I add a new post to my old Writing Tips series of posts. This one focuses on simple tools to get the job done.

When I sat down to write this post this morning, I thought I was starting a new series. But then I looked at my list of about 3,100 blog posts here and realized I already had a series it could fit into: Writing Tips. So with a little bit of editing, I wrangled what I’d written today to fit that series. I do, however, want to start with a new introduction.

About Me and This Series

I’ve been a writer since I was 13. I was always driven to write, to tell stories. I wrote them in spiral bound notebooks with the Bic fine point pen I preferred. I filled notebooks with my writing and I still have many of them. This was before the days of computers and word-processing, mind you. The two choices available for writers was longhand or typing. (It wasn’t until much later that I’d get my first typewriter.)

I’ve been a published — yes, by a real publisher — author since 1987 when I was about 26. (You can do the math; it won’t bother me.) That was an article for a professional newsletter. No money, but my first published clip. (There’s a chance I still have it around somewhere; I should hunt it down.) My first paid writing gig came in 1991 as a ghost writer on a John Dvorak book. I wrote 4 chapters and was paid $500 per chapter. My writing career took off from there — I’m sure I’ve covered its trajectory somewhere.

(Hmm. I just did a quick search of blog posts and can’t find one with the full story. I guess I’ll have to write it. In the meantime, here’s a list of all of the books I’ve written and all of the non-blog articles I’ve written.)

The point is that I’m a real writer. I’m driven to write, I have written books and articles that have been published, and I have even made a good living as a writer. Although I’m mostly retired now, I still write; my most recent article for Passagemaker Magazine was published this past week.

I’m telling you all of this because I want you to know who this series is by and for. You know about me; how about you? This series is for people who want to write. It’s not for people who are being forced to write by a boss or job requirements or even a teacher. It’s not going to cover things like grammar and spelling and style. It assumes you already know enough about all that to actually write.

This series is for the folks who, for some reason, want to write but are having trouble doing it. If that’s you, read on.

The Simplest Writing Tool and System

I’ve been thinking a lot about writing lately, looking at ways to get more motivated about doing it and seeing all kinds of weird (to me) posts on my social media platform of choice (Mastodon) from people writing or struggling to write. Two of them just came up in my feed today. I’m including screen grabs of them here for reference and have purposely omitted anything that identifies the poster. (I don’t want to embarrass anyone. If this is your post and you want to be identified as its author, let me know and I’ll modify it for you here.) These posts motivated me to sit down and answer the questions they posed, which I also answered online in replies to them.

In both examples, the author is talking about writing tools or systems to help them write. One mentions bullet journaling, the other lists software platforms. Neither seems ready to acknowledge the simplest tool/system available to them. I’ll give you a hint: it’s the same tool I used when I was 13 years old.

A pen and a notebook.

The Writing Tool/System Trap

Let me start by saying that I was also caught up in the writing tool/system trap that these folks are in. In today’s world, there is always an established and promoted solution for a problem. If someone can cash in on potentially solving a common problem, they will.

That’s what I think of bullet journaling, anyway. It seems that the guy who “invented” it is writing books and creating videos and selling notebooks with dots (instead of lines or grids). You can’t look up “bullet journal” without seeing him pushing his products.

As for software tools, it seems like a race to give writers the perfect app or platform to get their thoughts out there. Even Apple got into the game with its disappointing Journal app. (Why disappointing? Am I the only one who doesn’t want to lock my writing into a cloud-based app?) I even went so far as to spend good money on the app that every writer seems to use and rave about. And then not use it for writing. (I did find another use for it.)

Julia Cameron and The Artist’s Way

A few months ago, still searching for something to help motivate me to sit down and write, I began reading The Artist’s Way by Julia Cameron. This book has been around for 30+ years and still sells well. Every professional writer I know — including me — would love to have written and published a book like this.

Not only did I begin reading it, but I also listened to the audiobook version at the same time — a process some people refer to as “immersive reading” — and took notes. I wanted to squeeze everything I could out of this volume.

