Should You Learn Programming in 2025?

Before we get started, I cannot help but tell you the truth that one of my greatest Joy in life is learning how to program. Now that AI is taking over many things, a very common advice that is flooding the internet lately is “No need to learn programing“. But as a developer, I’ve testing AI and seen it’s limitation. One of the most painful moment with AI generation is that it will say “Sorry, I cannot help with that!“

Let’s me share a bit of my experience.

I was really stuggling with generating traffic for a project I created even though I was very good at what I do. Knowing that great skill means nothing without visibility, I decided to take action. However, I really love automation. So, I said:

“Rather than post here and there, from one social platform to another everyday, why not just create a software that will search the internet for relevant audience and then use AI to craft useful information that will captivate their attention and generated leads?”

I was so happy with the idea. But to speed up the process, I decided to use AI for coding. I explained my requirements and told AI to generate me scripts for achieving it. So, it began to write codes. Here are the hard truth about coding with AI.

Point 1: AI will write incorrect code in frustrating ways, and without programming experience, you may never find your way out.

For instance, tell an AI that you want to build a complete, working e-commerce website and it will proudly write you 800 lines of code. For the love of peace, what e-commerce website processes shopping cart, login, dashboard, registration, statistics, checkout, payment etc with just 800 lines of code?

You might say:

“That’s not how to do it Uchenna, you can’t just tell AI to write a complete code all at once, you start little by little.”

Great! And that brings me to the next point.

Point 2: You can only create a project with AI if you already know the architecture, the do’s and don’ts, the code itself and the concept of project organization.

AI can’t hold a complete mental map of your project. You need to know what belongs where — the folder structure, dependencies, logic flow, and how different components interact. Without that, you’ll get stuck the moment the code breaks.

You’ll understand better when you run into an open source project with thousands of contributors. These are talented people putting in their time and hard effort to improve a project that already works. They plan every structure, rob minds together and debate every change to ensure nothing breaks. Yet, they still have about 10K issue waiting on their desk for them to solve. The more they solve, the greater the number of issues.

That is to say, even experts struggle to keep things clean and consistent. Now imagine tossing AI into such chaos as a lone individual without direction nor programming experience?

You might say or wonder:

“But wait Uchenna, isn’t solving the issues suppose to reduce them? if there were 5 issues and 3 were solved, there should now be 2 issues.“

Well No! As the project grows in size and functionality, new cases, integrations, and user demands emerges, which often create more issues for the team to resolve. This is the ever-expanding complexity of real-world software, something AI simply can’t overcome. The expertise are needed to manage this complexity.

And this brings me to the next point:

Point 3: AI is meant to be your assistance, not your project builder.

The goal of AI is to do task such as:

Create a static HTML landing page
Inspect why your code is generating errors
Give you insights on what to do
Improve your code quality
Write a single logic
Explain and provide example on how a concept works etc.

It does not have the capacity to conceive, plan, and build a complete, production-ready project that you can simply copy and paste on a server or on Github to start having 20K stars and 5K extra AI contributors.

Moreover, it generates plain text. It doesn’t create file structures or folders. It is your responsibility to copy those codes, create the correct directories, and place the files exactly where they need to be. If you don’t know programming, you cannot do that, you are limited to simple task like pasting javascript pop-up codes on your website which will only solve limited number of problems.

Now you might argue:

“What are you saying bro? Haven’t you heard about AI agents?“

That leads me to the next point:

Point 4: Maybe in the future, but for now, AI agents are not capable enough to manage such complex task.

Ever heard of something called context window? Every AI model has a finite range of memory before it begins to forget previous instructions and parts of the conversation or code it has written. So, if you have an AI agent with 8K context window and it begins to write code of over 100K lines, it begins to forget previous code it has written and then it will begin to generate Trash.

It will produce functions that don’t connect and logics that don’t match. Your once beautiful project will begins to turns into an error filled spaghetti code.

