← Back to Projects

2025 in CodePens Recap: 10 Months of 'Creative Director' Coding

In 2025, I started doing CodePen challenges for the first time in my life. I’m somewhat familiar with coding—it’s always occupied a reasonable chunk of my career in web—but this year, I took a deep dive into creating and exploring the potential of AI assistance.

To be honest, the year was not a nice one. I think that applies to many people right now, especially those of us battling unemployment. It just so happened that I stumbled upon these challenges and thought to myself: Why not?

The Overview

I managed to maintain a pretty steady pace, completing 10 months worth of challenges. Along the way, I picked up 7 CodePen Picks and one Collection addition, so I’d say it wasn’t a bad start.

The first few months were more about getting back into the swing of things rather than finding my creative “juice.” After quite a while of not doing much coding, this served as a nice refresh. By the end of the year, I felt I had actually found some specific angles and aesthetics worth exploring and polishing.

And yes, I did use AI. It can truly speed things up or offer helpful suggestions, but like all tools, it has its grievous limitations. You still need a director’s eye to know when it’s leading you down the wrong path.

The AI Assistance

Oh, the bloody AI. It’s perhaps the most demonized or celebrated tool of them all. But once you actually understand what you are doing, you see its limits fast.

The Pros:

The biggest advantage of this collaboration? You can get the boring, basic syntax done much faster. You no longer need to write repetitive code like nth-child over and over again. It’s a massive time-saver for quick typo checks or finding that one missing semicolon that’s breaking the layout.

The Cons:

It is incredibly limited. If you ask it to suggest new features, it often suggests everything except what you actually asked for. Then there is the “comprehension gap.” My understanding of what a “cat” looks like is not at all the same as what the AI provides.

This might sound strange—people think, “Hey, it generates a pretty accurate cat if you ask for it.” But that’s the learning gap: a generated image is just a combination of millions of photos. To the AI, the cat is just a high-probability guess. If you ask for a more exotic animal, it starts to stutter.

The real problem arrives when you want to build something out of shapes. Shapes are constructed with coordinates and percentages; there isn’t enough logic in the machine to understand why a certain polygon looks “cat-like” to a human eye. It can’t see the soul of the geometry.

But ironically, even this text is using AI to check my grammar and typos.

The Learning Potential

After finding my rhythm with these challenges, I eventually started hitting those “Aha!” moments. The discipline of pushing yourself to finish really does force you to learn. It’s been interesting to observe my own progress—moving from “Oh no, where do I even start?” to approaching a project with, “I’ve done this before, maybe I can reuse that logic, but how do I make X do Z instead?”

Journey-wise, it’s been rewarding. I can feel a genuine sense of confidence building up. Beyond the code itself, this has been a great opportunity to broaden my understanding and keep pace with what’s happening in the industry. It keeps you sharp.

The 2025 Challenge Archive

Here are the deep dives into each month’s challenges:

TL;DR;

What started as a “why not” project during a difficult year of unemployment turned into a 10-month journey that yielded 7 CodePen Picks and a Collection addition, proving that a strong creative vision often resonates more than “perfect” code.

While I leaned on AI to handle the boring syntax and repetitive tasks, the experience highlighted a massive comprehension gap: a machine can guess what a cat looks like, but it cannot understand the soul or geometry of a hand-crafted shape.

Ultimately, the real growth came from the discipline of showing up, turning those early moments of being overwhelmed into a functional toolkit of logic and confidence. 2025 was the year I stopped being a “wannabe” and finally started directing the code.