Tech Alchemy

This interactive piece called Neural Network by Kim Seonghyun is pretty cool—it’s from Design Korea and shows how a neural network actually works, but in a visual, real-time way.

It reacts to people and generates visuals that mimic how the brain processes information, going through steps like analyzing, classifying, and predicting. The part shown here is the inference stage, where a trained model takes input and gives an output.

What’s interesting is how it was built—the creator () learned the basics of deep learning, trained a model using the MNIST dataset with PyTorch, then turned it into a visual experience using Processing. They also used OpenGL instancing so it can handle rendering a lot of elements smoothly.

It’s a really nice way of taking something that usually lives in code and making it something you can actually see and understand.

What do you think about it? 🤔💬

This project uses OpenStreetMap data to recreate real-world environments inside Minecraft, generating roads, buildings, and terrain based on actual coordinates.

You can build your neighborhood, a full city, or any area block by block to match real life. It’s also fully open source, so you can run it locally, customize it, or build on top of it.

GitHub link: https://github.com/louis-e/arnis