Michael Leznik·3d agoForty-Six Thousand VarietiesThe Bayesians could not agree either — and what your 97% means to each of them
Michael Leznik·Oct 18The Grammar and the BookkeepingWhy artificial intelligence will not destroy mathematics
Michael Leznik·Oct 7Two Men Who Could Not Stand Each Other Built Your A/B TestFisher, Neyman, and the hybrid nobody designed
Michael Leznik·Sep 28The Ladies’ DiaryOver three hundred problems from an almanac for women, and the women who set them
Michael Leznik·Sep 12Model-First, Not Bayes-FirstA point-by-point reading of Spotify’s “Why Spotify Is Not Using Bayesian A/B Testing” in the light of Kruschke’s “Bayesian Estimation…
Michael Leznik·Sep 11Your Two Favourite Metrics Have the Same BugAverage Revenue per Paying User (ARPPU) conditions on being a payer. Average Revenue per Daily Active User (ARPDAU) conditions on being…
Michael Leznik·Sep 4The Paradox That Destroys Data-Driven Decisions (And How to Spot It)Two treatments. Two stone sizes. One reversal that should change how you read every dashboard you own.
Michael Leznik·Jul 5Integrating T-ShirtRecently, my son spent a week at an Olympiad Maths training camp at Oxford University, delving into a treasure trove of challenging and…
Michael Leznik·Jun 2The journey of a token, in search of attentionHow a single word finds its meaning inside a Transformer — told for people who already know information retrieval.
