Writing
Field notes on AI integration, WebGL performance, and shipping high-craft web work.
Next.js • Edge • LLM
Streaming LLM responses from the edge
Why the assistant on this site answers in real time — and the small amount of code that makes token-by-token streaming feel instant.
2 min readWebGL • Performance • React Three Fiber
A WebGL hero that doesn't melt mobile
Three techniques — lazy canvases, adaptive resolution, and visibility-gated render loops — that keep an interactive hero at 60fps on a mid-range phone.
2 min read