> there are some heuristics you can use to guess whether an obsession might be one that matters. For example, it's more promising if you're creating something, rather than just consuming something someone else creates. It's more promising if something you're interested in is difficult, especially if it's more difficult for other people than it is for you. And the obsessions of talented people are more likely to be promising. When talented people become interested in random things, they're not truly random.

You can be very rigorous in trying to understand some topic, such that you won’t be satisfied until you understand it from first principles. Since so few people do this, it can lead to new discoveries.