Plane Math — does the speed pay for itself?
An interactive cost-of-ownership model for the general-aviation fleet. Set how you fly, compare any two aircraft head-to-head, and find the planes nobody beats on both cost and speed.
Open Plane Math →The first long-form essay is in the run-up.
Deep dives on AI strategy, judgment, and the view from the flight deck — the long-form that doesn't fit in a post. Subscribe, and the first one lands in your inbox.
Subscribe on Substack →Most companies don't need an AI strategy.
IBM, Zillow, Amazon didn't have technology failures — they had decision failures with expensive models. If you can't name the decision your model improves, you have a science experiment with a budget.
I forfeited my CPA license. Now I build foundational AI.
Audits run on zero trust — every number traces to a receipt. That one instinct catches more bugs than any unit test I've written. Your weird background is the lens your team is missing.
My career path makes no sense on paper.
Pilot → CPA → EY → Uber → Meta. It looks random — but it's the same muscle the whole way: high-stakes calls on incomplete information, which is exactly what AI needs more of right now.
What is vLLM?
An open-source project quietly cut LLM serving costs ~20x — not with a better demo, but with a smarter memory allocator. The boring infrastructure is what decides who can afford to compete.
AI demo vs. AI in production.
It's the gap between driving in a parking lot and driving in Hanoi. The demo works because you control everything; production works because you anticipated what you can't. Most companies are still in the parking lot.
The AI consulting bill is coming due.
Thousands of agents, millions in fees — and now the question is "did it make money?" Deploying models stopped being the edge. Tying them to decisions is the whole game.
Substack is for the long stuff. The shorter takes live elsewhere.
I post free observations, mini-essays, and field notes on LinkedIn. Premium long-form lives only on Substack — by design.