It's even weirder than that, it's like if Google received your messages for you and bundled them up with everyone else's, then allowed you to read your share of the bundle but never actually delivered them to you as far as the protocol is concerned.
The email protocol isn't complicated though. Back before spam was a thing, lots of people could just open a port and manually talk SMTP (you could send messages as Santa, or whatever). The only service Gmail provides that's not trivial (and I mean, something that can be done in a whiteboard interview) is the spam protection.
With the accessibility of ML these days, I'm less convinced that spam protection is that hard, if you have a little bit of patience. The spam I get is repetitive and bad, and I get enough of it that I think I could get a workable model simply by using Gmail's spam tagging to train it.
Back in the day ('noughties through 2015 or so) there was POPFile, which did Bayesian classification of emails for spam filtering. John Graham-Cumming wrote it, and it worked reasonably well. I still think of him as the POPFile guy first, then remember he's the Cloudflare CTO.
A more advanced classifier certainly do better these days.
The email protocol isn't complicated though. Back before spam was a thing, lots of people could just open a port and manually talk SMTP (you could send messages as Santa, or whatever). The only service Gmail provides that's not trivial (and I mean, something that can be done in a whiteboard interview) is the spam protection.