A capacity split for a team's roadmap between two different jobs: building to learn creates a safe space to run experiments and test hypotheses where being wrong is cheap and expected, while building to earn is operational-excellence work meant to be stable, performant, and directly revenue-generating, where being wrong is expensive. Treating both as the same kind of work — holding experiments to production-reliability bars, or holding revenue-critical systems to move-fast-and-learn tolerances — misapplies the wrong standard to each. The split gives a team and its leadership a vocabulary for deliberately allocating capacity between discovery and delivery instead of discovery quietly losing out to the delivery work that's visible on a keep-the-lights-on ledger.