Insights
Write-ups from production
Architecture, AI, and engineering decisions, written from real projects. Some write-ups include working models.
ai
Your team does not need more AI-generated code
AI compressed implementation, not thinking. Spend some of the time it saves reading the code together and making sure the software reflects reality.
architecture
This portfolio: how far can AI take a production website?
I built this portfolio to test how far AI could carry a real production system: Next.js on Lambda, an ISR pipeline, and a procedural WebGL sky.
ai
AI coding rewards preparation: research before you build
My seven-stage AI-assisted development workflow: expand the idea, read the sources, refine it, run deep research, add expertise, then build.
infrastructure
Enterprise k3s homelab: GitOps, SSO, zero open ports
A five-node k3s homelab applying enterprise platform patterns: three-node HA, Argo CD GitOps, SSO, observability, network policy, and zero open ports.
ai
GroundNet: an agentic AI safety experiment about trust
A playable agentic AI safety experiment testing what happens when an AI has tools, conflicting goals, and a human who knows how to pressure it.
ai
Ziggy: building a personalized RAG agent with long-term memory
How I built a personalized RAG agent from scratch using graph memory, routed reasoning, MCP tools, and deliberate forgetting—and why it is working.
What you’ll find here
These write-ups come from three places. Some explain how I built a project. Some break down a decision from client work. Some are ideas I gave a client that changed how their business runs. Every one comes from a real system.