I didn’t last long. You see, Julia’s philosophy leans heavy on “spirituality,” her poorly disguised religious beliefs. In her world, a person’s success as an artist (including writers) relies on support from a Great Creator. Although she denies throughout the book that she’s talking about God, you’d have to be an idiot to believe her.

I’m a firm believer that we are each ultimately responsible for our own successes and failures so her Great Creator did not sit well with me. I stopped reading after the Week 2 chapter where her list of 10 “Rules of the Road” included three affirmations that rubbed me the wrong way:

In order to be an artist, I must:

…
6. Be alert, always, for the presence of the Great Creator leading and helping my artist.
…
8. Remember that the Great Creator loves creativity.
…
10. Place this sign in my workplace: Great Creator, I will take care of the quantity. You take care of the quality.

I found #10 especially offensive. (Here’s a newsflash for anyone waiting for a Great Creator to edit their writing: He’s already busy not stopping wars, famine, and injustice. What makes you think your writing is more important to Him than all of those other things that He apparently ignores anyway?) And since it became clear to me by then that a certain amount of religious belief (which I lack) would be necessary to get the most out of the book, I decided not to waste any more time with it.

Anyway, religion aside — I’m certainly not willing to discuss it here so keep your comments about it to yourself, please — I did come away from the book with one thing of great value: Morning Pages.

Morning Pages

The Artist’s Way program requires writers to write Morning Pages every day.

Morning Pages are three full pages, handwritten, about anything you want to write. It could be what you’re thinking or doing or seeing. It could be stream of consciousness. It could be notes for a book you want to write or are already working on. Anything. Words on paper, handwritten.

Since this requirement is stated early on at the beginning of the book (before I decided the book wasn’t for me), I started writing Morning Pages every morning. I had a spiral bound, college ruled notebook — the 8-1/2 x 11 inch kind, not the smaller ones they make to save money on paper. I was already using it for occasional notes, including (ironically) the notes I was taking while reading The Artist’s Way.

I did my Morning Pages every morning during my coffee in bed time with my pups. I’m an early riser, often up before 5 AM. I get up, make coffee, and bring it back to bed. I had been using that time to do some puzzles, catch up on Mastodon, and write a quick journal entry in a blank book I had for that purpose. But instead, I started doing Morning Pages and following that up with the journal entry.

It was easy. I can’t begin to tell you just how easy it was. If I can keep myself distraction-free, I can knock it out in 30 to 45 minutes. That might sound like a long time, but we’re talking about three densely packed pages of writing, roughly 1,000 words. With distractions, the time stretched. But in the 71 days I’ve been doing it — I number the entries — I only missed two days and did a short entry one day — all due to scheduling-related time constraints.

One of the reasons I think it’s so easy is because it doesn’t matter what you write. I usually write about the previous day and what I got done. I write about things I thought about, people I communicated with, and things I want or need to do. I write about the weather and the air quality — it’s smoke season here, after all. I write about my health. I write about ideas for blog posts and articles. I scold myself for not doing things I should be doing.

Seriously: I write about anything that comes to mind. It’s rapid-fire stuff with no formatting or editing. Heck, I don’t even create paragraphs. In one sentence I might wish for rain and in the next I might start writing about a new decorative paper process I’d like to try in my studio.

What is this doing for me? Well, it’s a brain dump, to be sure. Isn’t that what both of the Mastodon post examples I shared above seem to be looking for? It helps me organize my thoughts and create threads that often lead to new thoughts or solutions or even action plans. (I’ve actually begun writing my Morning Pages with my To Do list for the day nearby so I can add items to it.) It has also replaced my journal, which I had been keeping quite faithfully for nearly 2 1/2 years. I was a bit sad to see it go, but I realized that there was a lot of repetition. The additional space I have in that spiral notebook makes it possible to go into detail I could not fit in the old journal page format.

