I must say I'm a bit bemused by some of the earlier reviews...so let's at least try to cut the confusion a bit. First, the book "The Atrocity Archives" (note the plural) contains two separate items: the short novel "The Atrocity Archive" (singular) and the long short story/short novelette "The Concrete Jungle". They are part of a series, i.e. they make use of the same characters in the same world, but there is no reason to expect plot continuity, any more than there would be reason to expect plot continuity between two separate episodes of Star Trek or two Agatha Christie Poirot stories. The separateness of the stories is quite clear from the layout of the book: why some earlier reviewers wanted to read them as one beats me completely. Oh well.
Second, this is actually Charlie Stross's first book, though it's clearly been reissued on the back of his later success, and yes, it does show. This is a book written as a side project by an IT professional, and one feels that other IT professionals were the intended audience. It does, indeed, work better if you're a geek (I'm not, but I am a university physicist so I got most of the in jokes). When it was originally published, the publishers obviously felt, probably rightly, that an introduction by Ken Macleod would help to sell this unknown author - the subtext, that if you like Macleod you're likely to like Stross, is completely justified in my opinion. Yes, the intro could have been dropped for this reissue, but it would probably have cost money to do so.
The stories in this book (and its sequel, "The Jennifer Morgue") are written as affectionate pastiches of classic spy novels, as the Afterword makes clear. The basic idea is that magic actually works, mostly by tapping into alternate universes (probably the "many worlds" of quantum mechanics). Hence, people working in particular branches of applied mathematics (especially geometry and algorithm theory, as you might expect given that magic uses formulae and diagrams) are apt to get more than they bargained for. The Laundry, a branch of British Intelligence, is tasked with (a) heading off mathematicians and computer scientists who are straying into the wrong areas before they accidentally do something catastrophic and (b) dealing with the consequences if they fail to manage (a). However, it's also a government department, which means in addition to this it is expected to conform to government standards for administration, staff development, etc. Been there, done that, and feel exactly the same as Stross evidently does about it!
Disagreeing completely with an earlier reviewer, I think this does work, and (given the historically-attested strongly mystic ethos of large parts of the SS, plus the whole spy-novel pastiche idea) a late-WW2 Nazi SS plot is an entirely sensible choice of "baddie". However, it is true that, after a highly entertaining build-up, the actual climax does seem to me to happen too quickly and be resolved rather too easily. I think this is first-book inexperience; I doubt he'd write the last quarter of the story in the same way if he were writing it now. It's not really a serious fault, because the entertainment value of this book is mostly in the details: the sly digs at various office administration buzzwords (and assorted Microsoft products - a view I suspect Bob Howard shares with his creator). Anyone who's worked in IT or in a university department should find this book very funny indeed. (Incidentally, the mathematicians whose names I recognised - not all of them - are being used in completely appropriate contexts given the book's premise: yes, O previous reviewer, I'm pretty sure Stross does know what he's talking about there.)
If you're looking for sophisticated high-concept hard science fiction as per Accelerando, Glasshouse, etc., this book is not for you. If you're a fan of Dilbert, buy it immediately!
