AI, Work, and the Next Downturn

The argument is that AI is already displacing paid tasks and that continued automation will shift income from wages toward owners of scarce assets - land, location and power - so the next downturn could destroy jobs that do not return. A public quantitative model projects large labour‑market losses under plausible shocks: the scenario presented yields 14-16% of the US workforce out of work by end‑2028 (versus a 10% central case), while an extreme automation path pushes unemployment toward about 46% by mid‑2032. Empirical specifics include health and social care supplying 107% of US net job growth from Aug 2024-Aug 2026, labour’s share of US business output at a record low (93.4, 2017=100), a 0.6‑point drop in prime‑age participation in June 2026, and a 4.8× divergence between wages in durable goods and in shelter since 1964-2024.

The current assessment finds long‑term borrowing costs rising across rich countries faster than economies are weakening, and the labour market appearing calm on the surface but hollow underneath. The near term will bring higher rates for longer, fiscal tightening in Europe, and a material risk that the AI investment boom stalls on financing costs; a stock‑market bust larger than the dot‑com crash is considered plausible and could lock in permanent job losses. Eight of 29 tracked indicators have triggered - jobs outside care, labour’s share, people leaving the workforce, the premium on US government debt, France’s borrowing costs, oil near $125, US mortgage rates ~7.28%, and Japan’s bond yields - and the model and underlying data are published and reproducible on GitHub.