Writing
Documenting the journey from enterprise architect to AI practitioner. Real experiments, honest takes.
11 articles
An experiment in running multiple AI agents in parallel on real backlog items from the recipe manager — what worked, what collided, and the one boundary that mattered more than any prompt.
The AI skills gap narrative sounds reasonable until you look closer. In many organizations, the real issue isn't that employees can't use AI — it's that they aren't allowed to.
A behind-the-scenes look at a full site audit session: image optimization, self-hosted fonts, SEO fixes, accessibility improvements, and two CI build failures I had to hunt down and kill.
Twenty years of experience makes some things easier and some things harder. Here's the honest version.
GPT-5, Claude Opus 4, Gemini 3, and Grok 4 all dropped within weeks of each other. Benchmarks are nearly tied. The question has shifted from which model wins to which model fits your use case.
94% of enterprises running AI agents say they're worried about agent sprawl, technical debt, and security risk. That's an architecture problem. Here's what it actually looks like to govern a fleet of agents.
Most developers use Claude Code as a smarter autocomplete. The real unlock is the Agent tool — spawning parallel sub-agents that explore your codebase independently and keep your main context clean.
A full-stack Next.js app that imports recipes from any website, generates weekly meal plans, and consolidates ingredients into a smart shopping list.
A honest account of using Claude Code alongside VS Code, GitHub, Netlify, and Cloudflare to get a site from zero to live.
After 20+ years in enterprise architecture, I'm going deep on AI. Here's why I'm doing it in public.
What two decades of database architecture taught me about thinking in probabilities instead of certainties.
// No posts match that filter.