[bjourne]

You are right. Still, you need intimate knowledge of the state of the art in whatever field you're approaching to make your dent. And type theory, which TS gave as an example, is a field where all low-hanging fruit were picked decades ago so you'll have a lot of reading and understanding to do to get started. Something more "applied" may be more approachable. Like type theoretic construct X as applied to TypeScript to prevent integer overflows in domain Y (I don't know anything about type theory). A good supervisor could help TS pick a realistic subject. Unfortunately for TS, not every supervisor is good... But even with an awesome supervisor, awesome subject, awesome work ethic, and awesome talent, it's a lot of work.