James van Geelen on His Viral AI Doom Scenario

What It Covers

James Van Geelen of Citrini Research joins Odd Lots to explain his viral Substack piece outlining a 15%-probability AI disruption scenario set in 2028, where accelerating AI capability curves trigger white-collar unemployment above 10%, a 40% market decline, and cascading private credit stress.

Key Questions Answered

• •AI Capability Acceleration: Autonomous AI agent task duration expanded from 2 minutes to 8–16 hours of complex intellectual work within two years. Investors should monitor this metric as a leading indicator — when multi-day autonomy arrives, white-collar displacement risk crosses from theoretical to economically actionable within a single budget cycle.
• •Scenario Probability Framing: Van Geelen assigns only 15% probability to the doom scenario, framing it as a bear-case stress test rather than a forecast. Investors benefit most from mapping all three scenarios — bull, base, bear — before a crisis, not after. The most dangerous position is being unable to visualize the downside at all.
• •S-Curve Misapplication: Traditional S-curve adoption models measure breadth of new technology uptake but miss intensity of capability improvement within already-adopted platforms. AI embedded as a feature inside existing tools — like spell-check in word processors — bypasses consumer adoption friction entirely, making displacement faster than standard diffusion models predict.
• •Private Credit Vulnerability: High-earning white-collar workers targeted for AI displacement carry approximately 780 FICO scores — a demographic not historically modeled as default risk. Private credit portfolios with ARR-based software lending assumptions should be stress-tested against scenarios where recurring revenue becomes non-recurring as AI erodes software pricing power.
• •Agentic Commerce and Moat Erosion: AI agents optimizing purely for lowest price eliminate the friction that sustains network-effect moats in delivery and payments. Unlike comparison shopping websites requiring active user effort, an agent executing "get me the cheapest burrito" autonomously bypasses incumbent platforms entirely, compressing margins for intermediaries who currently extract rent through consumer inertia.