Summary & Key Takeaways ​

• Simon Willison experimented with Claude Opus 5.5 Max to generate a game.
• The AI created "Tree Appreciator," a Three.js-based spot-the-difference game.
• Claude spent over three hours generating the game's assets and logic.
• The prompt included "make sure it is fun," which yielded compelling results.
• The game features music, animation, and an achievement system.
• This showcases LLMs' advanced capabilities in creative and complex game design.
• It highlights the surprising effectiveness of high-level, abstract prompts.

Our Commentary ​

"Make sure it is fun." That's the prompt. And Claude delivered. We're genuinely a little unnerved by how effective that was. It's one thing for an LLM to write code, but to grasp and implement "fun" in a game? That's a whole different ballgame. I'm left wondering what other abstract concepts these models can now interpret and execute. The implications for creative industries are immense, and honestly, a bit scary. Where does human creativity fit when an AI can just... make something fun?