The blog
Notes & observations.
On data, uncertainty, and the limits of models.
Subscribe via RSS →
Topics
All writing
Harari’s Imagined Imagined Communities
A critique of Harari’s account of shared stories and large-scale cooperation.
Common Arguments About the AI Bubble
Five common AI bubble arguments, from valuations to workplace productivity.
Supervised Learning, RLHF, and DeepSeek R1
A math-class analogy for how three training approaches score an LLM’s answers.
A Simple Stock Market Trick: Use Logarithmic Charts
Why logarithmic stock charts make percentage changes easier to compare.
Farewell, Yann LeCun
On Yann LeCun, scientific metrics, and having skin in the game.
Yuval Harari's Dataism: Why is it a Bad Idea?
Why more connections and more data do not necessarily mean more information.
Estimate Book Review Entropy from Pre-trained Embeddings
An experiment in measuring the diversity of book reviews with embeddings and entropy.
A Simple Way to Estimate DDP, A Better Alternative to GDP
Estimating Democratic Domestic Product with the Gini index and a Pareto distribution.
Fat Tails, Survivorship Bias, and the Illusion of Exponential Growth
How focusing on the winners can turn fat-tailed outcomes into a story of exponential growth.