I Became Stronger, So Why Can I No Longer Make Games?#

Before#

Last summer was when I had only just begun seriously learning Godot.

I knew almost nothing. I followed tutorials to set up the environment, learned Godot by following tutorials, searched whenever I did not know how to write the code, and asked the community when the tutorials could not solve the problem.

Later I joined Game Jams. AI once broke a project I was working on; I managed to break Git myself; at one point I even lost three days of development.

Looking back, I was quite bad then. My code was messy, the architecture was barely an architecture, I did not know how to implement many effects, and I could not make art or music.

But there was one simple fact: I was the one making the game.

I knew why the scene was built that way. I knew roughly what a script was doing. When something went wrong, I often had no idea what to do, but I still had to search for the answer and use my own strange little brain to make it work. Creation was painful and happy at the same time. Even browsing an asset store when bored could feel exciting.

At the end of last August, I noticed that Hatsune Miku’s birthday was coming up, so I spent a few days making a small game and uploaded it to Bilibili.

AI was nowhere near as capable as it is now, and I knew less than I do today, but I could still think through a theme and finish the whole thing.

This year I had stronger tools and was better at using AI. I should have been stronger than before, yet I ended up with an unfinished failure.

Now#

Looking back, perhaps I have already experienced two major collapses of long-term game projects. Each failure changed me in its own way.

After the first one, I decided that I simply did not understand game development well enough, so I tried Game Jams one after another, from November 2025 to March 2026. I probably joined more than ten. Eventually I stopped wanting to do quite so many, and, full of confidence, started a second long-term project.

That project eventually became boring too. I escaped back to Game Jams, but this time they left me feeling even emptier. To fill the emptiness, I branched out from game development into servers, websites, videos, communities, open source, AI, drawing, music, and many other fields.

In June, this still felt wonderful. Everything seemed connected: I began using AI to make games; AI led me into communities; the communities introduced me to servers, websites, open source, and more strange areas. But a person’s energy is limited.

Even when I was learning and collecting useful resources, I could not explore every field by leaning hard on AI. Worse, I was getting farther and farther from my original goal.

I became increasingly good at writing prompts and describing requirements clearly. Eventually that became counterproductive: when AI did something badly, I started treating it like a person and insulting it. Insulting a tool? What an incapable thing to do.

When I reread my July diary, one sentence felt especially sharp: “Keep playing the GodotHub game and let Codex fix it later.” At the time I saw nothing wrong with that. Slowly, I had changed from “the person making the game” into a strange kind of product manager. The loop was probably efficient.

But the problem was this: I was gradually losing the ability to understand why my game had become what it was.

At GMTK Game Jam, the problem exploded.²

There was not enough time. The gameplay needed to change quickly. The game was not fun. Yet when I looked at the scenes and code AI had assembled, I did not even know where to begin modifying them. I had not written the important code myself for too long.

When I finally had to rescue the project, I discovered that I was the author of the game, but I did not have enough control over it. In the end I cut away a lot and submitted it anyway.

I thought I had learned my lesson. Then, at the very end of summer, I started another Hatsune Miku game and failed to finish it.

Last year, the version of me who knew almost nothing finished a Miku game in a few days.

This year, the version of me with stronger AI, more Godot experience, more plugins, and more knowledge could not finish one.

That hurt more than I expected. I had to admit that I really did become stronger, but some important ability had become weaker.

I Outsourced Too Much of My Brain#

I used AI last year too. The Miku game was not made without advice from AI. So why did this year go worse?

Maybe I outsourced too much of my brain.

In the past, I would think through a problem myself and then bring my conclusion to AI, asking it to think with me.

Now I bring AI the demand to think. There is a big difference between sending it a conclusion that I have participated in forming and sending it a request whose conclusion I have not helped create.

Before, I wanted to implement something, researched the principles together with AI, and tried to understand them as we went. My work was rough, but I had complete control over it. Now I keep describing what I want and hoping AI will move everything in that direction. When it does not, all I can do is get angry at it. It is a sad way to work.

The difference is larger than it looks. Perhaps this is how I slowly started thinking less and less.

Does That Mean I Will Stop Using AI?#

Of course not.

If I really had to return to the age of searching for every fact myself, hand-writing repetitive code, and organizing tens of thousands of words of material on my own, I cannot imagine it. It would be like asking a modern person to return to the primitive world.

AI is genuinely useful, and it will only get stronger. I do not want to pretend to be a “purely handmade creator.” The question has never been whether AI is allowed to participate.

The question is: which parts can be outsourced, and which parts cannot?

