A good seminar should have a lengthy introduction at a level somewhat below the general level of knowledge in the room and then transition into showing off how clever your approach is. The first part is critical because it makes everyone feel smart. The second part is critical because it shows everyone that you are smart.

I think the lesson is "Read the room". Different types of seminars have different purposes.

For some, it is indeed desirable to dazzle and confuse. In others, to be approachable.

This seems counter to almost every CS hiring job talk I have attended.

The professors almost always seem to be receptive of people who lay down the basics for everyone in the room, and then systematically place their work using this newly situated academic landscape.

I mean, how is a databases professor supposed to understand a SOTA CNN, if they don't understand CNNs to begin with.

Maybe OP was simply stuck in the wrong (or right) kind of toxic workplace.

I can see that perspective too. I had interpeted it as "don't make it seem that our research (via seminar talks) can be understood by undergrads!"