Hidden Technical Debt of AI Systems: Agent Runtime
Eleven years ago, Sculley et al. drew the diagram everyone in MLOps has seen: a tiny black box labeled “ML Code” surrounded by a sprawl of much larger boxes — data collection, feature extraction, configuration, monitoring, serving infrastructure. The point of the diagram was that the model code is the smallest piece of a real ML system, and that e...
This analysis surfaces a critical blind spot in AI system design: the assumption that the model is the primary locus of value, while the runtime is treated as mere plumbing. The pattern here echoes historical cycles in software engineering—where infrastructure debt accumulates until it becomes the dominant cost—now accelerated by the unique risks of agentic systems. The article’s strength lies in its concrete taxonomy of isolation primitives and the trade-offs between them, grounding abstract co...