Two of three, and you’re defensible.
A better model drops every few months. These are the three walls it can’t climb. Hold two and you have a business. Hold one or zero and you’re a prompt wrapper.
Distribution
The heaviest moat, because every other one flows through it. If your users came to you instead of the model directly, that distribution is borrowed, and it weakens every time a lab ships a better native feature.
Proprietary state
Your data and context: every edit a lawyer makes, every patient chart, every workout logged. Even if a better model drops tomorrow, it doesn’t have your three years of annotations. The data belongs to the workflow, not the model.
Regulatory authority
The legal right to act in the real world. A self-driving car needs government permission; a surgical robot needs FDA clearance. No amount of better model can shortcut years of regulatory runway.
The winners won’t build the best model.
They’ll own the layer above it or below it. Vertically integrated leaders like Amazon, Tesla, and Google are the most protected, because a new model can’t unseat them: the model is only one of the things they have. The stack is the moat.
Billions in private vehicles, deployed one flywheel at a time.
We're raising the capital to buy more software, roll up more operators, and lend against more hard assets. If you build, operate, or back these businesses, we want to talk.