We discuss Chroma, an AI-native database designed to function as "programmable memory" for AI systems. We argue that the current computing landscape is undergoing a radical shift toward an AI-Native Stack, demanding new infrastructure that existing databases cannot adequately provide. Chroma distinguishes itself from "vector databases" by focusing on optimizing for machine consumption rather than human queries, aiming to solve the challenge of productionizing AI by simplifying the management of vector embeddings. The company's open-source model and serverless architecture are highlighted as key differentiators, fostering a community-driven distribution flywheel and positioning Chroma to dominate the emerging market for Context Engineering and become a foundational layer for future autonomous AI systems.