How to Approach Learning a New Skill at Three Different Levels

This is a follow-up to my earlier post on the 70-30 split between AI assistants and YouTube playlists. That one was about why. This one is about how - and more importantly, how the approach should change as you go from beginner to advanced.

Most people use the same method at every stage. That’s the mistake. The way you learn on day one cannot be the way you learn in month six. The tools are the same. The mix is different.

I’ll use programming as the running example (and touch on Python here and there because it’s the easiest first language to talk about), but the same idea works for almost anything - a new language, a new framework, design, even a non-tech skill like writing.

Level 1 - Beginner: lean on structure

When you’re brand new, you don’t know what you don’t know. You can’t ask the right questions because you don’t know the right words yet. This is the most important thing to accept at this stage.

So the approach is simple: find one structured course or playlist, and finish it. Not five. One. The order the teacher chose for you is the gift here. Resist the urge to jump around.

Your split at this stage is roughly 40% AI, 60% structured content. The structured content gives you the shape. The AI fills the small gaps.

Use the AI for three things only at this level:

Re-explaining something the teacher said too quickly
Quizzing you on what you just learnt
Giving you small exercises on one topic at a time

Don’t ask the AI to write your code. Don’t ask for advanced tips. Don’t let it pull you ahead of where you are. It will, if you let it, and you’ll end up confused.

A prompt I’d give to an AI at this level:

Beginner AI Prompt: "I’m a complete beginner learning [topic]. I just finished a lesson on [specific sub-topic]. First, ask me to explain it back to you in my own words. Then point out what I got wrong or missed, using simple language a teenager would understand. After that, give me 3 small practice exercises on this sub-topic only - no advanced concepts, no shortcuts. Send them one at a time. After I answer, review my work but only mention beginner-level issues. Do not introduce ideas I haven’t learnt yet."

Build small things at this level. Not impressive things. Small ones. A calculator, a number guessing game, a to-do list. If you finish three of these without copying code, you’re ready for the next level.

Level 2 - Intermediate: lean on building

Once you have the vocabulary, the way you learn flips completely. Now you can ask real questions. Now the AI becomes much more useful than any playlist.

Your split here moves to about 65% AI, 35% structured content. You still pick a good playlist for each new big topic (a framework, a database, a deeper concept) because the shape still matters. But the bulk of your time is now spent building - and the AI sits next to you while you do.

The shift at this level is from learning to do to doing to learn. You pick a project that’s slightly harder than what you can comfortably build, and you fight your way through it. The AI is your pair programmer - not your code generator.

The trap at this level is letting the AI write everything. It feels productive. It is not. You’ll finish projects you don't understand, and the moment something breaks in a way the AI gets wrong, you’ll freeze.

A prompt I’d give to an AI at this level:

Intermediate AI Prompt: "I’m an intermediate learner in [topic/language]. I want to build [describe the project in one or two lines]. Act as my senior pair programmer. Do not write the code for me. Instead, start by asking me how I plan to structure it. After I answer, point out what I’m missing, what could break, and what a senior would think about before starting. Once I begin coding, I’ll share my code in pieces - review each piece by asking me “why did you do it this way?” first, and only then suggest improvements. Keep the conversation honest. If my approach is bad, say so."

Three or four solid projects at this level, all pushed to GitHub with proper READMEs, are worth more than any certificate. This is also the stage to start writing tests, using version control properly, and reading other people’s code.

Level 3 - Advanced: lean on depth

At this level, you’re past tutorials. The questions you’re asking are specific, contextual, and often have no single right answer. “Should I use this pattern or that one for my case?” “Why is my service slow under this specific load?”

Your split here is roughly 50-50, but the kind of learning changes completely. You read. Real books. Real engineering blogs. Real source code. Conference talks. You watch how experienced people think, not how they teach.

The AI’s role shifts too. You don’t use it to learn topics. You use it to challenge what you already think you know. Bring it your assumptions and let it push back.

This is also the stage where you start contributing - to open source, to internal teams, to blog posts, to mentoring juniors. Teaching is the final form of learning. You don't fully understand something until you’ve explained it to someone else and watched their face.

A prompt I’d give to an AI at this level:

Advanced AI Prompt: "I’m an experienced developer working on [specific topic]. My current understanding is: [explain in your own words, with examples]. I’ve read [book/blog/source if any]. I want you to challenge my understanding. Point out where I’m oversimplifying, what edge cases I’m missing, and where my mental model would break in a real production system. Use concrete examples. If you think I’m mostly right but missing nuance, say exactly what nuance. Don’t agree with me just to be polite."

That last line matters. AI tools tend to flatter you. At this level, flattery slows you down.

The thread that runs through all three levels

Notice what didn’t change across the levels: you’re always doing some structured learning, you’re always using the AI, and you’re always building. What changes is the ratio and the kind of question you ask.

Beginner: “Teach me.”
Intermediate: “Build with me.”
Advanced: “Challenge me.”

If you keep asking “teach me” at the advanced stage, you’ll stay shallow. If you ask “challenge me” at the beginner stage, you’ll get crushed. Match the question to where you actually are, not where you wish you were.

That’s the whole post. Pick your level honestly, pick your prompt, and start tonight.