The Age When Anyone Can Build Software with AI

sungwook — Tue, 23 Jun 2026 10:47:50 +0000

Hi, I'm a developer from South Korea.

Lately, I've found myself thinking about the first time I ever touched a computer.

It was 1991.

There was a small computer academy in the apartment complex where I lived. It was located on the third floor of a dim commercial building. Inside were rows of CRT monitors glowing with a faint green light. On the walls hung disassembled 8-inch and 5.25-inch floppy disks, displayed almost like scientific exhibits.

To eight-year-old me, it felt more like a laboratory or a hospital than a classroom.

The truth is, I wasn't there because I wanted to be.

My parents encouraged me to go—more specifically, my father.

Looking back, I don't think he had any particular vision about the future of computing, nor was he deeply interested in technology.

He simply wanted to play a PC game called HEXA.

Buying a computer was expensive, and sending his son to a computer academy was probably a convenient justification.

Whatever the reason, that's how I found myself learning GW-BASIC on DR-DOS at the age of eight.

Before long, a 286 PC appeared in our living room.

It had a color CRT monitor and, incredibly, a 20MB hard drive.

Unlike the computers at the academy, I didn't need to insert a boot floppy disk every time I wanted to use it.

My father spent nights playing HEXA.

I spent mine drawing triangles on the screen and making melodies with the PC speaker.

Now that I'm in my forties, I sometimes look back and realize how many technologies I've been fortunate enough to experience.

GW-BASIC, Turbo C 3.1, 80x86 Assembly, HTML, Netscape 4.2, PHP, Linux, NASM, Reverse Engineering, Hacking, Java, C#, SQL Server, MySQL, Networking, WPF, Compiler Design, Kubernetes, MongoDB, InfluxDB, Redis, Kotlin, Spring...

The list goes on.

I spent countless hours in bookstores.

Whenever I bought a technical book, I read every page and followed every example.

Over time, I stopped seeing technologies as isolated tools and began seeing the architectural patterns behind them.

Eventually, most APIs, function names, and parameter orders became second nature.

Not because I memorized them intentionally.

I had simply used them so many times that they became part of how I thought.

That accumulated experience became one of my greatest assets.

I could learn faster.

Build faster.

And solve problems that others considered impossible.

At least in my mind, very few things were truly impossible in the world of computers.

Some things were difficult.

Some things were tedious.

Some things took time.

But impossible?

Rarely.

That belief stayed with me for decades.

And then, almost overnight,

that asset disappeared.

Or perhaps "was taken away" is a more accurate description.

AI Agents arrived.

And within a single year, the software development landscape changed more dramatically than anything I had witnessed in the previous three decades.

This wasn't another programming language.

It wasn't another framework.

It wasn't even a new generation of tools.

It felt like something fundamentally different.

Not the evolution of a tool.

The replacement of a person.

Traditionally, a tool exists to help its user.

A compiler helps a developer.

An IDE helps a developer.

A framework helps a developer.

But AI Agents feel different.

They don't merely assist developers.

They increasingly perform the work that developers used to do.

For employers, this may be an incredible advancement.

For developers, the feeling is more complicated.

The gap between someone with thirty years of experience and someone with two years of experience is becoming smaller.

Of course, experience still matters.

Judgment matters.

Architectural intuition matters.

But sometimes I find myself wondering:

How much longer will that remain true?

Perhaps this is just a form of self-justification from those of us who have spent decades in the industry.

Perhaps there will soon come a day when AI designs scalable systems better than we do.

When AI writes more maintainable code than we do.

When AI chooses architectures more effectively than we do.

And if AI is doing all of that, another question emerges.

Do we even need human-readable code anymore?

Who exactly is the code being written for?

The next developer?

Or the next AI?

Faced with that thought, I started asking myself a different question.

What is AI still bad at?

What can't it do well yet?

And more importantly:

Why?

After thinking about it for a long time, I arrived at a surprisingly simple conclusion.

The problem isn't intelligence.

The problem is infrastructure.

More specifically:

There are very few platforms designed for AI itself.

The cloud platforms we use today were built for human developers.

The databases we use today were built for human developers.

Most SaaS products were built for human developers.

They provide flexibility, customization, and endless choices.

But choices require expertise.

To choose correctly, you must first understand the problem.

To answer a question, you must understand what is being asked.

Humans can do that.

AI often struggles because those platforms assume a human decision-maker exists somewhere in the process.

Meanwhile, we're entering a world where non-developers can create software simply by describing what they want.

AI can build applications surprisingly well.

But those applications often remain trapped inside a local environment.

A prototype.

A demo.

A proof of concept.

Turning them into real services introduces a new set of challenges:

User authentication
Data storage
Permissions
Sharing
External API integrations
Service operations

Suddenly, the simplicity disappears.

Users face a wall of decisions they don't know how to make.

Many eventually stop and say:

"That's enough."

And the project never becomes a real service.

That observation led me to another idea.

What if we could remove most of those decisions?

What if the backend itself was designed primarily for AI Agents rather than human developers?

That question eventually became the starting point for a project I'm currently building called Weegloo.

The idea is simple.

You build whatever you want with your favorite AI Agent.

And when you're finished, you simply say:

"Integrate it with Weegloo."

Or:

"Turn it into a service using Weegloo."

The goal is to allow the AI to handle much of the backend complexity on behalf of the user.

As I said earlier, I still believe there is very little that is truly impossible in computing.

Things are usually just difficult, tedious, or time-consuming.

Weegloo is no different.

Today it certainly cannot support every type of service imaginable.

But I believe it can continue growing toward that goal.

It will take time.

And it will probably be painful.

But that's true of most worthwhile things.

At the time of writing this article, Weegloo has been publicly available for only about a month.

It's still a very young project.

But every time I see someone build a blog, an image-generation service, or another application with an AI Agent and then connect it to backend functionality through a simple instruction, I become a little more convinced that this direction may be worth exploring.

I still remember that eight-year-old kid sitting in front of a glowing CRT monitor.

Back then, I had no idea where computers would take me.

Today, I feel something similar when I look at AI.

I don't know exactly where this technology will lead us.

I don't know what software development will look like ten years from now.

But I do know that we are witnessing one of the most significant shifts our industry has ever experienced.

And honestly?

I'm excited to see where it goes.

If you've been thinking about similar problems, I'd genuinely love to hear your thoughts.

And if you're curious about the project I'm building, you can find it by searching for Weegloo.

Thanks for reading.

DEV Community: sungwook

The Age When Anyone Can Build Software with AI