The High-Density Information Diet

...with apologies to Tim Ferriss for taking his concept and bastardizing it

Many years ago now (in internet time), Tim Ferriss introduced the idea of a low information diet. The premise is described succinctly in his modern classic The 4-Hour Work Week:

Focus on what digerati Kathy Sierra calls “just-in-time” information instead of “just-in-case” information.

The idea is intuitive: it’s bad for your attention to be regularly commanded by events outside your control or by people outside your own social sphere, and most of the time the media is interested in maximizing readership rather than reporting true or useful things. So, in response, don’t read aimlessly! Just read the things you expect to be relevant to you on the short- or medium-term. Understood this way, the argument is distilled to two keywords: read in a pragmatic and deliberate way; avoid the rest.

I enjoy this mindset and have my own information diet of sorts. It looks something like this:

  • I only read newsletters I can consume by email, so I don’t hunt for things I want to read—they find me

  • I only have a Twitter and not a Facebook, Instagram, TikTok, etc. so I’m not overwhelmed with social feeds to check when procrastinating—there’s only one

  • If I think a book might be interesting, I always buy it and send it to my Kindle, but I read at my leisure and don’t pressure myself to finish anything in particular—consequently, I finish a lot of books

  • Anything linked or referenced from any of the above sources is fair game, and this is the primary way I find new people to follow or feeds to subscribe to

  • All my other reading is suspended if I do indeed have a need for just-in-time information due to my job or personal life, such as quickly learning a new skill1

In practice, I still have multiple feeds to check regularly: my personal email, my work email, my productivity app of choice (currently Google Tasks), Twitter, LessWrong, Slack and Linear for work, and sometimes my text messages, Telegram and Signal messengers, or certain IRC chats. This list is much too long for my liking, and probably longer than mid-2000s Tim Ferriss would agree with, but aside from possibly convincing everyone to use either Telegram or Signal and neither the other nor SMS, I don’t see any obvious ways to improve it.

One thing that is nice is that, excepting the several paid Substack newsletters I read, nothing on the above list costs anything. I feel pretty well-informed for the trouble, and at least somewhat better off in terms of attention span (and average degree of clickbait-fueled rage) than some people I know who rely more on social media.

The other nice thing is that most of the above lend themselves well to responses. One should always declare their beliefs publicly, and ideally bet on them, in order to best learn from experience—this incentivizes you to pay attention, to think about your beliefs more carefully than you’d need to do if you didn’t have any skin in the game, and thus to better calibrate one’s future beliefs. So I enjoy chatting with my friends and coworkers, replying on Twitter, responding to other Substack writers, making public predictions, and so on.2 My information diet still casts a wide net.

Primary sources are inefficient

Sometimes I hear a particular criticism that everything I read is someone else’s commentary and thus less valuable than primary sources. I want to address this directly here. I disagree.

A primary source is someone directly talking about their own lived experience or otherwise closely-held knowledge. A primary source would, for instance, be the account of a courtroom appearance from a court stenographer, or the transcription of a speech, or someone’s exact words from an interview, or an ancient writer speaking about their own time and place. So most of what we read is already at least one degree removed from primary.

Journalism standards used to be more strict about presenting opinion in a piece outside an opinion column, but now loaded language is used everywhere in anything you read. (Far be it from the New York Times to refer to anything a Republican says without necessarily including the word “controversial,” lest someone suggest in The Discourse that the paper of record stooped so low as to tacitly agree with a Republican. But anyway.) Even those sources which hold a high degree of cultural authority aren’t necessarily any more neutral than, say, Matthew Yglesias.

Put another way: the signals that society broadly accept as being well-read or well-informed don’t necessarily correlate with actually being informed. Let’s remember also how bad the press is at reporting accurately in the first place.

So this is my first point: a piece written and reviewed by an editor for the New York Times is not inherently more valuable than something written on an individual’s blog. Members of the rationality community online may think of the individual work of such people as Scott Alexander and Eliezer Yudkowsky this way. I struggle to think of any defensible way to argue that the readers of the New York Times are more well-informed than followers of the ACT blogroll.

Second, a potentially obvious point: primary sources are less efficient. With the caveat that one must carefully consider the potential biases in the author’s head, reading summaries is much, much more efficient for information acquisition per unit time. It’s much easier to read a three-sentence (or even one-page) debrief than sit through footage of a three-hour congressional hearing, unless it’s critical that you learn the information live or before someone else. That’s not a game most people need to play.

Put simply: for most knowledge, there is a point of diminishing returns that you easily hit prior to reading primary sources. Reading summaries and narratives produced by other writers also introduces the possibility of learning from the insights of people who have spent years thinking about a certain topic—with the above caveat still in mind that those insights are subject to bias.

But I would much rather be susceptible to whatever biases Zvi has than, say, those of Cade Metz.

What are you optimizing for?

Back to Tim Ferriss: how pragmatic and how deliberate is my information diet? Is it really low-information?

I answer as follows: it’s quite deliberate—I have a system for taking in new things to read and it’s about minimizing the amount of time I spend looking for reads of interest and maximizing my opportunity to respond. It’s pragmatic, but on the level that I enjoy life the most when I’m also reading and writing—this newsletter isn’t my job, but I care about keeping the content high-quality.

I’m not sure if it’s low-information, but it’s certainly high-density information (which is time-efficient) with a strong filter (mostly no social media & no mainstream media) to make sure I’m taking in a satisfying amount of new information per unit time. This works for me.

I also enjoy the premise of having reading days and writing days. I tend not to read anything on the days when I write, and vice-versa. There isn’t any thought behind this position; it’s just something I naturally do. This is also an okay optimization: it seems that when I do things this way, I make the most use of my time in terms of both output and enjoyment.

I invite the reader to consider the deliberateness and pragmatism of their own information diet, in the interest of making sure their interaction with the endless flood of content from the internet is healthy and enjoyable. This post was originally much longer, with two more sections to accompany the one on primary sources, but the other two weren’t all that helpful. I kept the section about primary sources because I want to fight against the idea that mainstream news is automatically more legitimate than other sources of news. It’s important that everyone find a system that works for them and doesn’t preemptively judge their own information diet to be bad or wrong (But please, no arguing in long comment threads on Facebook, okay?)


For instance, last week I started a new full-time job and have barely read anything since, because I’m instead quickly learning a new codebase and internal company culture.


My next post this week will address public predictions.