I went deep on how developers are using AI to create full games. Some are building fully functional games in as little as a few days to a few weeks.

Takeaway: Opus 5.5 is the Gold standard 🏆🏆

Games..

I Just Wanted to Fish :
( https://store.steampowered.com/app/5350560/I_Just_Wanted_to_Fish ):
This cozy simulator was built over two weeks using Claude paired with Blender MCP. The developer generated 3D assets directly within Blender without using external AI model generators.

Nightshoal :
( https://profactor.itch.io/nightshoal ):
Developed in Godot over 11 weeks, this exploration and fishing game used Claude for coding, with the developer implementing a strategy of using separate chat sessions for different tasks to maintain efficiency . Some complex models were generated using Tripo AI, while others were created procedurally.

Emberfall — Luminati
( https://broad-grass-ea4b.supportinfo-593.workers.dev/editor ):
This project utilizes a "2.5D pixel art" diorama style. It was built using Opus 5.5, with nearly every asset described through code. This approach resulted in an extremely small initial file size of only 3.5 megabytes .

How you can build:

1. Master Your AI Workflow

Don't just rely on AI to generate everything. Use it as an integrated partner.

Segmented Development: Like the developer of Nightshoal, create separate chat sessions or environments for specific tasks, such as one for core mechanics and one for visual assets. This keeps token usage efficient and the AI focused on the problem at hand.

Iterative Prompting: Use an AI coding assistant like Opus 5.5 to build mechanics one at a time. Start with a foundational architecture before asking the AI for deep, complex features.

Specialized Tooling: Combine AI coding with tools like Tripo AI for 3D generation or Blender MCP to keep assets consistent and game-ready without leaving your main production environment.

2. Prioritize Design Foundations

Even the best AI can't replace a solid design vision.

Define the Experience: Focus on the emotions you want players to feel. Before building, define your game's core loop: what is the main action the player takes that is actually fun?

Start Small: Use the game jam mindset. Build playable prototypes quickly instead of getting bogged down in large, complex systems.

Learn the Basics: You don't have to be an expert coder, but understanding basic programming principles like clean code and system architecture will help you catch AI errors and debug more effectively.

3. Professional Best Practices

Source Control: Always use version control like Git. When you build with AI, code can change rapidly, and you need a way to revert when something breaks.

Clear Context: Give the AI as much relevant context as possible. Explain your game's genre, your intended art style, and the limitations of your chosen engine (Godot, Unity, etc.) to get higher-quality output.

Iterate and Playtest: Don't wait until the game is finished to play it. Play your own game constantly during development to feel the gameplay loop and refine the controls.