Is It Still Worth Learning to Code? Fewer People Think So. That's the Opportunity.

Ask a teenager today if they want to learn programming and you’ll often hear: “Why? AI does that now.”

They’re not alone. Coding bootcamps are quieter, programming courses have fewer students, and for the first time in about twenty years, fewer people are choosing computer science. My answer is still yes, learn to code. But the reason has changed, and so has the way you should learn it.

The students are leaving

The numbers are clear. According to the National Student Clearinghouse, undergraduate enrollment in computer and information sciences at four-year colleges fell 8.4% in spring 2026. Over the same period, overall undergraduate enrollment grew 1.3%. Students didn’t stop studying. They just picked something else.

The reasons make sense from the outside. AI writes working code from a sentence. Junior job postings got harder to find. The headlines say programming is “solved”.

But follow the logic a few years forward.

Fewer people learn to code

AI generates more code than ever

→

Fewer people who understand it

Supply goes down. The amount of software that needs maintaining, securing and fixing goes up. That’s not the end of a career path. That’s a shortage on its way.

”AI writes code” isn’t the same as “AI replaces understanding”

Here’s code an AI will produce in seconds for a search box. It works in the demo.

async function search(query: string, render: (items: string[]) => void) {
  const response = await fetch(`/api/search?q=${query}`);
  const items: string[] = await response.json();
  render(items);
}

Type “an”, then quickly “angular”. Two requests fly out. If the first one comes back last, the screen shows results for “an” while the input says “angular”. The query isn’t encoded either, so a & or # breaks the request.

let controller: AbortController | undefined;

async function search(query: string, render: (items: string[]) => void) {
  controller?.abort();
  controller = new AbortController();
  const url = `/api/search?q=${encodeURIComponent(query)}`;
  try {
    const response = await fetch(url, { signal: controller.signal });
    render(await response.json());
  } catch (error) {
    if ((error as Error).name !== "AbortError") throw error;
  }
}

To spot that bug, you need to understand async code, network timing and URLs. No prompt teaches you that. You learn it by writing code, breaking it and fixing it. The person who can’t read code can’t tell when the AI is wrong.

What’s worth learning now

Learning to code in 2026 isn’t about memorising syntax. AI is great at syntax. It’s about the things underneath.

Matters less now	Matters more now
Memorising syntax and APIs	Reading and judging code fast
Typing boilerplate quickly	Debugging and asking "why?"
Knowing one framework deeply	Fundamentals: data, state, async, networks
Writing every line yourself	Designing systems and splitting problems
Solo coding speed	Explaining trade-offs to people

The left column is what AI took over. The right column is what makes someone worth hiring.

How to learn to code with AI around

The trap for new learners is letting AI do the learning for them. You get working projects and zero understanding. Here’s a better order.

Write it yourself first

Small programs, no AI. Feel the struggle.

Use AI as a tutor

Ask it to explain, not to solve.

Review AI code

Find the bug before running it.

Build with AI

Now it multiplies skill you actually have.

The order matters. AI multiplies what you already know. Zero times ten is still zero.

Key insight: When everyone has AI, the advantage goes to the people who can tell good code from almost-right code. That skill only comes from having written code yourself.

Will people still learn to code?

Yes — just fewer of them, at least for a while. And that’s exactly why it’s a good time to start.

We’ve seen this cycle before. After the dot-com crash, people said programming was finished, and computer science enrollment dropped for years. Then demand exploded with mobile and the cloud, and companies spent the next decade fighting over developers.

Today, people who learn to code properly will enter a market with fewer competitors and far more software to care for. That’s not a bad bet. It might be one of the best ones available.

The takeaway

Learning to code is still worth it. Learning to type code isn’t. Focus on understanding: how programs run, why they break, and how to judge what an AI hands you.

If you’re starting today, try this: build one small project — a to-do app, a calculator, anything — without AI. Then ask an AI to build the same thing and compare. Every difference you can explain is something you’ve really learned.

Sources:

National Student Clearinghouse — Computer Science Enrollment Is Cooling — spring 2026 enrollment decline in computer and information sciences
MDN — AbortController — cancelling in-flight fetch requests