Blogging – a colossal waste of time?

I was struck by a comment another Inkhaven participant made during one of our first days here: that perhaps blogging is just a waste of time, especially if you want your work to be enduring.

It feels hard to disagree with the notion that most blog posts are just noise, and that even exceptionally good blog posts mostly just get thrown onto the pile of other such blog posts and will be lost to time.

There are of course some exceptional bloggers, whose work will stand the test of time. Scott Alexander is the canonical example given here; Slate Star Codex articles are lore and their ideas have spread far and wide across the memeplex. His current blog has more than 100,000 subscribers, his posts are taught at Harvard, and his readers sent him enough crypto during the boom that he can fund a $200,000-a-year grants programme.

But by mainstream standards he’s still just a very online psychiatrist whose most famous moment was getting doxxed by the NYT. So what does it actually mean to “stand the test of time” as a blogger, compared to say an author of classical literature or even someone who writes textbooks?

Screaming into the void

I’ve heard some people here say that they don’t read any modern works of literature at all, which is pretty fucking funny if you’ve decided to go on a writing retreat for bloggers. But if you care about reading things which are lindy, then it makes sense that you probably shouldn’t read blogs at all – they’re probably among the forms of media which are the least enduring (arguably podcasts are worse, and I do think a lot of people fill their brains with slop by listening to podcasts every minute of the day).

Substack has the same creator dynamics as X or OnlyFans – where a tiny sliver of creators monopolise almost all the attention. The median number of X followers is 1 and the median OnlyFans creator makes $100 a month (it has a Gini coefficient of 0.85, higher than South Africa’s). The top ten authors on Substack collectively make $40 million a year, but only about 5–10% of Substack readers are paid subscribers – the vast majority of writers are giving away content for free.

Although there’s a lot of amazing content on Substack, inevitably most of it will be lost to the void. There are a lot of writers who are literally world-class who get basically no attention at all.

Inkhaven theory of change

Ben Pace talked about how “we all produced exceptional writing and I want to see more of it”. Which was very flattering, and motivating. But it’s not clear that simply writing more, and becoming better as a writer, helps you produce better insights.

The theory of change isn’t that the daily 500 words of slop we each vomit out at 23:59 is going to change the world, but that the program will create new, exceptional bloggers – “the next Scott Alexander or Gwern”.

This is a problem because even Scott Alexander isn’t the next Scott Alexander. It’s possible that Scott’s best work is behind him at this point – that he’s captured all the lindiest, lowest-hanging fruit. And I’m sure while Gwern will produce more great stuff seems unlikely that any of it will be as influential as his earlier work on modafinil, darknet marketplaces, and nicotine.

I do buy the idea that blogging more often is a way to become a better blogger, both through raw honing of the skill, and also because taking more shots on goal increases the chances that you produce a post which is a banger. I also think that writing every day, or at least regularly and often, is likely needed for this to happen for a particular type of blog post. I think Bentham’s Bulldog is a great example of this: lots of shots on goal (like 1000+ at last count), many psychotic takes, some excellent ones.

This is far from being the rule though. Brian Tomasik has dozens of posts on his website, and a great number of them are excellent and convey genuinely novel insights. It was his work which convinced me and many others to go vegetarian and take animal suffering seriously (and I often cite them in conversation).

Likewise, David Pearce in his day was incredibly prolific, and his writing on abolishing suffering with biotechnology put suffering reduction on the map for a great many people, including Tomasik.

Producing better insights

There are other bloggers who have also produced a relatively small amount of extremely insightful writing. I’m not sure the method of posting often as a way to improve your skill as a blogger really generalises to them, except insofar as it makes you a better writer, which is of course not the same as having better insights.

The skills that make a blogger good aren’t really writing skills at all: noticing things other people don’t, synthesising insights from across domains, having priors which turn out to be unusually well-calibrated. Writing well is necessary but not sufficient.

It does seem the case that writing more improves your thinking. Paul Graham talks about putting ideas into words as a mechanism not just for writing things down, but for thinking them through, for making them better and more complete.

That said – it seems hard to believe that simply writing more can produce Scott Alexander levels of insight. He reports that he essentially sits down at the keyboard and has a post flow out of his fingers, no draft, no substantial edits. He can one-shot blog posts that get hundreds of thousands of views. I’m suspicious that any amount of writing can get you to this point – more likely it comes from thinking obsessively for years, gathering interesting experiences from the outside world, and (the real blackpill) having a very specific type of brain.

Eaten by the machine

There’s another way in which blogs can be lindy: by being incorporated into LLM training data.

LLM base models (i.e. the raw model before it’s transformed into a helpful AI assistant by post-training) use huge text corpora as input, including the bulk of information written online. That includes your sick-ass blog! So by “writing for the LLMs”, as some call it, you can be guaranteed immortality in the form of residual tiny changes across their outputs.

The algorithms by which the ‘importance’ of different sources are determined are pretty opaque, although interestingly LLMs themselves do a pretty good job of labelling the importance of different sources. But Wikipedia is intentionally weighted higher as a high-quality source, and gets another boost because it’s mirrored and repeated across the web in various guises.

If it seems odd that an encyclopedia which anyone can edit is weighted so highly, consider that if your edits persist for any appreciable amount of time without being reverted or overwritten, those edits themselves are lindy. Lindy reinforcing lindy!

So if your intention is to have your ideas persist in the substrate of future AI, you might be better off editing Wikipedia than writing a blog post.

Don’t optimise for longevity

Of course it’s possible that one shouldn’t optimise for longevity. I would personally rather write about some weird fascination of mine, or a piece of personal non-fiction that moves other people, than edit some Wikipedia articles on the margin just so they get included in some LLM corpus.

Plausibly there are lots of other reasons to write: to refine your own thinking, to give people useful advice that might not apply in many years’ time, to move the Overton window on a given topic. Many pieces won’t stand the test of deep time but can still provide a huge amount of value and fun in the near-term, and it would be a shame if they didn’t get written.