100% AI-generated: Our workflow with Unity 6, .NET 8 and Python
Updated: Jul 6
One of the most frequent questions we receive is: How exactly do you build a complex multiplayer voxel game entirely with AI? The short answer: With very strict guidelines.
The AI is fantastic at writing code, but it quickly loses sight of the architectural perspective. That's why I enforced a strict client/server architecture from day one.
Our tech stack at a glance:
Client: Unity 6 for rendering, input, and voxel meshing.
Server: .NET 8 Dedicated Server for chunk generation and server-authoritative UDP networking.
Logic & Code: Exclusively prompted via Claude and ChatGPT.
Assets: We had Python scripts generated that allow the AI to independently access APIs at ElevenLabs (sound) and OpenAI (textures/graphics).
Our most important principle is the MVP (Minimum Viable Product) approach. For every feature (e.g., a crafting recipe or planet generation), we force the AI to first build the smallest possible working version before expanding the feature.
This keeps the code clean and the AI's hallucinations in check.
You can view the complete code on our GitHub repository!
Comments