Building a Model Router: Picking the Cheapest Model That Still Works
A practical, step-by-step guide to building a model router that picks the cheapest LLM likely to succeed on a given request, complete with TypeScript code and fallback logic.
A practical, step-by-step guide to building a model router that picks the cheapest LLM likely to succeed on a given request, complete with TypeScript code and fallback logic.
Migrating a Node.js API to FastAPI: A Real Cost/Benefit Breakdown
I've now run this migration twice for clients adding heavy AI features to an existing Node backend. Here's the honest ledger — what it actually cost in time, what broke, and when it wasn't worth it.
WebGPU + WebXR: Rendering Realistic 3D Scenes in Browser-Based AR
Full native WebGPU rendering inside a WebXR session is still catching up across browsers. Here's the pragmatic hybrid pipeline teams are actually shipping today, and how to structure it so you can drop the hybrid part later.