Machines at Boundaries In 2016, AlphaGo defeated the world champion at Go, a game so complex that brute-force computation seemed impossible. The victory felt momentous: machines mastering domains requiring intuition, pattern recognition, strategic depth. Then researchers tried applying the same system to StarCraft, a real-time strategy game. It struggled. Same underlying architecture, different domain; and the framework that dominated Go couldn't transfer. This isn't a flaw in AlphaGo. It's a demonstration of contextual stratification in artificial systems. The AI learned F_Go at λ_board-game with M_Go (measurable game states, valid moves, winning positions). That framework produced brilliant Q_Go (optimal strategies, creative plays). But F_Go doesn't apply to F_StarCraft at λ_real-time with different M_StarCraft. The boundary between frameworks isn't crossable by mere scaling. It requires different architecture, different learning, different framework. AI system...