Learn Coding or AI First? Here's What I Would Do

What learning to code actually means

If I had to start from scratch today, I would lean on first principles thinking: go back to the fundamentals, and embrace AI as a powerful tool on top of them.

Learning to code is not about memorising syntax. It is about understanding how computers, the internet and the web work at a fundamental level. What is a web server? How do apps talk to each other? How do APIs act as bridges between software systems? Oracle describes an API as an interface that lets applications communicate and share data, and that idea sits underneath almost everything you will build.

When I tutor students or build products, I start with these core ideas. The front end is what users see. The back end is the invisible logic: HTTP requests, databases, algorithms. Once you understand those pieces, everything else becomes much easier to learn. If you are completely new, our explainer on what coding actually is is a gentle place to start.

Your goal decides the balance

At the end of the day, it comes down to what you are aiming for. The two goals I see most often pull in slightly different directions.

Why the founder path still needs foundations

I never became a full-time software engineer, but I learned Vue, React and back-end stacks like Express and Ruby on Rails. Knowing what each piece does gave me the technical literacy to hire developers well and build businesses around software.

Later I co-founded an AI-powered job search startup after seeing how painful job hunting was for some of my students. It uses AI to help write CVs and screen jobs, but under the hood it is still a software product with a database, a user interface and back-end logic. If I had not learned those core skills first, I could not have architected it.

This is similar to how self-taught builder Pieter Levels works. He taught himself just enough by searching how to connect a database or make a button work, then built real projects and learned along the way. That hands-on approach is powerful because every bit of code you write drills the fundamentals in, like learning a language by moving to the country.

AI is a powerful amplifier, but it needs a strong human hand to guide it. Go AI-first with no foundations and you will feel productive at first, then hit walls. I have seen students blindly accept AI-generated code and introduce messy, buggy code without noticing, because they were just pressing Tab and saving whatever the AI gave them.

Tools change. Principles do not

Remember when jQuery was all the rage, then Vue, then React? If you only chase the latest trend, you will spend your life chasing your tail. Serverless and microservices are popular now, but decades-old concepts like HTTP requests and data structures are still the foundation. Fundamentals are a reusable toolkit: once you have them, you can pivot into AI or cloud much more easily.

Jason Fried and DHH, bootstrapped founders I admire, preach doing more with less and shipping quickly. Pieter Levels built very successful products with plain HTML, CSS, a bit of JavaScript and PHP, because he was building a product, not a CV. Users do not care what you built it with. They care whether the button works.

So learn the why, not just the how. If you learn AI, understand roughly what is under the hood. If you learn to code, learn the logic, not just the syntax. You do not need to memorise every library method. You need to know where to find things and how they work, so that when you get stuck you can trace back to first principles. Charlie Munger's inversion principle helps here: ask what not to do, and the answer is usually don't skip the fundamentals. If an AI agent hallucinates and you do not know what it is doing, you cannot fix it. For more on this trade-off, read vibe coding versus fundamentals.

Stack skills beyond code

If I started again, I would not silo myself in tech. I would learn to write about what I was learning, even a weekly blog or newsletter, because teaching others solidifies your understanding and gets you noticed. Writing guides for students forced me to clarify what an API is, why you use GitHub and what Git actually does.

Communication is a keystone habit. One commenter on Pieter Levels' story pointed out that the best programmers he had worked with also understood business and marketing, and built things people wanted instead of blindly following the scope. So if you learn Python, also learn to explain your projects on LinkedIn and in interviews. If you are diving into AI, learn basic statistics.

And pick one language and stick with it until you can build something meaningful: Python if AI interests you, JavaScript if you want to go deeper into web development. Be good at one before you hop. Jumping from resource to resource is the fastest way to get overwhelmed. Our guide on choosing your first programming language can help you decide.

Is this still relevant when AI writes code?

Obviously yes. Stack Overflow's 2025 developer survey found that 84% of developers use or plan to use AI in their workflow, about half of them daily. AI is everywhere in tech jobs and you cannot run away from it. But that also means everyone has similar tools. What sets you apart is still your own understanding of software. Ask two people to debug an issue with Copilot, and the one who truly understands the logic will give me the faster, safer fix.

The same survey showed developer sentiment towards AI dipping slightly. I read that as experienced developers seeing AI as a big plus but not a replacement, which matches my view. In practice I would use AI for repetitive work like boilerplate code or summarising documents, but design the system architecture, key algorithms and tests myself, and verify every line before it goes into the team's repository.

When we built the back end for my startup, AI suggestions helped us write SQL queries and Node.js endpoints faster. But conceptualising the whole pipeline, how the CV parser feeds data into the matching algorithm, came from me and my developers. I once worked on an encryption feature where I had to trade off speed against security. AI did not know our product's requirements or what our client needed. Understanding the algorithms and the context did.

My step-by-step playbook from scratch

Start with a real project, not tutorials. Pick a small but meaningful problem: a personal website, a micro business idea, a daily coding newsletter app. Building it forces you to set up a server, connect a database and design a front end, so you learn HTML, CSS and JavaScript plus something like Express or Flask on the back end. When you get stuck, search and read the docs. Doing an entire unrelated course to solve one problem is, in my opinion, procrastination.

Pause to understand each piece. What is the HTML really doing? Why these CSS rules? How does JavaScript interact with the DOM? What does this SQL query fetch? Get comfortable with Git and the command line. Treat code reviews from mentors or peers as gold. Every evening, write down one insight, such as how HTTP GET differs from POST.

Then experiment with AI. Use it to generate test data or suggest optimisations, try an AI pair programmer in your editor and watch what it does. But always ask yourself whether you understand what each suggested line does.

Two sides of the same coin

It is never too late, and either path is valid if you approach it thoughtfully. I would not be dogmatic. I would learn coding fundamentals and play with AI as soon as I could, because they are two sides of the same coin. The internet still needs people who can build things, and the world needs people who can use AI safely and creatively.

Do not be afraid to build something imperfect. Launch it, get feedback and iterate. Every finished project, however small, is proof you are on the right track. That fundamentals-first approach is exactly how Sigmaschool teaches, and it is why we back it with a money-back guarantee. If you want to try it, the free trial is the easiest first step: write your first lines of real code, with AI alongside you.