If I Had to Learn Programming Today, This Is How I’d Actually Do It

Short answer: I’d spend about 70% of my learning time with an AI assistant like ChatGPT or Claude, and the remaining 30% on carefully picked YouTube playlists. The AI handles the questions, the doubts, and the small experiments. YouTube still wins for the slow, deep topics where you need a real human to walk you through the shape of things.

That’s the gist. Now let me tell you how I got here, and how you can use this split without losing the depth that makes a good developer.

A bit about where this comes from

I’ve been writing code for close to ten years. Built systems from a blank folder to running in production. Sat through bad architecture calls, fixed my own messes, mentored juniors who are now seniors themselves. So when I share this, it isn’t a hot take from someone who just discovered AI. It’s what I genuinely believe after watching the last two years change how all of us learn.

If someone handed me a fresh laptop tomorrow and said “start over, learn programming from scratch,” I would not learn the way I did the first time. Not even close.

How I used to learn (and why it worked for years)

My method was simple.

I’d pick a topic. Kafka. System design. A new framework everyone was talking about. I’d open YouTube, type the name, and look for a full playlist. Not a single video. Twenty videos, sometimes forty.

Then I’d pick the one that felt right. Usually the teacher who didn’t shout, didn’t use flashy thumbnails, and spoke like a normal human. I’d watch one video a night. Pause, try the code, break it, fix it, move on.

That routine carried me for years. It got me to where I am.

The moment things shifted

Sometime last year, I was halfway through a playlist on a topic and I had a small doubt. I opened ChatGPT in another tab and asked.

The answer I got was better than the next eight videos in that playlist.

Not flashier. Better. It answered exactly what I asked, in the context of what I was building, in the words I used. The YouTube teacher was teaching a room of a thousand people. The AI was teaching me.

That was the moment something clicked. Not “AI will replace teachers.” Nothing dramatic. Just a quiet realisation that the shape of learning had changed.

So how would I start today

I’d split my learning roughly 70% with an AI assistant, 30% with YouTube playlists. The split isn’t random. Each side does something the other can’t.

The 70% - where AI assistants win

Most of learning is asking small, sometimes silly questions and getting a straight answer.

Why does this loop not stop?
What is the real difference between let and var?
Why is my Docker container exiting right after it starts?

YouTube cannot answer these. A blog post cannot. You have to scroll, search, and hope someone on Stack Overflow had the exact same problem in 2017.

An AI just answers. And if the answer is wrong or unclear, you push back. “No, that’s not what I meant, here’s my actual code.” It adjusts. That back and forth is the closest thing to having a senior developer sitting next to you that I’ve seen in this industry.

Here’s what I’d move to the AI side:

Syntax doubts and quick concept checks
Explaining the same idea in five different ways till one clicks
Reviewing my own practice code
Comparing two approaches for my specific case
Generating practice problems on what I just learnt
Translating cryptic error messages into plain English
Walking through a piece of open source code line by line

For all of this, an AI is faster than any human teacher I’ve had. And it never gets tired of the dumb questions, which is honestly the bigger gift.

The 30% - where YouTube still wins

Now here is where I have to be careful. If I told you “just use AI for everything,” I’d be lying, and you’d figure it out in three months when your foundation feels hollow.

YouTube playlists still do three things AI cannot. At least not yet.

1. They give you the shape of a topic

When you’re brand new to something, you don’t know what you don't know. You can’t ask the right questions because you don’t even know the right words. A good playlist takes you from “what is this” to “okay I see the landscape” in a structured order. That order is the gift. The teacher already thought about it for you.

2. You get to watch someone think

A senior building something on screen, hitting an error, muttering, fixing it, moving on - that teaches you how to be a developer. Not just how to code. The pauses, the second guesses, the small habits. You absorb that by watching. An AI cannot give you that.

3. The slow, core topics

Data structures. Operating systems. Networking. Databases at a deep level. System design. These are topics where you need to sit with one teacher for hours and let them build the picture in your head. If you jump around with an AI here, you’ll end up with a bag of facts and no real understanding.

For these, I’d still hunt for a good playlist. Same as I always did. Find the teacher who speaks like a human. Watch one video a night. Build alongside.

A simple week, if you’re wondering

To make this real, here’s what one week would look like if I were learning backend development today.

Monday to Wednesday: Watch two videos from a solid Node.js or Spring Boot playlist. The boring kind, where the teacher explains why, not just what. Take notes by hand. Don’t code along yet. Just listen and absorb.
Thursday: Sit with the AI. Tell it what I watched. Ask it to quiz me. Ask it to give me a small project that uses exactly what I learnt. Build it. Break it. Ask why it broke.
Friday: Read one piece of real code on GitHub. Paste the confusing parts into the AI and ask, “explain this like I’ve been coding for one year.” Not five. One. Because if it explains for five, it’ll skip the parts I actually need.
Weekend: Build something tiny of my own. No tutorial. No playlist. Just me and the AI, like a pair programming buddy who never sleeps.

That’s the loop. Repeat it for a few months and the difference is real.

How to pick a good YouTube playlist (since this is half the battle)

A quick checklist I use:

The teacher speaks at a normal pace, no shouting.
The thumbnails are plain, not full of arrows and shocked faces.
The first video explains why the topic matters, not just what.
The code on screen is readable and they actually run it.
The playlist is finished, not abandoned after five videos.
Comments under the videos show real learners asking real questions.

If a playlist passes four out of six, it’s worth your time.

How to use an AI assistant without becoming lazy

This part matters more than the rest. Because the AI is fast - dangerously fast. You can finish a topic in two hours that used to take two weeks. That feels great, but speed is not the same as depth.

A few rules I follow:

Write the code yourself first, even if it’s wrong. Then ask the AI to review. Never the other way around.
Ask “why,” not just “what.” “Why does this work?” teaches you. “Give me the code” doesn’t.
Re-explain what you learnt, out loud or in writing. If you can’t, you didn’t learn it.
Once a week, build something with no AI at all. Just to check what stuck.

If you only use AI, you’ll be the developer who can build anything and explain nothing. You’ll freeze the moment the AI is wrong, because you won’t have the base to catch the mistake.

If you only use YouTube, you’ll be slow. You’ll spend weeks on things that should take days.

The 70-30 isn’t a magic number. It’s just a reminder. Use the new tool for most of the work. Use the old tool for the parts that build the spine. Both are needed.

Frequently asked questions

Is YouTube still worth it for learning programming in 2026?

Yes, but in a smaller role. Use it for the slow, foundational topics - data structures, system design, databases, networking. For everything else, AI is faster and more personal.

Can a complete beginner learn programming using only ChatGPT or Claude?

You can, but I don’t recommend it. Beginners don’t know the right questions to ask. Start with a structured playlist for the first month, then bring in AI heavily once you know the basic vocabulary.

Won’t I forget things if AI explains everything?

Only if you let it. The trick is to write the code yourself first, then use AI to review. Treat it like a senior, not a vending machine.

Which AI tool is best for learning to code?

Any of the big ones - ChatGPT, Claude, Gemini - are good enough. The tool matters less than how you use it. Ask better questions, push back when answers feel off, and always test the code yourself.

How long before I can build real projects?

With this 70-30 method and one focused hour a day, most people can build a small but real project in 8 to 12 weeks. Not a clone of YouTube. Something small and yours.