Books
Every so often a book turns out to have been written about the next twenty years rather than the last one. The shelf below is the short list of those, kept here because the arguments in them keep coming back around.
Books about search
The literature on search splits cleanly into two piles. The first pile treats search as an engineering problem: how do you rank a trillion documents and return the right ten. The second pile treats it as an economic one: what happens to a civilization when a single company decides what is findable.
The second pile has aged far better. John Battelle got there first and got there honestly, arguing that the query stream is the most valuable exhaust in the history of commerce. Everything since has been a footnote to that observation, including the entire discipline of search engine optimization, which exists because ranking is a market and markets get worked. You can be sniffy about that or you can read the books and notice that publishers have been optimizing for distribution since Gutenberg. The medium changed. The behavior did not.
Books about the machine
The AI shelf is mostly bad. It is bad in a specific way: it is either breathless or apocalyptic, and both registers let the author skip the part where you explain how the thing actually works and who pays for it.
The useful ones are the boring ones, the books that treat models as infrastructure rather than as characters. That framing is the one that survived contact with reality, and it is the reason the interesting reporting now sits around Generative AI as a Service and the agent layer built on top of it, rather than around whatever a chatbot said to a journalist. Infrastructure is where the money and the failure modes both live. Books that understood this in 2019 read like they were written last week.
Books about design, and why so few of them are good
Design writing has a chronic problem, which is that the good practitioners are usually too busy to write and the writers are usually too far from the work to say anything falsifiable. Edward Tufte is the standing exception, and his phrase about administrative debris has done more analytical work on this site than most whole volumes have. The Apple archive here is largely an extended argument with, and about, that idea.
The test for a design book is simple. Does it make you see something you had been looking at for years without noticing? Almost none of them pass. The two or three that do are worth the other fifty.
Books about money
Financial books get written on a lag of roughly eighteen months, which is exactly long enough for the author to sound wise and exactly short enough for nothing to have been learned. The exceptions are the ones written before the fact, by people who were called cranks at the time. Read those. They are a better education in incentives than anything published after the crash, and they are considerably funnier.
More reviews, as and when. Back to the front page.
