I'm learning to make my own games in Godot and currently use AI as a programming tutor. I don't want to blindly copy and paste generated code, so I ask for explanations and try to understand what each part does before using it. My goal is to be able to recreate the solution later rather than depend on the tool forever. Is this an effective and ethical way to learn game development, or would I be better off using books, courses, documentation, coding exercises, or another approach?
4 Answers
There are legitimate ethical questions around how AI systems were trained on publicly shared code, licensing, attribution, and the environmental cost of running them. However, for a learner, the more immediate concern is whether the tool is replacing your thinking. Code has long been shared through books, documentation, examples, and open-source projects, but you still need to understand the license and avoid presenting other people’s work as your own. For personal study, focus on learning the concepts and gradually writing most of the code yourself.
The main learning risk isn’t necessarily ethics or copyright—it’s that explanations can create recognition without real recall. Seeing code and understanding it line by line does not prove you could design and write it yourself. A good test is whether you can put the answer away and reproduce a simpler version, modify it, or use the same concept in a new project. If your dependence on AI decreases as your projects become harder, it is probably helping; if dependence increases, it is becoming a crutch.
For learning Godot, build small projects from scratch: a scene with player movement, a simple UI, a basic enemy, and a tiny game loop. Learn how scenes, nodes, signals, the editor, and your chosen scripting language fit together. Use official documentation and rewrite small examples in your own way. Coding challenges and deliberately imperfect toy projects will teach you more about problem-solving than assembling a large game from generated snippets.
Using AI can be reasonable if you treat it like a tutor rather than a code vending machine. Try solving the problem yourself first, then ask for conceptual hints, questions, or feedback instead of a complete implementation. Afterward, close the AI response and recreate the solution from memory. Also verify everything against Godot’s documentation, since AI can confidently give you incorrect or outdated advice.
The idea of asking for hints instead of finished code sounds much better. I’m going to try limiting myself to that so I still have to do the actual problem-solving.
You don’t need the perfect architecture immediately. Start with the simplest design you understand, then improve it when the project gives you a reason to learn a better data structure or algorithm.