Systems are easier to
understand than to explain.
paperboy is a small collection of interactive three.js / WebGPU field guides — built to turn systems that are usually just diagrams and manpages into something you can orbit, zoom into, and click apart.
Field guides
Interactive 3D explainers for systems that are usually just diagrams and manpages.
GNU / Hurd →
A live system console on the left, the machine it is describing on the right. Watch the microkernel boot one user-space server at a time, run a shell command and follow it hop by hop across the servers it touches.
Memory accounting →
What a tensor actually costs: rank, shape and strides on one flat buffer, the bit layout of float32 through nvfp4, and the trip from host RAM across the bus into GPU memory, and the arithmetic intensity that decides whether a kernel waits on the bus or on the tensor cores — every number sized live.
The LM head →
The last layer of a language model: how one hidden vector times one d×V matrix becomes a score for every word, why the layer is linear, what temperature does to the distribution, and why the “small head” is often the biggest matrix in the model.
About
I joined the GNU Hurd project back in 2021, and the learning curve was brutal — a multiserver microkernel, translators, capability-passing ports, decades of mailing-list history. This repo is the field guide I wish I'd had: an interactive explainer of the Hurd's microkernel core, built so the next person doesn't bounce off the learning curve the way I did.
It's also a small experiment about AI, authorship, and what "free software" means once you compare it to how "open" software gets built today. Read the full story on GitHub ↗