On distributing reading time

Suppose you've read 90% of Alice's output, 50% of Bob's, and 0% of Charlie's. You have an hour available for reading: whom should you read?

Other things equal, I'd argue for: Alice, then Charlie, then Bob. That is: choose a T-shaped consumption strategy. I think it's broadly accepted that broad sampling in reading is a good idea: the controversial part is advocating for reading the 100th hour of Alice instead of the 30th hour of Bob. For one thing, you have less fresh material from Alice to choose from (other things equal).

But I find myself more and more confident in the T-shaped strategy. Some arguments:

  1. It feels to me like I get outsized value from the thinkers I read the most.
  2. One of my favorite questions to ask impressive thinkers is: "Who is the most obscure person whose output you read all of?" I've never posed this question without getting a good answer. That is, impressive thinkers are pursuing the "consume everything from at least some people" strategy, and I should mimic that strategy.
  3. There is a valuable reward, governed by a threshold, to reading enough of someone's output: you get a mental model of how they think. (Patrick McKenzie put it that way in this interview.) Moreover, it pays to invest in those models even after you cross the threshold, because that improves the model.
  4. The Internet improves the value from both the top and the tail of the T more than it does the middle. There's more and better content to sample broadly from, while the best thinkers are becoming absolute intellectual powerhouses (in part by using the Internet to consume and reply to objections to their work). This is obviously speculative, but I'm very confident in it.
  5. The best thinkers are not only improving the quality of their thought but also finding more and more channels to distribute their output. It's striking to me how often I see footnotes from a generation ago that say things like: "Professor X always said this and argued for it brilliantly, but she never published on it in any form;" I suspect that happens less now. (There's a lot of analysis about how the newsletter renaissance is changing flows of money. That's interesting and important, but I'd love to read more about how it's improving the intellectual outputs of the people best suited to it.)
  6. Many of the arguments for long books also apply to large bodies of work.
  7. Rereading is underrated, so it's a mistake in the first place to view what you've already read as an exhausted resource. And the returns to rereading the best thinkers (best for you) are much higher.
  8. Reading output from a favorite thinker tends to have both a high ceiling and a high floor, and (for me, at least) these increase my motivation to do the reading in the first place. So I'm more likely to read more if I'm reading Alice than if I'm reading Bob, and shifting the marginal time to reading from whatever else I'd be doing is probably valuable.

Nozick famously said of Newcomb's Paradox that almost everyone agrees that the answer is obvious, but they disagree about what the answer is. I suspect this issue is similar.


Home page