Member-only story
A decision-theoretic view of long-term investing, with path-dependent risk, capital growth, and adaptive allocation
7 min readJust now
–
Motivation: why long-term investing is not a prediction problem
Most retail discussions of “AI investing” quietly start from the same assumption: that the core task is prediction.
Predict tomorrow’s return. Classify the next trend. Time entries and exits more precisely.
After years of real-world investing, I came to a different conclusion.
The hardest part of long-horizon investing is not forecasting individual price moves. It is designing a decision process that:
- survives repeated regime changes,
- avoids catastrophic, path-dependent mistakes, and
- remains actionable under realistic frictions — transaction …
Member-only story
A decision-theoretic view of long-term investing, with path-dependent risk, capital growth, and adaptive allocation
7 min readJust now
–
Motivation: why long-term investing is not a prediction problem
Most retail discussions of “AI investing” quietly start from the same assumption: that the core task is prediction.
Predict tomorrow’s return. Classify the next trend. Time entries and exits more precisely.
After years of real-world investing, I came to a different conclusion.
The hardest part of long-horizon investing is not forecasting individual price moves. It is designing a decision process that:
- survives repeated regime changes,
- avoids catastrophic, path-dependent mistakes, and
- remains actionable under realistic frictions — transaction costs, liquidity constraints, limited attention, and the psychology of uncertainty.
Once framed this way, the problem stops looking like supervised learning and starts looking like sequential decision-making under uncertainty.