While I thought I'd be focusing 100% on discrete math in my computing life, I was really happy to have been exposed to gradients and convolutions when neural networks came to prominence. I've also recently gotten interested in power electronics and have been really grateful for having previously been exposed to several varieties of transistors and having had half a semester of thermodynamics. Not to say that this background was really enough to be productive, but it was really nice to already know that these subjects exist and be familiar enough with their broad outlines to not have to completely bootstrap multiple fields at the same time.
I'm a little salty about this though because, at the time, there were a ton of classes that I thought were boring and useless. I was wrong (at least about the uselessness) about for vast majority of them.