Past Automation and Future AI: How Weak Links Tame the Growth Explosion

Charles Jones and Christopher Tonetti provide a much-needed injection of macroeconomic reality into the often-hysterical predictions of AI-driven hypergrowth. For those of us in the tech bubble, it is incredibly easy to model AGI as a magic bullet that automates everything instantly and unlocks exponential growth. But this paper brilliantly models the "weak links" in any economic production process—the physical, safety, and regulatory bottlenecks that cannot be automated away by a neural network. No matter how intelligent your software becomes, it still must interface with a messy physical world constrained by energy grids, resource scarcity, legal approvals, and human safety limits.

As a startup founder and engineer, this perspective is deeply grounding for strategic planning. It suggests that the value in an AGI-driven economy will increasingly concentrate around these physical and regulatory bottlenecks—the "weak links" that resist simple automation. If intelligence becomes a cheap, abundant commodity, then physical infrastructure, clean energy, proprietary data, and complex human coordination will become the ultimate premium assets. Instead of trying to build yet another purely digital automation layer, we should be looking at how our technology can integrate with and resolve these stubborn, physical bottleneck constraints that govern the real-world economy.

What stuck with me

• Economic weak links: Physical bottlenecks and safety regulations naturally constrain the growth multiplier of purely algorithmic automation.
• Commoditization of intelligence: As AI scales and becomes cheap, economic value shifts toward hard-to-automate physical and regulatory assets.
• Systemic production constraints: Explosive macroeconomic growth is fundamentally limited by the slowest, most stubborn component in the production chain.