Proposes 'Amdahl’s Law for AI,' proving that human effort in AI-assisted work is bottlenecked by the fraction of 'novel' tasks rather than agent capability.
March 31, 2026
Original Paper
The Novelty Bottleneck: A Framework for Understanding Human Effort Scaling in AI-Assisted Work
arXiv · 2603.27438
The Takeaway
The model predicts that human effort transitions sharply from linear to constant with no intermediate scaling, explaining why better agents don't always yield proportional productivity gains. It identifies the 'novelty fraction' as the key parameter for organizations to measure when deploying AI agents.
From the abstract
We propose a stylized model of human-AI collaboration that isolates a mechanism we call the novelty bottleneck: the fraction of a task requiring human judgment creates an irreducible serial component analogous to Amdahl's Law in parallel computing. The model assumes that tasks decompose into atomic decisions, a fraction $\nu$ of which are "novel" (not covered by the agent's prior), and that specification, verification, and error correction each scale with task size. From these assumptions, we de