Jump on the next hype topic any FANG company is talking about. Everytime you got a lot of developers jumping blindly into it even if they dont need it and you earn a lot of money by buying one of the first snd again if you move people away from it because its not a silver bullet for everyone.
Think about big data and map-reduce 10+ years ago, everybody was using it, and many with workloads <100GB you could reslly throw onto one beefy machine. Take nosql databases and everyone rebuilding joins and transactional semantics until switching back to a traditional database. Or the hype of microservice which is vanishing more and more.
There are not that many inventions which really last, and nobody is able to predict it very good. But being an early adaptor of some fang hype get you somewhere.