What I Learned Trying to Make a Game with AI — Only Half the Truth About 'Claude for Game Dev'

#ai #phaser

I initially wanted to make a side-scrolling game like MapleStory. YouTube was flooded with "I made a game with AI (Claude)," so I thought it would be easy. But when I tried it myself – it turned out that people with existing game development knowledge were just using AI to improve quality and speed, not that you could just "make it for me" without any knowledge. In the end, what I completed wasn't a playable game, but an "auto-battle" spectator game (like raising a mushroom) that you just watch.
This post is about that honest journey – where I got stuck, why I pivoted, and what I learned. (And you can try out the completed version via the 🎮 Play Now link below.)

I'm a developer in Korea building an AI chatbot alone. I only write about things I've actually tried and experienced.## 1. The First Wall — AI-Generated Characters Can't 'Walk'
Moving characters are essential for games like MapleStory. So, I first tried AI image generation (gpt-image) to create chibi characters and then generated walk cycles (4 frames of walking animation) for them.
This is where I got stuck. With each frame, the character subtly became a different character – the color of the clothes, the proportions, the face all changed slightly between frames 1, 2, 3, and 4. When stitched together in a game, the character wouldn't walk; it would just tremble erratically.

This is exactly the same ceiling I hit in Making AI Videos (Dev Log #3) – AI image generation cannot create consistent character animation (multi-frame movement). The same wall in videos, the same wall in games.

2. Pivot ① — Abandoning AI Characters for Pre-made Sprites

So, my first surrender: I gave up on AI-generated characters and switched to CC0 licensed pixel art sprites. The protagonist became Hero Knight (by LuizMelo), and the monsters became Pixel Adventure (by Pixel Frog) – these are consistent animation sheets drawn by hand by game artists, so the frames don't wobble.
Lesson learned: "Making game assets with AI" isn't viable for character animation yet. For things like skill effects (barriers/sparks) that are static/procedural VFX, I could code them myself, but for moving characters, pre-made assets were the answer.## 3. Pivot ② — Abandoning Playable (MapleStory-like) for 'Spectator Auto-Battler'
A bigger pivot was still ahead. MapleStory is a game where you directly control your character, but for a playable game, physics, collision, input, level design… it was too much for me to handle alone with AI. So, I changed the genre entirely – to an "auto-battle" spectator game (Mushroom-growing style). The character advances and auto-battles on its own, and I just watch. Not having to control it drastically reduced the burden of animation and controls.
But this "second best option" ended up becoming even more "me." The monsters the character encounters are populated with actual development data:

"It's not about creating something amazing, but about how you overcome challenges in that environment." — f-string crash loops, deployment OOMs, 100% disk usage – these major disasters become bosses, and the defensive measures I created to survive them become skills. The systems I built (chatbot, NFT, autonomous lab...) become milestone monsters + equipment. The game itself became a record of my build-in-public journey.

4. The Honest Conclusion — Only Half True About "Games Made with Claude"

Looking back, this is the core lesson:

"I made a game with AI" on YouTube is mostly the result of someone with existing game development knowledge using AI to enhance quality and speed. You can't make a proper game by just saying "make it for me" without that knowledge. AI is a tool to amplify your skills, not to replace foundational knowledge.
Still, I have no regrets. My mindset has always been:
Try it first. If the best (MapleStory, AI characters) gets blocked, do the next best thing (pre-made sprites, auto-battler) first, and keep tracking and documenting the technology.
And that "next best thing" has become a realistic outcome, not a mirage – and in a way, more "me" than the original goal (playable). I confirmed the animation ceiling of AI images (the same wall as in video #3) in a game, and I learned what kind of design circumvents that limitation.
🎮 Try out the finished version yourself → /play (Riel's actual issues, disasters, and improvements flow as an auto-battler.)