Bonsai Garden

Bonsai Garden is a gamified tree-growing simulation where you can spend your points earned from Energy Planner interactions to tend to miniature trees, and watch them grow over time. Making the trees look anything like the real thing — with authentic branching patterns, trunk shapes, and species-specific characteristics — turned out to be a fascinating design challenge.

I put a lot of thought into how to guide the AI effectively on research for each species. Rather than asking broad questions, I gave specific direction on what botanical factors mattered most: branching angles, bark texture variation, seasonal foliage. This focus produced much richer species parameters than open-ended prompting would have.

One of my favourite moments was the watering can cursor attempt. The first pass produced two entirely dysfunctional-looking cans — a tiny one for the cursor and a random large one stuck in the corner, with no water droplets anywhere in sight. I've kept those early screenshots as a reminder that when an initial request is unclear, the AI will sometimes implement something entirely different rather than ask for clarification.

This project also gave me my first real hands-on experience with Claude hooks and the remote control feature, which opened up exciting new possibilities for agent-driven development workflows.