Everything Is a Probability Distribution
Certainty isn’t a state you achieve — it’s a parameter you update.
Everything worth thinking about is a probability distribution, not a point estimate. This sounds abstract until you start applying it, and then it becomes impossible to unsee.
Markets As Probabilistic Systems
Most people approach financial markets looking for answers: will this stock go up? The question itself is malformed. The right question is: what is the probability distribution of outcomes for this asset, given what I know, and how does it change as new information arrives?
This is exactly what all my years of working with data and studying the markets taught me. When I build a forecasting pipeline, I don’t produce a number — I produce a confidence interval. When I model risk, I produce a probability, not a label. The business wants the label because it’s actionable, but the probability is what’s true.
Apply this to a portfolio: you’re not picking winners. You’re constructing a distribution of outcomes and deciding how much variance you’re willing to carry given your time horizon.
Astrophysics as Calibration
I find astrophysics interesting because it’s the most extreme calibration exercise available. We infer the composition of stars we’ll never visit from the light that left them thousands of years ago. We build models of black hole interiors we can never observe. The uncertainty bars are vast; the inferences are still meaningful.
It’s a good reminder that not knowing the answer with certainty isn’t the same as not knowing anything. You can reason under uncertainty. You can act under uncertainty. You just have to be honest about what you actually know versus what you’re projecting.
The Mental Health Dimension
Probabilistic thinking changed how I relate to mental health and anxiety. Anxiety is, fundamentally, a miscalibrated probability estimate — the brain assigning high likelihood to low-probability catastrophic outcomes. The fix isn’t to suppress the fear; it’s to update the estimate with evidence.
This is not a dismissal of mental health challenges. Anxiety disorders involve broken updating mechanisms, and that’s a medical reality, not a calibration problem. But for everyday worry — the kind that’s mostly noise — asking what probability am I actually assigning to this outcome, and what’s the evidence? is genuinely useful.
A point estimate is a comfortable lie. A distribution is a harder truth — and the only one worth building on.