Server setup, repetitive work, research organization, mechanical code, and tools I do not plan to master long term can all be handed to AI.

But if something is a skill I truly want to own — game design, Godot, core code, scene construction, and creative judgment — I must leave myself room to think.

Most importantly: do not outsource confusion.

When I do not know how to solve something, I should not ask AI to hand me the answer in the first second. I can investigate first, try a version first, and build a model of the problem first.

Then I can ask AI: where is my reasoning wrong? Is there a better way?

That way AI can help me move quickly, while I remain the person walking.

The Other Mountains#

I have written before that I like many things. Game development is one mountain. Servers are another. Videos, websites, open source, drawing, and music are all different mountains too.

I do not want to discover that I walked in the wrong direction and then blow up every other mountain, telling myself, “From today on, I can only make games.”

Those other fields are assets to me now. They do require time to maintain, but I am still willing to keep them.

What I truly need to remember is this: when I decide to enter the game-development mountain, I have to be present in the making.

Summer Is Over#

So was this summer good or bad?

I still feel that it was not as good as I imagined:

• I did not learn many of the things I originally planned to learn.
• I did not make as many videos as I wanted.
• Game development did not progress as far as I expected.
• I did not finish the final Miku game.
• There is still a make-up exam waiting for me (sad).

I kept telling myself that I needed to become stronger. I studied new things, looked for stronger tools, stronger models, and more efficient methods.

After walking in a circle, I realized that perhaps I had taken the wrong road. Stronger tools and more resources do not automatically make me a better creator.

What I have always wanted to become is a content creator, and the medium I most want to create with is games.

The version of me last year who knew nothing and built games one Godot tutorial at a time was not stronger than the current me. He was simply forced to face each problem himself, and that brought him closer to the person he wanted to become.

Now I have more choices and more resources, so I need to learn something I never had to learn before: when not to press the “do it for me” button, and to think for myself instead.

I Want to Believe Again#

As I write this, I suddenly realize that last year’s version of me had something I am slowly losing: determination.

Back then I really believed I could make a good game, perhaps even a game that sold well. Then two projects were interrupted. After the first, I buried myself in Game Jams and used small games to maintain the feeling that I was still making games. Along the way I met people. After the second, I looked toward a larger world: communities, new fields, more people, and more things made with AI.

None of that was meaningless. I learned a lot and saw things I could not see before. But the more I saw, the more clearly I saw my weaknesses: I had not played enough games, could not make art, had immature taste, did not know how to compose music, and lacked experience with stories. Even programming, once the skill I trusted most, was being changed by AI.

So I kept adding equipment. I studied art, joined Jams, researched AI, joined communities, met other creators, and explored every field that caught my interest. I thought I was armoring myself. Looking back, part of that armor was also a way to hide.

I no longer believed in the same simple way that I could really make the game.

I kept thinking that I would return after learning a little more, preparing a little better, and becoming a little stronger. Preparation slowly became a form of escape. Especially after returning from Hong Kong, I seemed trapped in that state: doing many things, meeting many people, touching many tools, but rarely moving toward the original goal.

I do not want to continue like this.

I cannot become the person who knew nothing and therefore dared to believe everything. The two interruptions really happened, and I now know that independent game development is much more complicated than I once imagined. That is exactly why I do not need to pretend those failures never happened. I need a new kind of confidence.

Not “I can do everything.” Not “this game will definitely sell.” Not “I will make something perfect.”

Instead:

This time, I will finish it.

Even if it is not as good as I imagined, even if the art is still immature, the music still needs outside assets, the story still has flaws, and not many people play it, I first need to prove one thing again: I am someone who can finish his own game.

AI should not walk the most important part of the road for me anymore. From now on, I want to define the problem myself, make my own first version, and make the key design decisions myself. If I do not know something, I will hit the wall once or twice before asking AI to explain, research, review, or suggest options. AI can be a toolbox, a dashboard, or a repair manual, but I have to become the driver again.

The last year was not entirely wasted. The Jams, communities, friends, AI, art practice, and all the exploration were real experience. The problem was simply that I walked too far away and almost forgot why I started.

Now I remember.

I still want to make games. I still want to make a truly good game. I do not want to throw away the old, naïve dreams of making a game that sells well or moves someone through its story.

This time I will change the order: first prove that I can finish, then prove that I can make it good, and only after that reach for the farther goal. Before any of that, I need to do the next thing well, study, pass the make-up exam, and come back. This is the third time.

This time, damn it, I will finish.

That is enough of this summer summary for now.

I need to go study for my make-up exam.

Some things, after all, still cannot be delegated to AI — like taking an exam. (laughs)