I like the Baroque Cycle as much as anyone and it is true that it's a standalone story. However, unless you're a die-hard scholar of late seventeenth and early eighteenth century European history, I think you'd be much better off reading Cryptonomicon first, then easing into the Baroque Cycle later. There are many advantages to this approach. First off, Cryptonomicon is about 1800 pages shorter than the 2700 page Baroque Cycle. Second, Cryptonomicon's settings (WWII and present day) are more accessible. Third, the entire series focuses on characters from around 2 major families, with several other side character families, and I think the introduction to these families is better made in Cryptonomicon. This provides you with a sorta shorthand when encountering characters in the Baroque Cycle, allowing you to focus on all the other stuff Stephenson is throwing at you without being totally overwhelmed. Finally, I think Cryptonomicon is just plain better than the Baroque Cycle, though I really enjoyed both. But this also begs another question - is Cryptonomicon the best place to start? If not, what is?
It's a truly tough question. I think Shamus really nailed Cryptonomicon and Stephenson in general with this statement:
In fact, I have yet to introduce anyone to the book and have them like it. I’m slowly coming to the realization that Cryptonomicon is not a book for normal people. Flaws aside, there are wonderful parts to this book. The problem is, you have to really love math, history, and programming to derive enjoyment from them. You have to be odd in just the right way to love the book. Otherwise the thing is a bunch of wanking.When I read Anathem, I got a similar feeling, but with different subjects. And when I think about the rest of his work, I find myself struggling to find an ideal starting place for Stephenson. I've come up with some ideas below, but I'd certainly be interested in any of my 5 readers' (at least a couple of whom have read some Stephenson) thoughts on the subject as well. In any case, I think the best place to start (perhaps not coincidentally) is the same place I started: Snow Crash. It's more accessible than most of Stephenson's later novels, and it's not nearly as long either. It's also a lot of fun.
Now, there are some things about Snow Crash that might be off-putting to new readers. For instance, it belongs to a specific sub-genre of science fiction called Cyberpunk. To be honest, I'm not especially in love with that sub-genre. William Gibson popularized the concept with his novel Neuromancer, which was kinda like futuristic Raymond Chandler, and that's widely considered to be the best cyberpunk novel. Snow Crash is almost (but not quite) a parody of cyberpunk tropes, while still being an excellent example of the sub-genre. One of the things I don't like about Cyberpunk is that it's infused with a sorta earnest nihilism or cynicism. Stephenson doesn't take it as seriously and has a lot of fun with the typical tropes of the sub-genre, which makes some of the more ridiculous stuff go down easier. There's a satirical element to the book that I don't get from a lot of other cyberpunk, and that makes the proceedings more interesting to me. Once you get past the initial culture shock at the beginning of Snow Crash, things rocket along pretty quickly. There's plenty of action and even the occasional info-dump doesn't slow things down too much. The characters are fun and the ideas are interesting. What's more, I know lots of people who have read and enjoyed this book, which seems to indicate that it's perhaps not as narrowly focused as something like Cryptonomicon. It's also widely considered to be one of his best novels and also one of the best SF novels of all time. For all these reasons, I think this is probably the best place to start. After that, you could go any number of directions.
I suppose one purist way to look at it would be to read his books in the order they were written. The big issue there is that you start with The Big U, which I did have some fun with, but which is really only for Stephenson junkies who have read everything else.
However, you could make a compelling case for starting with Zodiac, which I think is one of Stephenson's more underrated or at least, forgotten books (perhaps because it was written before Snow Crash). It's also probably his most accessible book, and it's subject matter is surprisingly relevant even today (it's about a group of environmentalists). If the concepts behind Snow Crash turn you off, you might still enjoy Zodiac a lot. It's a present day story, and not nearly as stylistic as Snow Crash. It also might be his shortest book.
The Diamond Age is a good book for those who loved Snow Crash and it makes for an interesting bridge between Snow Crash and Cryptonomicon (not surprising, as it's the book that was written between those other two). It has a similar Cyberpunky setting, though you are also starting to see a real historical influence, as Stephenson establishes a Victorian undertone layered on top of a more typical SF setting (with nanotech and immersive interactive books, etc...). The one bit of warning about Diamond Age though: I'm convinced that Stephenson's undeserved reputation for bad endings is due to this book (which has a deservedly bad, or at least strangely abrupt ending). It's something I want to revisit at some point to see if the ending makes more sense upon rereading, but still.
Cryptonomicon is great, but as previously mentioned, it's relatively long and it seems to rub some people the wrong way. Still, I consider it to be Stephenson's best novel and it's actually my favorite novel of all time. Following that with the Baroque Cycle makes sense, as they're both part of the same series.
Anathem is his most recent novel, and it is very good. Perhaps not as good as Cryptonomicon or Snow Crash, but excellent in its own right. The only real caveat with this one is that Stephenson kinda invents a new vocabulary in the story, and it takes a little while to get used to the style. That said, it's not a gimmick and there actually ends up being a pretty good reason for it. It's up there towards the top of my rankings, but I also don't think it's an especially good one to start with.
One other interesting idea for a place to start with Stephenson would be the novels he wrote under a pseudonym (Stephen Bury) with his uncle, J. Frederick George - The Cobweb and Interface. They're both written in a more prosaic style and read more like a techno-thriller than Stephenson's other novels. They start with absurd premises (the blurbs about their plots make the books sound awful), but the authors make them seem realistic and populate the world with good characters, then have a less realistic ending. I actually really enjoyed them a lot more than I thought I would, and you can clearly see Stephenson's influence, but they're not as deep as the rest of his stuff. I'd recommend holding off on these until later, but they're definitely worth reading if you're a fan (and maybe even if you're not).
I think that covers all his fiction novels. In terms of Non-Fiction, he actually has a few great books (or, er, reallly long essays). In the Beginning Was the Command Line is horribly out-dated (it's about operating systems, but it was written 10 years ago - before OSX, Win XP, Ubuntu, etc...), but still an entertaining read. Despite being out-dated, it's still relevant because he spends a lot of time talking about cultures and history of the computer and operating systems, etc.. It's also available for free online. In the Kingdom of Mao Bell and Mother Earth Mother Board are two absurdly long articles that Stephenson wrote for Wired in the 90s. The most interesting thing about them is that you can really see how his experiences writing those articles influenced his later novels.
So in terms of a recommended order to tackle his books in, my thoughts seem to point to something like this: Snow Crash, The Diamond Age, Cryptonomicon, The Baroque Cycle, Anathem, Zodiac, Interface, The Cobweb, and finishing off with The Big U. It's a little top-heavy in that his best works are at the front of the list, but I think that's generally how people approach authors anyway.
That list is, of course, purely subjective. I'd love to hear other people's thoughts on the matter...