The Reasons for Learning Coding Have Changed in the Age of AI
After Completing This Topic
You will understand why coding fundamentals are still necessary in the age of AI and gain a clear direction on how to approach this course.
Coding Education in the Past
Coding was long considered the domain of computer science and software engineering.
C, Java, Python, JavaScript – there were dozens of languages, each requiring you to learn its syntax, master frameworks, write thousands of lines of code, and debug it. It was nearly impossible to acquire all of this on your own. You had to spend years, even decades, pursuing a university degree, accumulate experience in real-world projects, and brainstorm with your team members to create a single service.
Therefore, the education method of the past was naturally structured to build from the ground up, step by step. You learned about variables, conditional statements, loops, functions, and objects – you had to master all of these steps before moving on to the next. Harvard University’s CS50 teaches binary and C language pointers from the very first week, and Korea’s leading course, “Coding for Everyday Life,” guides you through the process of creating a web server in 12 weeks, starting with HTML. This approach was undoubtedly systematic, but it had one premise: you ultimately have to write all the code yourself.
What AI Has Changed
But then AI emerged.
With the release of OpenAI’s ChatGPT in late 2022 and the widespread adoption of GitHub Copilot in 2023, the landscape of coding has completely changed. AI now writes code at a level that surpasses humans. You can describe your intention without having to memorize syntax, paste in errors to analyze the cause, and even ask for the code to be optimized to make it faster.
Andrej Karpathy, Tesla’s AI chief and OpenAI co-founder, said:
"The hottest new programming language is English." — Andrej Karpathy, January 2023, X (formerly Twitter)
We now live in an era where we work alongside the world’s best programmers.
This change does not mean that “learning to code is unnecessary.” On the contrary, it means that the type of things we need to learn has changed.
So, Why Are Fundamentals Still Needed?
For very simple code, it’s not a problem. AI will create a login page if you ask it to.
But when the code grows to tens of thousands or hundreds of thousands of lines, the story changes. When there are hundreds of files, database tables intertwined, and servers and front-ends intricately connected, neither AI nor humans can handle it alone.
If you don’t know what refactoring is, don’t understand the structure of the code, and don’t know about concepts like normalization or optimization, what happens then?
You can’t give instructions to AI.
You can say, “Fix this code,” but to say, “Separate this component and refactor it to be reusable,” you need to know what refactoring is. You can say, “The DB is slow,” but to say, “Solve the N+1 query problem in this table with a JOIN,” you need to know what JOIN is and what N+1 is.
In the age of AI, people no longer need to memorize every detail of the code. AI can handle everything from explaining concepts to writing code, designing basic structures, and optimizing it. However, if you don’t know the terminology and don’t understand the fundamentals of computer science, you won’t be able to debug even minor issues, and you won’t even be able to judge whether the direction AI is heading in is correct.
What’s Actually Happening
This is not a theoretical story. It’s happening right now.
A Lawyer Who Couldn’t Verify AI Code and Was Taken to Court
In June 2023, lawyer Steven Schwartz was sanctioned in the Southern District of New York (Mata v. Avianca, Inc.). After entrusting legal research to ChatGPT, he submitted six non-existent cases generated by AI to the court. Judge Kevin Castel ruled, “It is free to use tools, but the user is responsible for verifying the results.”
The same thing happens in coding. Even if the code written by AI “works,” you need someone who knows the fundamentals to determine whether it has correct logic, handles edge cases, and is free of security issues.
GitClear Report – The Hidden Costs of AI Code
In January 2024, software analysis company GitClear released a report analyzing 150 million lines of code. After the introduction of AI coding tools (such as Copilot), code churn increased by 39%. The number of codes that were quickly written but soon deleted or rewritten increased sharply.
The reason is clear. If the code written by AI is not reviewed by someone who understands the code’s structure, a vicious cycle of writing quickly and breaking quickly will repeat. Someone who understands concepts such as refactoring, modularization, and separation of concerns is needed to ensure that the code created by AI survives.
A Non-Major Researcher’s Data Pipeline
This is not a specific incident but a story that happens every day in research labs around the world. A researcher who knows almost nothing about Python writes data analysis code with the help of AI. It works well with 1,000 test data points. But when 1 million real data points are entered, the server crashes.
The cause is usually similar – a double for loop written by AI, i.e., code with a time complexity of O(N²). If someone knew what time complexity was, they could see the code and immediately judge that “this will break when the data grows,” and instruct AI to rewrite it with a hashmap-based O(N) solution.
Like Learning to Speak
When we learn our native language, we learn to speak naturally by listening to and imitating our parents and people around us. This is sufficient for everyday conversation.
However, at some point, we learn ABCD, learn to write, and learn grammar and sentence structure. Without this process, we cannot read contracts, write reports, or accurately convey complex logic.
Coding is the same. You can use AI to create simple things. But what is a variable, what is a function, and how does a database work? – without these fundamentals, you can’t give AI the right instructions when faced with complex problems.
So, the Way We Learn Has Also Changed
In the past, you had to study from the basics, write code, execute it, fail, and repeat the process for years.
It’s different now.
Learn the fundamentals, and then immediately dive into the vast ocean of coding with AI, based on your own domain knowledge.
If you’re a biologist, you can use it for analyzing genetic data. If you’re a chemist, you can use it for molecular simulations. If you’re a businessperson, you can use it for creating a sales dashboard. If you’re a designer, you can use it for creating an interactive portfolio – you can immediately apply coding to your area of expertise. People with a solid foundation can predict what the problem is when it arises, and people can guide AI even in structural problems that AI cannot find on its own.
AI handles the technical implementation of the coding. However, intent, direction, and planning based on domain knowledge – this is something only humans can do, and if you can give these instructions correctly, AI can implement them.
The Purpose of This Course
Let’s be honest.
Even if you complete this entire course, it may be difficult to create a website from scratch on your own. In the past, that would have meant failure, but now it doesn’t.
Build this foundation and work with AI, and it’s definitely possible.
You can forget it. You can remember it later. You can stop in the middle, create something with AI, and come back when you encounter something you don’t know.
But at least you need to understand what AI is saying. If AI says, “I will process this part asynchronously,” and you don’t know what asynchronous means and just say “yes,” you won’t be able to figure out where the problem is when it arises.
Being able to understand what AI is saying and being able to provide leadership and direction is the purpose of this course.