Building a production-ready hybrid search for static blogs - combining BM25 keyword search with OpenAI embeddings, all without a backend server.

Building a RAG-powered AI assistant for my personal website using hybrid search, Cloudflare Workers, and OpenAI API.

Adding real-time AI translation to a static blog with GPT-4o-mini and Cloudflare Workers, supporting 10 languages with dual-layer caching.

Documenting the overnight journey of migrating a shell script-based static site generator to Deno

Building a zero-backend hybrid search engine for static blogs: BM25 keyword retrieval + browser-side semantic expansion with cross-lingual Chinese-English support

静态博客语义搜索架构升级方案：计划将浏览器端 ONNX 推理迁移到 Cloudflare Workers 边缘计算，预期客户端下载从 156MB 降至 2.6MB，彻底解决移动端崩溃问题

Moving blog semantic search from browser ONNX to Cloudflare Workers while keeping the browser approach as a switchable option

Implementing passwordless email OTP login, dual-mode commenting (explicit button + AI auto-detection), and a build-integrated comment pipeline for a personal website chat widget

Full-stack architecture upgrade from free tier to $5 production-grade solution, covering AI assistant, semantic search, and graffiti comment system

A creative approach to transforming traditional comments into graffiti-style sticky notes, featuring hybrid storage architecture and anchor positioning

How I used git-crypt to seamlessly merge my private notes and public blog into a single, secure, version-controlled repository on GitHub.

Password: openthedoor — How I built LUKS-style multi-password article encryption for a static GitHub Pages blog.

Building a fitness data visualization system for a static blog: parsing FIT/GPX/TCX files from an Amazfit smartwatch, creating Apple Fitness-style SVG rings, interactive GPS track maps with Leaflet.js, and a monthly activity calendar

A practical guide to implementing AI Engine Optimization (AEO) on a static blog — sitemap.xml, llms.txt, JSON-LD, Atom feed, IndexNow, Baidu push, and GitHub Actions automation.

Solving the problem of images not displaying in encrypted articles by encrypting images separately as .enc files and rendering them via Blob URL after browser-side decryption.

Extending the blog encryption system to support PDF, Jupyter Notebook, LaTeX, and Office documents with end-to-end encryption — type-based dispatch at build time, type-based rendering in the browser.

Implementing a short link system for a purely static blog with deterministic SHA-256 shortcodes and a persistent lock file to ensure links never break.

Building a build-time thumbnail preprocessing pipeline for a personal website gallery, reducing first-screen load from 8.5MB to 0.7MB — a ~91% performance improvement.

When AI Agent cluster configurations are scattered across runtime directories, and every model change requires manually editing multiple JSON files, it's time to bring in Infrastructure as Code.