilla JavaScript”

- Build all three yourself

- This reveals patterns vs syntax, principles vs frameworks

3. Constraint-Based Challenges

- “Build a todo app without using any libraries”

- “Implement array.map() from scratch”

- “Explain to AI how closures work before asking it anything about closures”

4. Code Review Loops

- Write code without AI

- Ask AI to review it like a senior developer would

- Study the feedback. Ask follow-up questions about suggestions

- Refactor and repeat

5. Explanation Challenges

- Ask AI to show you code that uses a concept

- Before asking for explanation, try to explain it yourself

- Then ask AI to explain. Compare. Learn from the gaps.

The Questions You Should Be Asking

Stop asking AI to solve. Start asking AI to teach.

Instead of:

- “Write a function that does X”

- “Fix this error”

- “How do I implement Y”

Ask:

- “What concepts do I need to understand to build X myself?”

- “What does this error tell me about how JavaScript works?”

- “Walk me through the thought process of solving Y”

Then:

- “I tried to implement this [paste your code]. What am I misunderstanding?”

- “Can you explain why your approach is different from mine?”

- “What should I study next to get better at this?”

The quality of your questions determines the quality of your learning.

Build Your Learning System

Don’t rely on ad-hoc AI conversations. Create a deliberate learning system:

Daily:

- One concept deep dive with AI (30 minutes)

- Implement something using that concept without AI (30 minutes)

- Review with AI, ask follow-up questions (15 minutes)

Weekly:

- Build one small project that combines the week’s concepts

- Ask AI to review your architecture and approach

- Write a brief explanation of what you learned (Feynman test)

Monthly:

- Build something meaningful that pushes your limits

- Use AI as a pair programmer and teacher

- Document patterns you’ve internalized vs patterns you still rely on AI for

This compounds. In 6 months, you’ll have both the portfolio and the knowledge to back it up.

The Meta-Skill: Learning How to Learn

The most important thing AI can teach you isn’t React, Python, or system design.

It’s how to learn anything quickly.

Use AI to understand your learning process:

- “I’m struggling to understand recursion. What are 5 different ways I could approach learning this?”

- “I learn best by building. Give me 10 project ideas that would teach me async programming.”

- “I just learned about closures. What should I learn next to build on this foundation?”

AI can be your personal curriculum designer. Use it to map your learning journey, not just answer individual questions.

Start Here This Week

Three immediate actions to transform how you learn with AI:

1. The 15-minute rule - Next time you get stuck, struggle for 15 minutes before asking AI. When you do ask, request a hint, not a solution.

2. The explanation test - After AI teaches you something, close the chat and explain it out loud. Record yourself if needed. Identify what you can’t articulate, then go back to AI with targeted questions.

3. The review loop - Write code without AI, then ask AI to review it like a senior developer. Focus on understanding the feedback, not just implementing it.

These small shifts transform AI from a crutch into a catalyst.

Key Takeaways

- AI amplifies your approach - If you’re learning poorly, AI makes you learn worse faster. If you’re learning well, AI makes you learn better faster.

- Struggle builds skills - The harder path creates stronger understanding. Use AI to guide your struggle, not eliminate it.

- Ask for teaching, not solutions - “Explain” and “why” questions build knowledge. “Do this for me” questions build dependency.

- Implement before you ask - Try yourself first. Use AI for review and refinement, not initial implementation.

- The Feynman test is your north star - If you can’t explain it simply without AI, you don’t understand it yet.

- Build a learning system - Deliberate, structured practice with AI as your tutor beats random questions to ChatGPT.

- Quality of questions determines quality of learning - Your prompts shape your progress. Ask better questions, build better understanding.

- Document your learning - Writing explanations forces clarity. Use AI to review your explanations and fill gaps.

The developers who become truly great won’t be the ones who avoid AI. They’ll be the ones who use AI to build deeper understanding, faster.

That’s it for today! 🎉

I’d love to hear your thoughts on this. How are you using AI to learn?

What techniques have worked for you? What traps have you fallen into? 🤔

My goal is to help software engineers build genuine expertise in this new era of AI-assisted learning. 💪

If you found this helpful, I’d love to hear from you. Drop me a line on X, LinkedIn, or through SubStack!

Thanks, and a happy journey to you! ❤️

Edo ✌🏻