Really nice thread. Three thoughts: 1. It’s now obvious why NV was the first millennium prize problem to go down. It’s just a giant PDE! (Which, duh, but I hadn’t thought about until now). Big game of guess and verify. If you can throw “400 years of work in a weekend” at it your chances are pretty decent. 2. Of course, none of that would have been possible without three centuries of human work and insight (since Euler, of course, because Euler) to get us here. Human-AI complementarity in action. Worth noting, when the entire narrative around this is about substitution of humans by AI. 3. Speaking of which, Joonas picks up on the “math as exploration” metaphor I’ve also been pushing. I like the extension to video games. Even if the AI can get us in a straight line from the known world to a signpost in the “fog of war”, it’s of the essence that we have players around that can wonder off the path, map the terrain, and find more signposts. Even if you’re the most AI-pilled person out there (and I’m probably at least one SD to the right), this should be glaringly obvious to you. The way we develop this tech HAS to emphasize the role of humans in the future. Otherwise we’re going to sacrifice human-AI complementarities in the future, like the ones that led to this breakthrough.