The Alpha Half-Life

Set how fast AI agents take over capital allocation and how alike they think. Watch how long a mispricing survives, how much easy return is left, and how violently the market reprices a genuine surprise.

How long the market takes

A new mispricing loses half its edge in
A genuine shock is fully repriced in
Easy alpha left per year
Year easy alpha falls below your cost
Capital run by agents
Overshoot beyond fundamentals in a shock
Drawdown from a 10% fundamental repricing
Share of edge still coming from friction

The pool drains

Friction alpha decays as agents absorb the mistakes that created it. The premium for things models can't manufacture rises in its place.

Easy alpha from friction Premium for scarce assets and patience Your cost of chasing it

Where the edge lives

How the model works
  • Agent share follows an S-curve toward the ceiling you set, reaching half of it after the number of years you choose from 2026.
  • Easy alpha = 10 bps floor + (today's pool − floor) × (1 − agent share)capability. The floor reflects the Grossman–Stiglitz point: prices can never be so efficient that nobody is paid to gather information.
  • Mispricing half-life starts at 45 days in a human-run market and decays exponentially with agent share × capability.
  • Shock repricing starts at three weeks and decays with agent share, capability, and signal overlap.
  • Overshoot = 1 + 1.6 × overlap × agent share × (0.4 + refinancing pressure). Correlated agents plus leverage turn an orderly repricing into a cascade.
  • Scarcity premium = 40 bps + 140 bps × agent share: the return migrating to proprietary information, physical assets, patience, and liquidity.
  • This is an illustrative instrument, not a forecast. The parameters are chosen to make the mechanism legible, not calibrated to any dataset.