From Model Scaling to System Scaling: The Harness Around Foundation Models Is Now the Bottleneck

arXiv·high signal

Argues that the next major bottleneck in agentic AI is system scaling — the design of auditable, persistent, modular, and verifiable architectures around foundation models. Treats the structured execution layer (the 'harness') as a first-class object of design, evaluation, and optimization. Directly relevant to practitioners building production agent systems where the orchestration layer matters more than model size.