Now

Right now I am deep in one shift: the job stopped being writing prompts and became building loops. Small systems that run an AI agent for you, on a schedule, while you sleep, that ship on green with no human in the seat. Prompts make one person faster. Loops make the whole team faster.

I am CTO at Omilia, and most of my thinking lately is running a fleet of these loops the way I once ran a crew on a warship. Scoped stations. Clear orders. A hand only on what you cannot take back. The goal is plain, ten loops running in parallel, constantly, doing the work nobody should still be doing by hand.

At Omilia that work centers on our agentic platform and the memory framework beneath it, the part that lets agents remember, hold context, and act reliably across long-running work.

I just finished a field guide on it, From Prompt to Loop. The whole framework, what a loop actually is, thirty five worth building, where they run, what they remember, what they cost, and how to stop one when it goes sideways. There is a real repo behind it you can clone and run, on Claude or Codex.

I write about all of it as The Warship CTO, and I post daily on LinkedIn and Substack. Crews, not committees. Loops, not turns.

The tool was never the skill. The thinking behind it was.