The core idea is almost stupidly elegant: take the golden angle — the irrational rotation that nature uses to pack seeds into sunflower heads — and use it to place 89 RGB LEDs in a spiral. Voronoi-tessellate the point cloud to get cell boundaries. 3D print the walls. Solder an LED into each cell. Diffuse the light through mulberry paper. The result is a physical object that looks like it grew rather than was manufactured, which is the whole point. What makes this more than a weekend project is the layered craft. The author wrote the phyllotaxis generation in Processing, exported geometry to Python's CadQuery for parametric CAD, split the shape into printable quadrants, post-processed in FreeCAD, designed a custom PCB, wrote firmware for an STM32 with ARM DSP instructions to run real-time FFT on audio from an I2S mic, and eventually rewrote everything in Rust on an ESP32 with a WebAssembly-based sketch upload system. That's five or six distinct technical domains woven into one artifact. The audio-reactive behavior is where the project earns its keep. Running a Fourier transform on-chip, splitting into frequency bands, applying auto-gain control, and mapping energy to a radial brightness pulse — that's a real signal processing pipeline, not a VU meter with extra steps. The Jon Hopkins demo video shows the thing breathing with the music in a way that feels genuinely organic rather than just reactive. The second version is where the engineering maturity shows. Switching from 3D-printed LED mounts to PCBs, exploiting 5-fold symmetry to match PCB batch quantities, learning PCB design from scratch, and ultimately leaving the finished piece as a communal installation at Recurse Center with a web interface for user-uploaded sketches — that's a project that kept teaching its maker new skills at every turn. The honesty about friction is refreshing. Hand-soldering neopixels with bottom-mount pads is described as a nightmare, not a fun challenge. The paper faceplate is fragile. The perfboard electronics were janky. Version 2 went on hiatus. This is real maker documentation, not a highlight reel. The explicit list of unsolved problems — easier LED types, PCB assembly services, acrylic protection layers, magnetic faceplates — reads like a genuine engineering backlog rather than a teaser for a Kickstarter. The write-up itself functions as a tutorial-grade walkthrough without ever becoming one. Code snippets are short and annotated. The Processing-to-CadQuery-to-FreeCAD pipeline is explained clearly enough that someone with moderate maker skills could attempt a similar project, though the breadth of tools involved is a real barrier to entry. The fragment-shader analogy for LED programming — iterating over cells with a position LUT and writing per-pixel color logic — is a genuinely useful mental model. As a creative object, this sits in the sweet spot between generative art and functional design. It's not trying to be a product yet (though the author clearly wants it to be), and it's not purely conceptual. It's a working thing in a room that reacts to sound and looks beautiful, built by one person learning as they went. That's the best version of what maker culture produces.
