Schmidt gives three tests for whether you’re safely off the Yellow Brick Road. The tools-and-steps test: few steps with a forgiving outcome (a Drive search) is lab territory; dozens of steps that must clear partner review (a legal redline against three years of precedent) is not. The system test is the sharpest: are you a system the customer runs their work through, or a tool that sits on top of a system they already have? Its litmus question — “would the customer still need you if a lab shipped something that supposedly directly competes with you? If yes, you’re a system. If no, you’re a tool — even if your ACV is high.” High ACV signals a system but doesn’t guarantee one.
The P&L / hedge-fund test: the customer judges you on their own P&L (did the deal close, did the policy bind), not on SWE-Bench or MMLU — “winning on alpha measured in customer P&L, not in benchmark scores.” These operationalize The system of work is the moat, not the model — the model is fungible underneath and Sell the work, not the tool — the services budget dwarfs the software budget, and every model improvement helps whoever sells the outcome, and the system test is the practical inverse of Frontier companies absorb every useful agentic pattern into their products — if a lab shipping the feature would kill you, you were on the road.