I write about software architecture, backend systems, applied AI, and tools I build.
A prompt for your AI agent that turns an Obsidian vault into shared, local memory for Claude, Codex and other agents: a four-question setup wizard, standing rules and free sync.
Read articleShowing 8 of 19 articles
A prompt for your AI agent that turns an Obsidian vault into shared, local memory for Claude, Codex and other agents: a four-question setup wizard, standing rules and free sync.
Coding agents write code in minutes; review takes hours. My PRD and ERD Agent Skills settle product and design decisions before code, in Claude Code or Codex.
Toolbox began as four native macOS utilities. Here is how it grew into a cross-platform desktop app, and why safe, local file handling shaped the way I built it.
Stop AI coding agents from hallucinating files and breaking your build: 11 ways to steer Claude Code, Cursor and Copilot with rules, plan mode, context and git.
Over 50% of small-to-medium projects ignore OWASP guidelines, and that is exactly what attackers count on. Here is how I hardened a real enterprise backend across nine security layers, with code.
How we built a layout-aware ingestion pipeline for enterprise contracts with Docling, Postgres pgvector and local LLMs, without any data leaving the server.
Build a natural language to SQL system using the Model Context Protocol (MCP) SDK, LangChain, Ollama, and PostgreSQL - enabling non-technical users to query databases using plain English.
A real-world account of migrating a Node.js + GraphQL monolith to microservices with zero downtime - covering database migration from ArangoDB to DynamoDB, code decomposition, and scaling strategies.
