A developer experimented with Claude Opus 5.5 Max by asking the model to build a Three.js game called "Tree Appreciator." The requested concept was a 3D game that displays two algorithmically generated trees and centers gameplay on noticing and appreciating the differences between them.
Experiment details
- Tool: the experiment used the default Claude iPhone app in chat mode (not Claude Code mode).
- Prompt: "Build a three js game called 'Tree appreciator' - it shows you two algorithmically generated trees and the game mechanics is about appreciating the differences between them. Think hard about how to make sure it is both fun and beautiful." The developer provided a short, high-level instruction.
- Model: Claude Opus 5.5 Max.
Outcome and timing
Claude processed the request for about 3 hours and 7 minutes before returning a playable prototype. The output was described as a cozy spot-the-difference style game that included music, animations, and an achievements system.
Context and significance
The developer noted some embarrassment that telling the model to "make sure it is fun" appeared to work surprisingly well. The experiment illustrates that contemporary large language models can transform concise, high-level prompts into relatively polished interactive prototypes after multi-hour computation.
The report does not provide the generated code's exact structure, licensing details, runtime environment, or a specific date. Nonetheless, it serves as a concrete example of how generative models can produce game-design concepts and working prototypes from brief instructions.
