To celebrate its 350th birthday, the British Royal Society put 60-odd papers online. I've been reading through them, as a method of procrastination which is hard to tell from my 'real work' of reading a slightly different set of papers. There are some real gems there; I particularly like the early medical papers on dogs, and the one about the effects of standing in a furnace on body temperature (read: homeostasis ftw!). But possibly the coolest thing about it is you can see the modern scientific format being born: those early papers read more like letters between friends than prestigious scientific publications, (Ben Franklin's 'I've heard you've been talking about these lightning rod things. Here's a cool experiment I did with a kite' is especially nice in that light) but slowly they become more rigorous and formalized. On the other hand, I still haven't gotten through the Bayes paper, if only because it is mathematics and hence much more formalized and technical. Oh, and it's fifty pages long.
They provide background info and commentary, too, and generally it's review articles that are reasonably accessible to a lay audience I think.