Blog
Sharing insights about front-end and full-stack development with TypeScript, Node.js, React, and friends.
AI Framework Comparison: Vercel AI SDK, Mastra, Langchain and Genkit
I built the same examples with Vercel AI SDK, Mastra, Langchain.js and Genkit. Vercel AI SDK won for simplicity, with Mastra a close second.
Build Your Own AI-Powered API with Ollama, Hono and Vercel AI SDK
In this post, I'll explain my tech choice and show you three approaches for building AI-powered APIs.
React-Select Search Box: Searchable Dropdown with Remote Data and Caching
Turn a React-Select dropdown into a searchable autocomplete box with custom styles, remote data, debouncing, and React Query caching.
Pipe Streams in Node in the Right Way
Pipe a file to an HTTP response in Node.js without leaking file handles when clients disconnect, crashing on read errors, or hitting ERR_STREAM_UNABLE_TO_PIPE.
Using Small LLM Models Locally with Ollama
The smallest models in the Ollama library are smollm2:135m (271 MB) and gemma3:270m (292 MB). Here's how to install Ollama, run small models, and call them from JS.
AI Image Generation Cost Analysis
Picking the wrong image generation model can cost you 34x more and take 12x longer. I ran the same prompt through 31 models from 8 providers to find out.
Moving PreVibe from Vercel to a DigitalOcean VPS
Why I moved PreVibe's dashboard and MCP server to a VPS, how I handled the cutover, and the deployment problems I worked through along the way.