And it’s easy enough to make into a habit. Isn’t that specifically what one of those Mastodon posters wanted to do? Set a time of day to do it, make sure you’ve blocked out enough time, and just get it done.

It also reminded me of how I wrote way back when I was 13. Those notebooks! Those pages of neat printed writing in ink! I was able to be creative back then, to write longhand without editing.

Oh, how I rejoiced years later when word processors made typewriters obsolete! But did they really help me? Or hinder me? Morning Pages makes me wonder.

How Writing Tools/Systems Stifle Us

I mentioned earlier that I was once stuck in the mindset that I needed to use a special tool or system to be productive as a writer. Like the folks I quoted above seem to be. And I tried (and set aside) many software-based tools over the years.

Why? Well I found that once I had decided to use a specific tool to do all my writing, I could not write unless that tool was handy.

For example, when I write on a computer, I want to use a keyboard. A regular keyboard that I can touch-type on. I’m a pretty quick typist when I have the right keyboard, a normal sized one that fits my fingers. I can type nearly as quickly as I can compose in my brain. But needing a keyboard means I have to have a desktop or laptop computer with me to write. I cannot write on a phone or tablet. Unfortunately, it’s not possible to have a laptop or desktop with me all the time. So without that tool, I don’t write.

That’s the most basic problem. The problem is worse when you have a cloud-based tool such as the sites listed in the second Mastodon post I shared above. They require you to be connected to the Internet to use them. Another hurdle. While many folks can’t imagine not being connected to the Internet, I can assure everyone that there are times and places where it simply is not possible to connect. What do you do then?

Using a specific writing app locks you up the same way. When you don’t have access to the app, do you find other ways to write? Or do you not write at all?

That pen and notebook are looking better all the time, no? Pretty easy to slip into a purse, messenger bag, backpack, or tote bag, no? (It doesn’t have to be letter-sized, you know. Try a steno pad.) Easy to pull out when you’re waiting for a friend to meet you for lunch or while killing time at the local DMV or even when heading out to the park for an hour of people watching or mindfulness exercises. I’ve begun taking a tote bag with me when I leave home to do my errands. Guess what’s in the bag? That same Morning Pages notebook; I can write anything in it.

If it’s always there with you, ready to receive what you want to give it, there’s no excuse not to use it to write.

Just Write

Even if you don’t like the idea of Morning Pages, there’s something you can take from them: the idea of writing longhand to get what you want to remember out of your brain and onto paper.

Isn’t that the goal of writing? To write?

Don’t worry about having your writing in a format that isn’t “usable.” You can make it usable later. Consider what you write a draft. Then, when you need to create a final work based on it, you can edit it as you type it into your word processor (or blogging platform or whatever) of choice to create your second draft.

Oh, and I need to clarify one more thing. If you’re already diligently working in a writing tool — like maybe a word processor? — to write a book, story, or article, the advice here might not apply to you for that project. If you’ve already established a writing routine that works for you to create specific pieces, keep using it! The advice I’m offering here is mostly for brain dumps, journaling, and getting the random thoughts floating around your head out on paper where you can consult them later.

Is Writing Important to You?

If it is, stop making excuses that revolve around finding the perfect tool or system. Buy yourself a note book and a good writing instrument and just sit down and write.

But wait! There’s more!

This is not the only thing I’ve written about writing. I’ve been writing about writing since the very beginning of this blog. If you want to do a deep dive, I recommend checking out posts in The Writing Life category or posts tagged writing. Remember, all posts are listed in reverse chronological order with the most recent ones first.

I’m done neglecting this blog, at least for a while.

Shame on me. After blogging regularly here since 2003, I stepped away for more than 3 months. I can feel the air of neglect here.

Understand that my blog has always been a place where I share my thoughts, opinions, and ideas, as well as providing an account of what I’m up to. It’s not as if I have nothing to write about. I do. I’ve been keeping crazy busy this summer, with plenty to blog about: boat trips, new client jobs, writing for hire, exploring artistic endeavors in my studio, and even an unexpected road trip. I’ve been keeping very busy. But what I haven’t been doing is making time to blog about any of it. And that’s a failure on my part.

