Forum Discussion
Threshold Reef - Interactive Ecosystem Simulator
You control three environmental pressures: temperature, fishing, and pollution - with sliders, and watch the reef respond in real time. Fish swim away when overfished, coral bleaches under heat, and algae spreads when the grazers disappear.
For example: you crank up fishing, the fish leave, and suddenly algae explodes everywhere. That's when it clicks - the fish weren't just decoration, they were eating the algae and protecting the reef. One change triggers a chain reaction. There's a bunch of these interactions to discover.
It's a simulation of how complex systems work - how pressures don't just add up, they multiply. While this one is about coral reefs, the same approach could be adapted to other ecosystems, or really any system where variables interact in non-obvious ways - supply chains, organizational dynamics, market forces. You could also dial up the complexity significantly; I kept this version streamlined for clarity, but the architecture can handle much more.
What Inspired This
I'm a biologist (PhD) moving into instructional design. I'm building a coral reef course in Rise 360 that takes learners from the basics (what reefs are, why they matter), through the threats they face, into a hands-on scenario as a reef manager, and finally into how to actually get involved - as a career, volunteer, or supporter.
Threshold Reef came out of that course. I wanted learners to explore how different pressures interact in a playful, visual way - just mess with the sliders and see what happens.
How I Built It
For Rise 360 Code Blocks, you can use any LLM. I used Claude (Opus 4.5 and 4.6). My usual workflow: I create a Project in Claude, do some upfront research on Rise 360 Code Block constraints and what to keep in mind, and drop that into the project as background knowledge along with a prompt that sets up the context, that we're building Rise 360 Code Blocks together. Then I start coding within that project.
Before any actual code, I brainstorm with the model. What are we building, what does it need to do, how could it look? Sometimes I'll even ask a different LLM for its take on the approach - kind of a feedback loop. Once the direction is clear, it's back and forth: I describe what I want, review the output, test it, describe what to change along with a screenshot of the application.
For prompting, I use a voice-to-text app called Wispr Flow. I don't write structured prompts - I just talk. I look at what's on screen and describe what I want, what's off, what should change. Over the past years I've gotten a feel for how to communicate with AI, but honestly it's mostly just explaining things naturally, the way you'd talk to anyone you're working with on a project.
What I Learned
Make sure your LLM knows the Rise 360 constraints upfront - sandboxed iframe, no CDN links, everything in one HTML file. Without that, you get code that silently breaks.
Start simple and iterate visually. My first version was scientifically rigorous and completely overwhelming. The current version uses direct cause-and-effect logic, and it teaches better. Look at the actual output, describe what's wrong ("the corals look like lollipops" etc.), and fix from there.
.