Agents onward: why I'm writing about production agentic AI

I've spent the last two years building agentic AI systems in production — not demos, systems a business actually depends on.

Along the way I noticed something: the gap between "agent that looks impressive in a demo" and "agent that's actually load-bearing" is where all the real engineering lives, and almost nobody writes about it.

This is where I'll write about what I've actually learned — orchestration, tools, harnesses, spec-driven development, testing agents properly — starting from "agents onward," not another LLM-101 post.

More soon.