I read a few papers, they're educational. Better on average than the stuff linked to by HN. That one was special though: Most CS papers are about things I don't know but I can read them, that's why I read them, but I know the foundations and get the point of each paper. The Transformers paper, though, has foundations and assumptions beyond what I know, and explained its point in terms of those foundations.

It's all based on NAND gates when you go down to the ultimate foundation, so in a way you could say it's all just a big fancy spreadsheet. Saying so stretches the words big and fancy well beyond their usual meanings. Saying so is misleading because of that stretching.