Looking Back on 2025

Looking Back on 2025

There are only a few days left in 2025.

Every year I try to set some kind of challenge for myself, almost like a light goal, and this year I took on a pretty big one.

The biggest challenge was changing the way I work from being a full-time employee to becoming a freelancer.

From Full-time Employee to Freelancer

For a handful of reasons I ended up freelancing more by circumstance than by design, but ever since I started this line of work I had wanted to experience freelancing at least once, so the opportunity gave me the push I needed and turned into a great experience.

What I feel most strongly is that you never know unless you actually try.

In the past I had thought about freelancing or even starting a company several times, but most of that stayed within the realm of researching, summarizing, and running numbers in spreadsheets.

Thinking has this trait where, once you cross a certain threshold, it can make you feel as if you have already done the thing.

If all you do is think and the risks still outweigh the benefits, you probably will not act, but the thinking itself is not wasted—that is the mindset I had.

I still hold that philosophy quite strongly and have no intention of denying it. If anything, I think avoiding risks that can be avoided is a healthy stance.

Because I had already thought through it several times, when I suddenly had to change jobs this time around, the transition to freelancing went surprisingly smoothly.

Once I actually became a freelancer, though, a world I had never seen before came into view.

You have to understand your standing as a freelancer inside an organization, stay on top of taxes and social insurance, handle invoicing—things I never had to think about as an employee suddenly became my responsibility.

At the same time I could see the downside of how much time gets consumed by tasks that are unrelated to the core work.

Thanks to modern accounting software and generative AI, it is far easier now than it used to be. (I honestly feel that without ChatGPT I would have lost much more time.)

Still, I want to personally understand the rules and the calculations for each payment, so it ends up taking a fair amount of time, and that is something I only grasped after becoming a freelancer.

Ultimately, whether I conclude that this style of work does not suit me after all or that it is actually a lot of fun, the fact that I have strong conviction either way after trying it once is the biggest lesson I gained.

Coding Agents and Design

One more major trend—and the change in my mindset—that I want to write about this year is coding agents.

Generative AI has become fully mainstream, and coding agents have spread widely throughout the software industry.

Their accuracy is now impressively high.

With that backdrop, I started seriously studying software design this year.

Until now I could not really say I had studied design systematically.

Of course, I occasionally designed features at work, but most of the time I relied on prior experience to guide my design decisions.

I often limited my thinking to ensuring data consistency and called it a day.

My growing interest in design was partly driven by career considerations, but it also relates to the rise of coding agents.

From a career perspective, it has been eight full years since I started working as a software engineer.

The web is my main battlefield, but thanks to startup experience I have worked across nearly every phase of software development, from cloud infrastructure to backend development.

At the same time, I could not always explain the meaning behind each choice.

For example, even when deciding how to structure directories or where to split logic, I could not justify my decisions with confidence and often relied on intuition and vibes.

I wanted to clearly articulate that vague feeling and develop a systematic understanding of design—that was the driver behind my growing interest.

With coding agents now available, AI can generate much of the code for you.

But if you give them a sloppy prompt, you will get disposable, low-quality code.

That is where design knowledge becomes valuable.

By communicating the background and intent of the design to the AI, you can get much better code out of it.

In other words, in software development where we co-create with coding agents, I feel that design knowledge will carry even more weight.

I gave two reasons for diving into design in earnest, and after studying it I already feel the benefit of how it changes my mindset even at the individual coding level.

Looking Ahead to 2026

I feel that 2025 was full of discoveries that could grow into something more.

That means 2026 will be a phase for nurturing those buds.

When it comes to design in particular, I am not focused on specific methods so much as learning different ways of thinking and expanding my perspective. That makes the content feel difficult at times, but because I can already see the results in practice, I want to expose myself to many more design philosophies in 2026.