From the Blog

I built a public step counter using Google Fit API and Cloudflare Workers. It shows my daily steps, weekly totals, and streak on my website. Here's how I did it and why public accountability works.

Learn how to run open source AI models (LLMs) directly on your computer. No cloud, no subscriptions, complete privacy. A beginner-friendly guide using llama.cpp.

I got tired of context switching between Drupal projects, so I built an open source MCP server that gives AI assistants instant knowledge about your site. Here's how I use one server for all my projects.

I wanted to build a pronunciation coaching AI and tried training Google's Gemma 3n locally on my M4 Pro MacBook. After multiple attempts and memory errors, I learned that even 48GB of RAM isn't enough. Here's what I tried and why I'm moving to Vertex AI.

Learn how to securely expose your local DDEV sites to the internet using Cloudflare Tunnel, without opening firewall ports or configuring complex networking.

Google released a model that can see your screen and click around like a person. I used Claude Code to build a test in 15 minutes. Now an AI checks if my translations load correctly.

Found a subtle bug where translated blog posts loaded empty in production. The issue: calling getPostBySlug with 'en' locale to access translatedSlugs failed because the manifest doesn't store English in the localized object.

I often use AI while reading on my Kindle, but there wasn't an easy way to ask questions without switching devices. Here's how I built Kindle-ChatGPT, a browser-based AI assistant optimized for e-ink displays that runs entirely on your Kindle.

Nginx shows 502 Bad Gateway, PHP-FPM won't start, and the usual config test passes. The real issue is often a stuck FPM process holding the socket. Here's how to find it and fix it safely.

If you change block output based on the current path, you must add the `url.path` cache context. Otherwise Drupal will reuse a cached version across pages and show the wrong content.