Counterintuitively, don't automate all the things until that automation is sure to pay off. Sometimes manually testing something is just faster, if you only need to test it a few times before the code never changes again (when was the last time you rewrote your password reset flow?)
To get prolific, don't do katas. Katas are the same problem solved in different environments. Do different problems in the environment in which you are productive. That is what will make you better and faster.
Are you sure about that? I thought it was more solving the problem, then solving it again more efficiently / differently in the same environment. Like using recursion the second time around, not changing languages.
"Don't code your CSS from scratch, use Bootstrap or similar. Don't create an auth system, use the one that comes with your framework."
Definitely true for productivity, but a lot to be gained in experience by knowing how to roll your own.
I always start off thinking it'll be easy, and every single time I'm amazed by the number of details and edge cases involved. It's usually where you'll get exposed to more interesting algorithms, since most applications are able to leverage libraries which usually hides the magic.