But I’m back and hope to post new article at least once a week. There are three swishing around in my head right now and I’ll likely get started on the first one this morning.

There is one thing in particular that I’d like to mention here — and will mention in all of my upcoming blog posts. That’s the simple fact that I wrote every single word that appears on this site (with the obvious exception of attributed quotes). I did not use AI in any way, shape, or form. This site is entirely human-created and it always will be.

If you’re sick of reading AI slop — I know that I am! — I urge you to support the human writers who maintain blogs like this by visiting frequently, sharing links to posts, adding your comments at the end of posts you have something to say about, and, when deserved, sending a small contribution via a tip jar, “buy me a coffee,” or paid subscription. This is currently an ad-free, subscription-free blog and I’d like to keep it that way. But your participation is what motivates me to keep writing.

1. This is a great and wonderful contribution and you seem really nice and I’d like to show some appreciation but your comment is so complete and flawless that I can’t think of anything intelligent to say so I guess I won’t respond at all, sorry.

2. This is a great and wonderful contribution and I have some further thoughts so I can respond naturally, great.

3. This is a great and wonderful contribution and I can’t think of anything intelligent to say but I have a joke that’s mildly amusing, I hope you don’t find it cringe or misinterpret it as disagreement or mockery or something.

4. This is a great and wonderful contribution and I agree with 90% of it, but I’d like to hedge on a few points, I guess I’ll point those out while stressing my overall agreement.

5. I utterly disagree with every single thing you said but this is still a great and wonderful contribution because you did a better job than me of representing the view I disagree with, still, it seems like everything has been said and we aren’t going to reach a consensus, so I’ll try to thread the needle of thanking you and conceding what I’m willing to concede without misrepresenting myself as being convinced or sounding dismissive or implying that we should have a lengthy back-and-forth.

6. Everything in your comment seems correct and I completely agree with it, but I’m confused because it seems like there’s some implied disagreement but I have no idea what that disagreement is.

7. This comment seems well-intentioned but it’s based on a epistemology so different from mine that the gap appears unbridgeable.

8. This comment explains what I was trying to say much better than I did, how did you do that.

9. This comment brings up a point that’s worth taking seriously, but the whole purpose of my post was to address this particular objection, so I’m confused why it’s being brought up as novel without any acknowledgement that I have at least attempted to refute it.

10. This comment politely brings up a minor-ish point that I did address somewhere, which is completely fine, it’s unreasonable to expect people to scour every nook and cranny of a post before responding, and other people are surely thinking the same thing, but given that I’ve already written my thoughts on this point, I’d like to link to them without implying that you did anything wrong, but I’m not quite sure how to do that, hmmm.

11. This comment points out a clear mistake, I should acknowledge it and thank you for the correction.

12. This comment points out a clear mistake but is also dripping with sarcasm and implied malintent, why you gotta be like that.

13. This comment is completely confused in a way that reveals to me that my post is itself confusing and I should have written it differently, damn it.

14. This comment is egregiously mean and makes no useful contribution at all, I guess I’ll delete it.

15. This comment is vaguely mean but also makes some interesting points, lest my garden die by pacifism I guess I’ll respond to the substantive points while also gently reminding you that I am a delicate flower and I enjoy human kindness, this will be super awkward but contrary to what you might expect, often works.

16. This comment is about aspartame, I can’t help myself, I absolutely cannot help myself.

17. Are you an AI agent?

18. This comment is thinly-veiled attempt to promote your own blog post, but you needn’t have veiled at all, I want more blogs and more bloggers and especially more blog posts engaging with each other, and I understand that there are today ~zero places you can promote yourself without immediately getting attacked, the social norm that it’s gauche to link to your own posts must change, so please go crazy provided it’s relevant, you aren’t constantly promoting the same thing, and (ideally) you aren’t blogging about a bunch of tweets.

