Choosing colors that work together is harder than it looks. This post presents a Python script that analyzes any hex color and generates harmonizing palettes based on color theory, while also calcu...

Make is a 48-year-old build tool that refuses to become obsolete—because dependency graphs are genuinely difficult to replace. However, Make’s user experience remains rooted in 1976. This post argu...

A self-documenting keybinding system that lives inside the editor. This plugin enables browsing 30+ categories of shortcuts through a fuzzy-searchable Telescope picker without leaving the workflow....

Part 6 of the PDF to Knowledge Graph series. Retrieval-Augmented Generation (RAG) grounds LLM responses in retrieved context rather than relying solely on parametric memory. Most RAG systems use v...

Part 5 of the PDF to Knowledge Graph series. Graphs are inherently visual structures, yet most graph databases offer only textual query interfaces. A list of nodes and edges reveals little about t...

Part 4 of the PDF to Knowledge Graph series. Manual processing does not scale. Hundreds of PDFs require processing, and more arrive regularly. This post presents a file-watching system that automa...

Part 3 of the PDF to Knowledge Graph series. Graph databases excel at relationship-heavy data, but most require server infrastructure. Kuzu is an embedded graph database—no server, no Docker, just...

Part 2 of the PDF to Knowledge Graph series. Large language models understand text remarkably well but produce frustratingly inconsistent output. Requesting “a list of entities” may yield JSON, bu...

Technical documentation exists in a paradox: more is generated than ever, yet extracting actionable knowledge remains stubbornly manual. Research papers, specifications, internal documentation—thes...

Part 1 of the PDF to Knowledge Graph series. PDF extraction is deceptively difficult. Standard libraries produce reasonable results on simple documents but fail catastrophically on technical paper...