How I read

Read clusters of five books. Visualize clusters as instruments to inspect the world. Collect instruments into a mental lab. Read ~40 pages/day. That's ~20 books/year, 40 new instruments per decade.

I try to read a lot, though I'm not nearly as voracious a reader as people who have a reputation for it. I have no idea how Marc Andreessen or Patrick Collison read as much as they do, especially given the enormous demands on their time. The most plausible explanation is that they're aliens with ungodly information processing abilities. I'm human, so I shoot for 40 pages every day, and fail maybe a third of the time. That works out to about 10,000 pages or ~20 books every year.

Reading twenty books a year gets you a lot. Consider: one book gives you more knowledge about a subject than almost every other person on the planet, because people don't read. Two books on the same subject give you more knowledge than almost any reader, because people don't read two books about the same thing. How many people who read The Power Broker went on to read a second book about Robert Moses? I'd wager not many.

Beyond two books on a subject, there is a long utility dip. People who bothered to read more than two books about Robert Moses, are either obsessed with him or study him for a living, which means they've read much, much more about him. So two books on a subject, at least with respect to competitive edge, is an inflection point. After you read two, you get diminishing returns.

There is a way to get around this limitation. A single book is a pinhole view of the world set up by the author. You have no input into its contents, and therefore cannot change the orientation of this view. But you do choose the books you select. That means you can stitch together multiple pinhole views into a unique lens to examine the world— one that no one else will have unless they use the same list of books to stitch together the same lens.

For example, you can look at the world through history of technology that became ubiquitous. Here is one possible list of books to stitch together this lens: The Victorian Internet, Empires of Light, The Wright Brothers, The Network, Hackers. Very few people in the world read all five of these books. Even within Silicon Valley, where everyone's living depends on creating new ubiquitous technology, your understanding of how technology becomes ubiquitous will be in the 95th percentile (and likely higher) if you read these five. You will now be armed with a unique instrument that few others possess, and assuming twenty books per year, it only took you three months to acquire.

Another example is to look at American history and the dynamics of failure by studying biographies of U.S. presidents who sought reelection and lost. There were ten: John Adams, John Quincy Adams, Martin Van Buren, Benjamin Harrison, William H. Taft, Herbert Hoover, Gerald Ford, Jimmy Carter, George H.W. Bush, and now Donald J. Trump. It may be difficult to stomach ten presidential biographies in a row, but it's doable to read five. That's enough to pick up a new lens. Another three months; another sophisticated instrument to examine the world at your disposal.

I settled on clusters of five and almost never read a single book in isolation. Less than five feel lacking; more than five gets repetitive. Every cluster has a goal of the form "study X through Y". Study American history through technological expansion, or study failure through one term presidents are just a few examples. I try to be creative and make Y unusual. For instance, everyone likes to read about presidents who are believed to be successful. A simple trick is to inverse it and read about unsuccessful ones instead. Or skip the presidents altogether, and read about vice presidents. It doesn't matter what Y is because you're trying to study X, and it's more fun to make Y unusual.

I don't bother diversifying books within a cluster by time period, cultural or linguistic background of the author, or anything like that. I simply try and find the best books on the subject. Sometimes they turn out to be diverse along some axes; other times they're homogenous. I never read forewords or prefaces and always finish every book in a cluster. I prefer paper to digital, and used books to new ones. I never take notes or, god forbid, create flashcards. My goal is to suck the juice out of a book, not to hold on to pieces of the carcass. If I ever need to remember a specific detail, it's always waiting for me on the shelf.

This system gives me four new instruments per year, each capable of inspecting the world in a different way. Rather than pick the instruments at random I try to be strategic and collect them into a mental lab. I visualize the lab as a room with multiple stations. Each station has a collection of related instruments on it. When I'm faced with a problem I can go to the right station and examine the problem through the instruments available on that station. All this is physically materialized on my bookshelf. Some day when I become less mobile, I plan to dedicate a literal room to it.

I don't maintain lists of books I want to read, and don't plan a sequence of instruments to acquire in advance. Once I'm done reading a cluster of books, I ask myself: what instrument can I add to double the utility of my lab? Another way to phrase this question is "what is the most important subject that I know the least about?" It's easy to think of many possible answers to this question; I then pick the answer that seems the most interesting, construct a new cluster of five books, and recurse.

Four instruments per year may not seem like much, but consider: in five years you will have about twenty mental instruments, all of which interrelate and reinforce each other. In ten years you will have an impressive lab of forty instruments. And it will be unique— no other mental lab on the planet will be able to inspect the world in the way that yours can.

Because the instruments are constructed from books, they are endlessly upgradeable. Sometimes you might choose to upgrade an existing instrument to increase its power and get higher resolution, rather than acquire a new one. You can do that one book at a time. To use the examples above, you can read another book about ubiquitous technology, or another book about a one term president.

If you organize your reading this way, your bookshelf won't be arranged by genre like a typical bookstore. Rather than having sections for e.g. biographies, essays, novels, military history, all sorted by author, you will instead have sections for addressing different problems, likely sorted by clusters of books in order that you encountered them. The arrangement won't make sense to anyone except you. But you'll be able to find what you need instantaneously, and your capacity to examine the world from different perspectives will dramatically increase.

Waiting five, ten, fifteen years to build a mental lab may seem like an impossibly long time. But you don’t have to wait. It’s a pleasure to read about the Wright brothers, or proliferation of the telegraph, or early computer culture! And in a measly three months you’ll have acquired a fresh way of looking at the world. Already you will be a different person. Then you keep going, and keep looking for new instruments to double the utility of existing ones. And in one year? You will look back at today’s version of yourself and find it unrecognizable.

TL;DR: Read ~40 pages/day, assume 30% failure rate. That's 10k pages and ~20 books annually. Pick a problem, and read clusters of five books to study that problem from a unique perspective. Visualize each cluster as an instrument to inspect the world. Collect instruments into a mental lab, with various stations for related instruments. You can upgrade the instruments one book at a time. Have your bookshelf reflect this mental image. Win the decade, not the day. Start now and never stop.