The AI Layoff Trap
2026 · Karl V.
A recent paper by Brett Hemenway Falk and Gerry Tsoukalas shows that firms replacing workers with AI can trap themselves. Each firm is better off automating, yet when all of them do it, both owners and workers end up poorer: the workers laid off were also the customers.
The paper is a mathematical model, and almost nobody outside economics will read it. This piece lets you play it instead, in a small ink-on-paper town of firms, workers and coins, without adding claims the paper does not make.
How this piece was made
Documented
Every number on screen is computed by an engine that implements the paper’s Propositions 1 to 6, checked against the paper’s own figures by 28 automated tests.
In the paper’s illustrative case (7 firms, AI costing 30% of a wage), firms automate 65% of tasks, while the level best for the firms themselves is 35%.
Artistic choices
The town, its people, robots and coins are a visual metaphor: the model itself has no geography.
Work in progress. Acts 1 and 2 (the game, and what more rivals do) are live; better AI, policy fixes and the model’s limits come next.