The devices around us are getting ever more powerful, with their embedded computers. We don't just interact with them, we engage in experiences. Also, they are used "in the wild", in ways unanticipated in the controlled laboratory settings. Buxton's thesis is that part of the development process for these devices requires design. Not just engineering development, but full-blooded up-front industrial design. He's not talking about prototyping, which takes a solution and refines it, but about a process that explores many alternatives, and chooses the best one -- where "best" may require a compromise between many conflicting requirements.
Buxton explains that this process should take a leaf from more traditional design approaches: that of sketching. A sketch is quick, timely, disposable. This gives a clear indication to others that it is low investment, and so can be easily critiqued. Sketches aren't used in isolation: they should also be plentiful: offering a choice between radically different alternatives, not simply an opportunity to tweak a nearly final concept. In interactive device design, a sketch may not be pencil on paper, but it should have these properties: clearly a quick, cheap, disposable mockup of a possible realisation. He gives some illuminating examples of how much experience can be conveyed with extremely simple tools. And these mockups have the advantage that they could never be mistaken for the finished product.
Buxton ends by bemoaning the lack of history and scholarship in the subject today. He gives some fascinating examples of what could be included in such a history, and some pertinent advice on consulting primary sources, not those filtered through others' biased readings. (So, ignore my points in this review: read the book!)
The book is beautifully written, and should be read by all with an interest in making better (more usable, more aesthetically pleasing, more engaging) devices. How much it will be taken seriously by an industry that finds it hard to invest in requirements capture, prototyping, or any "up-front" activity, remains to be seen. But those companies that do it right should prosper.