Stijn Masschelein

@stijn-masschelein.net

Researcher and lecturer in management accounting at the University of Western Australia. Working on a textbook. Team R. Owned by one cat.

From Terry Tao’s ICM lecture. While there are additional arrows later that aren’t shown here, verification informs exposition. If an AI proof is autoformalized before it is understood and reworked by a human, the result will be more conceptually brittle even if evidently correct. /1

Bild

Obviously now that it’s massive headline news a bunch of people who know nothing about the situation will comment stupid things about it (esp because it’s AI related) but the Leopold blowup is quite shocking to basically every professional investor who was following him for several reasons:

We're happy to announce our new #rstats package: lorax. If you fit tree-based models, lorax enables conversion to the party package's format (for making nice tree diagrams) and has methods to compute which features were used, the number of terminal nodes, etc. opensource.posit.co/blog/2026-07...

Introducing lorax: Speaking for the Tree-Based Models

lorax is a new R package that characterizes fitted tree- and rule-based models: extract their decision rules, see which predictors they actually use, and convert trees for plotting with partykit.

opensource.posit.co

These things are all decided by the framing effects; nobody ever actually offers you a gamble like this so nobody has good intuitions about what it would make sense to do. tbh it was a good day for economics when we tacitly agreed to give up trying to understand risk aversion