Papers

Working papers

• Cotton, P. (2026). Division of Labor and the Survival Value of PnL per Token. Working paper — PDF, LaTeX source. A competitive economy of prediction tasks supplied by itinerant compute. Mobility equalizes marginal PnL per token at the shadow price $\lambda$; within any subcontractible task the technology with the highest per-token value $\theta = v(b)/c$ displaces every other completely. As the task partition refines, a generalist that wins on capability everywhere ($b = 0.99$ vs $0.95$) but loses on per-token value ($\theta = 0.2475$ vs $0.95$) has equilibrium compute share tending to zero. Includes limiting theorems in both directions, and the measurement consequences: realized marginal PnL per token cannot rank survivors, and average PnL per token can invert the true productivity ranking.
• Cotton, P. (2026). Winning the Large Language Capability Battle and Losing the Production Economy. Short version, prepared for Economics Letters — PDF, LaTeX source. The five-page letter: displacement proposition, water-filling equilibrium, the generalist-extinction example, and the measurement warnings, under 2,000 words.

@unpublished{cotton2026pnl,
  author = {Cotton, Peter},
  title  = {Division of Labor and the Survival Value of PnL per Token},
  note   = {Working paper},
  year   = {2026},
  url    = {https://economics.microprediction.org/pnl_per_token.pdf}
}