Adam Barr writes well. I found myself agreeing with several of his analyses: esp. his dissection of MSFT's evangelistic activities and his keen understanding of the api-itis that afflicts MSFT products today.
The book is in four parts. The first is a look at MSFT hiring and interview processes, which is followed by a description of his time at Softimage (which includes a brilliant dissection of type-1 through type-4 demos), then a long and meandering recounting of his early involvement with computers and then an equally meandering final part which is a compilation of his observations about MSFT and the industry in general. I found the first two interesting enough to read, but found the final two not as compelling. He completely mis-understands the point about middleware and Java (see Lou Gerstner's book "Who said elephants can't dance?" for a different definition of middleware and business strategy). Perhaps his narrow, unappealing and unfocussed second half meanders so much because he didn't take his chances to widen his own career within MSFT as a manager or PM.
Like Adam with his interviewees, I agonized over whether or not I should give this book a "four" or a "three" star rating :). Ultimately, I had to go with the lower rating because as a developer, I was hoping to read about what "he" had actually "learnt as a developer" when I picked up the book. Unfortunately, while he talks about a whole lot of things (such as the importance of testing for product quality, and the importance of programmers getting a 'life' as they mature, the contributions of MSFT to the open source movement, etc. etc.) he doesn't at all talk about what he worked on, what was exciting and new about NT code he may have contributed to, or how methodologies and practices changed while he was there. Maybe MSFT prevents people from talking about such stuff, but by cutting out such professionally interesting bits, the book becomes a "missed opportunity" (esp. since Adam is a self described "systems guy"). Perhaps he really was writing only for the non-programmer crowd (but I doubt it).