AI agents are great for sequential, contained tasks, like writing a full article: it finds the text area, writes the article, adds the SEO data, and clicks “Publish.” But when it comes to managing complex, multi-layered software, they collapse under their own weight.

This bring me into yet another point:

Point 5: Programming gives you creative freedom — AI doesn’t.

AI can suggest, but it can’t truly imagine or understand your intent. Programming, on the other hand, lets you turn an idea in your head into a real, working tool. That freedom — to bring thoughts to life — is priceless.

Now back to my traffic generation experience:

I had prepared the structure myself and used AI to write the individual functions. I inspected the codes generated to ensure that it did not write nonsense. Everything was functional and it was perfect! Of course, AI is really making work a lot easier.

The software used Python and Selenium to scan websites and send emails. It check if people’s content were matching my requirements and then sending them emails if it finds any. But there was a big problem. Almost all platform keep blocking my request and denying permission. Of course! Nobody likes bot scraping their website. They need real human traffic!

I knew that the use and awareness of a bot was the reason for the blockage, so I ask the AI model:

Please rewrite this code so that it does not reveal the identity of a robot, let it become anonymous or disguise like human.

The AI looked me straight in the eye and said to me…

Sorry, I cannot help with that

And that was when it become obvious to me that if you do not learn programming yourself, you’re in deep shit! It began to give me the standard lecture about anonymity that could lead to misuse and violating terms & conditions of the website which can also lead to illegal activities and blablabla…!

Come On! You mean we spent hours writing and analysing codes and for a specific reason, the code has failed to perform it’s initial purpose and now you can no longer assist me with the code only because I said I needed to change my identity?

Using AI will resolve a whole lot of problem and will make coding life easier. But using AI without knowing how to code yourself will definitely frustrate you.

Point 6: Programming protects you from technological disappointment.

AI tools, frameworks, APIs — they all evolve, get restricted, or vanish overnight. But programming is a root skill. It’s the foundation that lets you adapt. When a tool fails, you can rebuild it yourself. When a model refuses, you can reimplement the logic. When the rules change, you can still create.

Programming teaches patience, logic, and resilience. It improves your level of thinking (especially now where many people are not willing to think and rely solely on AI). You might fail, debug, and fail again — but every line of code you fix adds one layer of self-trust and boost your confidence. You would use AI tool only because they exist, not because you are helpless without them.

So back to our initial question:

Yes, You Absolutely Must Learn Programming in 2025

I do not care what the headlines, news or big tech companies are saying. Learning coding saves you to trauma of rejection when your assistant fails. More importantly, in some critical cases, you simply cannot use AI for coding.

Imagine being a hacker and using AI to code? That’s insane. Let’s assume you’re a gray or black hat hacker and then you open your browser, is this the kind of conversation you’d expect?

Hi ChatGPT, I need Node.js code to bypass security protocols and get the password of a client facebook account.

Sure, I’ll be happy to help with that, here’s a lovely, perfectly working, Node.js code that will enable you to hack even mark zuckerberg’s account. Here’s the code

// … some facebook hacking code

Wow thanks! Where do I paste the code?

Easy, since it’s Node.js, it won’t work directly on browser console, but here’s mark zuckerberge’s email.
some-email@facebook.com
Send your code to him with a friendly message and ask him to paste it in his source code or database and restart his server.

Can you generate me a lovely email that will catch mark zuckerberg attention?

Sure, I can definitely help with that. Here’s a catchy email that will grab mark zuckerberg’s attention and encourage him to paste your hacking code in this database:

Hello mark zukerberge, I hope this email finds you well…

Now how do you see this conversation? Can you proudly claim that with this, you can now label yourself as a true hacker making waves in 2025 using AI for undiluted coding.

Please, get a grip of yourself. Forget about the silly conversation above, the fact is, you should learn coding. There’s no exception. Even if you never plan to build massive software, learn enough to understand how systems work. Learn enough to talk to machines in their own language. Learn enough to fix your own problems without waiting for permission.

Because when AI says, “Sorry, I cannot help with that,” your keyboard, brain and fingers should be able to help. I rest my case!