Why I'm Building in Public
This site is not a portfolio or a résumé. It's the public record of how an operator thinks, what he's learning, and what changes when a worldview meets a payroll run. The charter for everything else here.
Thoughts
Long-form essays. Worldview, business, leadership, technology, economics, philosophy.
The Founder Has to Become the Distribution
When creation gets cheap, trust gets expensive. A founder's public thinking stops being personal branding and becomes part of the company's distribution — the trust layer that features can't copy.
Companies Are the Real School
A company is not just an economic vehicle. It's a school where reality grades your thinking every day — and the only classroom I fully trust for judgment. This is why the Lessons section exists.
AI Is Turning Small Teams Into Institutions
The interesting question isn't what one person can do with AI. It's what a small, coordinated team can become — smaller in headcount, larger in capability, coordination, and institutional weight.
The Verification Economy
When generation approaches free, the scarce goods become provenance, audit, and the right to be believed. Every flood of cheap supply in history has created a verification industry. This one will too.
The Size of the Firm Is a Technology Variable
Coase asked in 1937 why companies exist at all. His answer — coordination costs — just got repriced by AI on both sides of the firm boundary. Company size is becoming a design decision, not a destiny.
The Service Company That Becomes Software
AI gives labor-based service businesses a one-time window to convert labor margins into software margins on the same contracts. Most will miss it — by cutting costs instead of repricing value.
The Entry-Level Problem
Firms buy junior labor for output, but society gets skill formation as a by-product. AI now supplies the output without the by-product — and senior judgment is manufactured from exactly the reps being eliminated.
The Dynamo Problem
Factories electrified in the 1890s and saw no productivity gain for forty years — until they stopped arranging machines around a shaft that no longer existed. Enterprise AI is in its line-shaft years.
The Cost of Intelligence Is Falling. The Cost of Judgment Is Not.
Machine cognition is repricing like bandwidth did in the nineties. Judgment — deciding what to want, what counts as done, what to trust — is not. Everything else on this site follows from that line.
Talent Is Evenly Distributed. Leverage Is Not.
Ability is a global constant. Tools, capital, rails, and networks are not. The defining arbitrage of this decade — economic and moral — is closing that gap.
Tacit Knowledge Is the Last Moat
We know more than we can tell, said Polanyi. That boundary is now an economic frontier: what can be articulated can be automated — so the durable edge of a firm shrinks toward exactly what it cannot write down.
Models Are Commodities. Workflows Are Moats.
The model layer deflates on a visible schedule. The workflow layer — context, permissions, exceptions, evals, trust — appreciates, because it encodes the organization itself. Build on the first; own the second.
Goodhart's Revenge
When a measure becomes a target, it ceases to be a good measure. AI is the most efficient target-hitter ever built — which means it breaks metrics at machine speed. Measurement design just became a core competency.
The Only Three Things
After a decade of building, almost every important decision collapses into three questions.
Global Employment Is a Protocol
The companies that win the next decade will treat hiring across borders the way the internet treats packets — as a default, not an exception.
On Quiet Discipline
The most underrated leadership trait isn't vision. It's the willingness to do unglamorous things, consistently, when nobody is watching.