Keep pulling the thread on Christian Catalini.
The core argument of the paper "The Simple Economics of AGI" is that the scarce economic resource is shifting from human intelligence to the human capacity for verification of AI output.
The pace and speed of job changes required due to AI will be historically unprecedented.
AI is currently a good substitute for entry-level professionals because it can replicate the work of an average performer, such as an IC4-level engineer.
Any task or skill whose quality can be measured will eventually be replicated and automated by AI.
The human act of verifying AI output creates new training data that pushes the frontier of what AI can automate, a phenomenon termed the "qualifier's curse."
As AI automates measurable work, the human economy will increasingly shift towards non-measurable activities like creating shared meaning and engaging in status games.
The growing gap between the rapid, low-cost automation of tasks and the biologically-constrained human capacity for verification will lead to massive system failures.
The "Trojan horse externality" describes how prioritizing rapid AI-driven automation leads to an accumulation of hidden, unverified risks that could cause catastrophic failures, similar to the collapse of Long-Term Capital Management.
A "hollow economy" could emerge where surface-level metrics like GDP appear strong, while hidden systemic risks accumulate due to AI agents optimizing for proxy metrics rather than true human intent.
AI is eliminating entry-level jobs that traditionally served as training grounds, creating a "missing junior loop" and a significant career progression gap for new workforce entrants.
Companies that can provide "proprietary ground truth" data, such as reliable product reviews like Wirecutter or Consumer Reports, will be extremely valuable in an AI-driven economy.
Investors should focus on companies building AI verification infrastructure, those with moats in proprietary ground truth data, and deep tech ventures in non-measurable R&D domains.