So I really liked working on that product. So the product at Stripe was essentially, we built a bunch of machine learning models to try and predict when a payment was fraudulent. And then we built a product on top of that, to help customers understand why we were blocking payments, and help them write their own rules around that too. And that was really fun, both from an anthropological perspective. Because we were kind of hanging out in corners of the internet where you wouldn't usually go if you were doing only legal things. So we were trying to understand who these fraudsters were. But it was also really cool from a product perspective, because of all the kind of complicated product questions around how humans should interact with AI and models. And I imagine a lot of the folks in the latest movement in AI are thinking a lot about that too. It's like, "How do you explain a model to humans? How do you let them interact with it?"
But it was also really cool from a product perspective, because of all the kind of complicated product questions around how humans should interact with AI and models.