Pillar

AI & Automation

AI agents, coding with LLMs, local models, MCP and workflow automation — what actually works when you build with it.

I handed an AI coding agent a real app to build — with a real deploy target — and logged exactly where it shined and where it fell on its face.

Demos are easy. I wired an agent into a real multi-step workflow with side effects and watched what actually breaks in production.

When does running a model on your own hardware beat paying per token? I break down the real trade-off — and it isn't the one everyone quotes.