Blog
Notes from rebuilding a company as a set of self-improving loops — written from inside the build, with the receipts.
The agency of one: the operating system I run a whole company on
Not a productivity stack — an operating system. The loops running the company, the business brain underneath them, and the revenue-per-person number that says one operator is enough.
The org chart of one to five — without hiring four people
Most owners scale by adding headcount. I scale by closing the next loop. The five roles a growing company fills, which loop absorbs each, and the one thing that never gets delegated.
How self-improving loops fail (and the gate that catches it)
Loops break — quietly, by drifting open. The common failure modes, and the quality gate and learning mechanism that turn an open loop back into one that gets sharper.
From a personal AI setup to a business brain
Everyone starts with a personal AI workflow. The leap is making it a company's brain — legible, shared, written back to by every run. What changes when knowledge, not the model, is the thing you invest in.
What running a company on loops actually costs
The real cost curve as you scale — tokens up and to the right — set against revenue per person climbing faster. The token bill, subtracted out loud. Not projections. The actual bill.
The market of one: when your customer is an agent, not a person
When an agent does the searching, comparing, and buying, the unit of a market becomes one. What that means for how a small business gets found and chosen — mapped from inside my own funnel.
Agentic commerce comes for small business first, not last
Everyone assumes enterprise gets agentic commerce first. I'll argue the opposite — small operators move faster — and back it with one transaction I ran agent-to-agent.
human.txt: writing for the agents that read your company
If agents are the audience, publish for them on purpose. The agent-readable brief I wrote about my company — what's in it, why, and a log of an agent using it correctly.
Migrating a real business to AI-native: the playbook
Not theory — the sequence I'd run to move a live business onto loops without stopping it. One real before/after operating map, the first loop closed, the open loop it replaced.