The Demon’s Footprint

A predictor tries to forecast a market. Make it smarter and bigger. At first it gets more accurate. Then its own actions start moving the market it’s trying to read — and the predictions it helped create come back to wreck it.

The decision in this simulation

You’re watching

A predictor grows from tiny to market-dominating while it forecasts that same market.

It stands for

How much of a market a powerful AI forecaster comes to influence as it acts on its own calls.

Why it matters

Past a sweet spot the predictor’s own moves reshape the market, so more power makes the forecast worse, not better.

Irreducible residue34%

▲ floor at 32% — no published predictor, at any size, drives the residue below this line

The market · legible

┄ the predictor’s forecast (published)

Unpredictability vs. predictor size

weak predictor— — floor: residue can’t go lowerpowerful predictor

Growing the predictor… accuracy improving, residue falling.

The point

Beyond a threshold, a more powerful predictor doesn’t shrink uncertainty — it manufactures it. The forecaster is part of the system it forecasts, and its own footprint is the largest new source of the unpredictability it’s trying to erase.

The research behind this

Academy of Management Review · 2025Are the Futures Computable?The computability boundary — including the competitive recursion this simulation dramatizes.

Academy of Management Review · 2025Do Androids Dream of Entrepreneurial Possibilities?How AI systems grapple — and fail to grapple — with Knightian uncertainty.

Journal of Business Venturing · 2025The Expectations GameReflexive belief dynamics in 302 AI startups: hype moves valuations, then breaks them.