Building in AI is running a trading book — you're long some curves, short others, and exposed to correlations that break when they matter

Gupta’s closing frame replaces the static-moat mental model: “Value in AI is never captured once and defended. It’s continuously repriced and relocated.” So “investing in AI, and building in it, starts to look less like underwriting a software company and more like running a trading book. You are long some curves, short others, and exposed to correlations that can break exactly when they matter most.” The operating instruction follows: “Choose which variables to bet on, know which ones can kill you, and build the company to recover faster than a wrong bet can compound.” Part of the asset is now research taste — in AI “talent is a raw input into production,” where a handful of researchers “can change the shape of the market,” not just execute against it.

This is the disposition required by AI strategy is a self-rewriting equation — solving one constraint changes which constraint matters next — if the equation rewrites itself, you manage exposures, not a fixed plan. It reframes Build for the model six months from now, not the model of today as a directional bet you’re long on, and pairs with Sand vs Stone — if models double in capability tomorrow, what washes away and what remains? as the position-sizing tool for which curves to hold. Where Gupta locates durability in knowing your long/short book, When products and models converge, the moat is the institution — how a company attracts people, distributes authority and compounds judgment locates it in the institution that can keep re-trading — fast feedback loops, small teams with authority, comfort reorganizing around what changed.