Central Deploy Manager
2026
The deployment plumbing behind this site and its smaller siblings: signed webhooks, health-checked Docker rollouts, and Caddy routing on one VPS.
Aquarium
2026
A small 3D aquarium for the browser, built with React and Three.js and deployed like a real app because apparently I cannot leave anything simple.
Bird of the Day
2026
A tiny daily bird site powered by recent eBird observations, a small Express API, and an unreasonable amount of affection for birds.
CentraID
2026
A six-person capstone connecting a NestJS/PostgreSQL backend, an Expo mobile app, and a classroom reader service for verified check-ins.
React Native Cloud Prototypes
2025
Two Expo/Firebase prototypes: one for vehicle access approvals and one for QR-based classroom attendance.
Wurmkickflip
2026
An extremely serious physics experiment about teaching a worm to ride a skateboard. The physics works; the worm remains a work in progress.
2026
Wurmkickflip
An extremely serious physics experiment about teaching a worm to ride a skateboard. The physics works; the worm remains a work in progress.
My notes
This started with a very reasonable question: could I teach a simulated worm to do a kickflip? It has since become a browser physics lab with procedural terrain, several questionable creatures, a skateboard, and a training setup waiting in the wings.
The default controller is still a scripted muscle wave, so I am not claiming the worm has achieved machine intelligence or even basic coordination. That is what makes it fun. The scene works, the parts flop around convincingly, and there is a clear path for testing trained policies without pretending the experiment is further along than it is.
Wurmkickflip is a React, Three.js, and Rapier prototype for experimenting with physics creatures riding a skateboard in a small procedural terrarium.
The browser app includes selectable creature shapes, terrain presets, rigid-body skateboard parts, and deterministic muscle-wave control. It also defines an observation and action contract for optional ONNX Runtime policies through WebAssembly or WebGPU, while keeping scripted control as the fast and reliable default.
A separate Python 3.11 training workspace provides Gymnasium and Stable Baselines3 scaffolding, ONNX export and validation, and an evolution experiment for creature and controller parameters. The learned-controller side is still experimental; the current public repo is best understood as a working physics playground with a deliberately unfinished training path.