The Stronger the AI, the Harder the Game? A 19-Year-Old Indie Dev’s Summer Reflection
The more powerful the tools, the harder it becomes to finish a project on your own. This counterintuitive conclusion came from Amiya after reflecting on her summer project: both AI and Godot are stronger than last year, yet she only managed to create an unfinished failure. The core issue wasn’t a lack of technical skill, but rather that she outsourced “thinking” to AI, downgrading from a creator to a “product manager” who merely submits requirements—ultimately losing control over the output.
Lessons Learned Twice a Year
Last summer, she learned Godot for the first time properly. Setting up the environment required tutorials, and when she didn’t understand the code, she figured it out herself. Though the code was messy, the architecture poor, and the art nonexistent, she made the game herself—she knew why the scene was built that way, what each script was doing, and when things broke, she used every bit of her own “stubborn brain” to find answers. Before Hatsune Miku’s birthday, she spent a few days making a mini-game on a whim and posted it on Bilibili. Even though she knew less, she could conceive the idea fully in her head and bring it to life completely.
This year, everything flipped. Armed with stronger tools, deeper familiarity with Godot, more plugins, and greater fluency with AI, she still couldn’t finish her game.
Two Critical Loss of Control
After the first collapse, she joined over a dozen Game Jams. The second collapse happened during GMTK Game Jam: time ran out, gameplay needed tweaking, and facing the AI-generated scenes and code, she didn’t even know where to start modifying. In the end, she cut away half the features and submitted what remained. The mini-game she’d started around Miku’s birthday still wasn’t finished.
From “Producer” to “Product Manager”
The crux of the problem lies in a qualitative shift in workflow. Before, she’d think things through herself first, then bring her conclusions to AI for further exploration. Now, she feeds raw requirements directly into AI, getting increasingly adept at writing prompts—and sometimes even lashing out at the tools. When she read her July journal entry that said, “Let Codex handle it,” she realized something: this loop is efficient, but she’s gradually losing sight of why the game is turning out this way.
Preparation as Procrastination
To fill the void, she branched out from game development into servers, websites, communities, open source, AI, drawing, music… In June, it felt magical because everything seemed interconnected. But human energy is limited. She can’t explore every field, and the gap between her current path and her original goal keeps widening. She began to see that she was building armor for herself—always feeling like if she just learned a bit more, prepared a little better, she could come back. Preparation itself gradually became a form of avoidance.
The Fix: Don’t Outsource Confusion
She isn’t trying to return to the era of hand-coding every repetitive line. The issue has never been whether AI participates, but what can be outsourced and what cannot.
- Can delegate to AI: server setup, repetitive tasks, research organization, mechanical coding
- Must keep thinking space yourself: game design, Godot core code, scene building, creative judgment
Her new workflow looks like this:
- When you hit a problem you don’t know how to solve, don’t ask AI in the first second. Instead, investigate yourself first, try a prototype, sketch out a mental model.
- Only then ask AI: “Where did I go wrong? Is there a better approach?”
- This way, AI can help you move fast—but you’re still the one walking the path.
The more powerful the tools become and the more resources are available, the more automatically they don’t make you a better creator. This time, she decided to stop pretending the breaks in progress don’t exist, and instead build a new kind of confidence: no matter how unpolished the art, finish it.
Source: Amiya’s Desk
Source · Amiya’s Desk: Read original post →