I agree with this approach particularly at the start of one's career but I will add that once you find a problem space that you really want to advance the state

I agree with this approach particularly at the start of one's career but I will add that once you find a problem space that you really want to advance the state of the art in, then you better get ready to spend a decade or more studying it. Sure, it's possible to pursue this even by changing jobs, but it is MUCH harder in that case, advancing the state of the art in computer technology is something that very few places in the world have a business need for and the resources to pay people to do that so I think it's much more likely to spend a lot more years in a single place to pursue such a goal.