There's no clear segmentation. There's symmetric and asymmetric primitives (and stuff that doesn't fit into these like ZKP), algorithms, protocols, research in many different types of attacks against each of these, research in design and defenses, and plenty of people will cover completely different subsets.
"don't" roll your own cover everything from "don't design your own primitive" to "don't make your own encryption algorithm/mode" to "don't make your own encryption protocol", to "don't reimplement an existing version of any of the above and just use an encryption library"
(and it's mostly "don't deploy your own", if you want to experiment that's fine)
I wonder if there is a concrete point at which it turn into “this is common sense security that even you should know about” like not conflating hashing and encryption, or “you should just have someone else do do security for you”? I guess at larger entities you have a CISO role but what about in smaller, scrappy endeavours, how does one know where one is at the limit of their due-commonsense and hand it off?