I work at a company that's very much not FAANG (more of a small, niche CAD-flavored Adobe), and I do a lot of what you describe.
Namely, a lot of "whatever the hell I want," almost never asking permission, and I don't recall ever needing to beg forgiveness. The end result seems to be that I now know lots of trivia about dusty corners of our massive, legacy code-base and my manager seems to be consulting me on a bunch of architecture-level decisions. Granted, I ... exercise savvy about playing in a leaf-node sandbox vs. a trunk-node jenga tower, and I'm a fluent English speaker with a PhD-level math background so maybe I get a little more latitude than other people.
Still, I find that I feel a lot more free working here, under considerably less pressure to ship, than I did in the PhD program.
Namely, a lot of "whatever the hell I want," almost never asking permission, and I don't recall ever needing to beg forgiveness. The end result seems to be that I now know lots of trivia about dusty corners of our massive, legacy code-base and my manager seems to be consulting me on a bunch of architecture-level decisions. Granted, I ... exercise savvy about playing in a leaf-node sandbox vs. a trunk-node jenga tower, and I'm a fluent English speaker with a PhD-level math background so maybe I get a little more latitude than other people.
Still, I find that I feel a lot more free working here, under considerably less pressure to ship, than I did in the PhD program.