To blog is to get dunked on. I accept this. I even sometimes wonder if I should be grateful, as I suspect my willingness to get dunked on may represent a kind of comparative advantage. (You can tell yourself that if you try to placate the haters, you’ll just ruin things for people who like you. But how do you feel when you’re staring down barrel of a 127 comment thread full of people debating how it’s possible that you’re such an idiot?)

Still, there’s one particular species of dunking that puzzles me. For context, Betteridge’s law states:¹

Any headline that ends in a question mark can be answered by the word no.

This is often employed as a sick burn, as in, You titled your article ‘Is this the world’s first gay caveman?’ because it’s not the world’s first gay caveman but you wanted it to be, because you want attention, you are so bad, har-har.

But I don’t quite understand the rules. Can someone explain the rules?

Question 1: Are question marks in titles always bad?

I’m just checking. I suppose I could see the logic, e.g. if you strongly feel that the bottom line should always come right up front. But I’m pretty sure that’s not the rule, because “this title used a question mark” is not regarded as a sick burn.

Question 2: Are question marks only OK if the essay ends with a full-throated “yes”?

Sometimes it does seem like this is the rule. But it’s strange. If it were universally enforced, we could all mentally convert “Do blue-blocking glasses improve sleep?” into “Yes, blue-blocking glasses really do improve sleep!” But then, of what use was the question mark? Why not just say they’re always bad?

If we’re going to allow questions that are actual questions, then it has to be possible for the answer to sometimes be something other than yes. On the other hand…

Question 3: Is Betteridge’s law useful at all?

I think so. At minimum, you can think of it as a convenient label for this theory:

1. Traditionally, news articles are written with the bottom line up front.
2. Traditionally, news articles have incentives to make a clear affirmative statement in the headline.
3. So if a news article uses a question, that’s because they couldn’t justify making a clear affirmative statement.

I don’t think this theory is 100% accurate. But it’s accurate enough to deserve a name. (On the whole, more theories should have names.) Still, Betteridge’s law isn’t usually invoked as a neutral observation about the forces that led to a given title. It’s usually invoked as a dunk. So…

Question 4: Is Betteridge dunking ever appropriate?

Again, I think it is. Here are some of the best/worst examples from John Rentoul’s book, “Questions to Which the Answer Is No!”:

• “Will Guam capsize?”
• “Is Osama Bin Laden in Chicago?”
• “Did Jesus foresee the US Constitution?”
• “Des smartphones bientôt équipés d’airbags?”

I think we can agree something is wrong with these. But what, exactly?

Question 5: Is it central that the answer is “no”?

Consider these made-up titles:

• “Is the Pope still Catholic?”
• “Do you need to sleep every day?”
• “Did Lincoln have personal qualms about slavery?”
• “Did the Rubicon even exist back when Caesar supposedly crossed it?”

These are anti-Betteridges. The answer is yes, but the title is irritating in the same way: It gives the impression of a live debate when none exists.

Question 6: What’s really going on here?

I think it’s pretty clear. Consider the title:

Is aspartame bad for you?

If you understood it to be a settled question that aspartame is safe, and the article ultimately concludes that aspartame is safe, then you might find that title annoying. On the other hand, if you understood it to be settled that aspartame is bad for you, and the article confirms that yes indeed it is bad for you, then you also might find that title annoying.

The answer is immaterial. What’s irritating is when a title suggests a novel, interesting possibility that the article does not substantiate as worthy of attention.

Question 7: So what’s the problem?

Here’s a proposition: The modern internet rewards people for being overconfident. I don’t know if you’ve noticed, but people with blogs are not constrained by the norms of traditional newspapers. On the contrary, if you start a blog, you will soon learn that the best way to get attention is to write spicy aggressive titles like, “No, creatine does not make you smarter despite what all the stupid dumb mouth-breathing supplement hucksters may tell you.”

Now, I do think you should say what you actually believe. If you truly are that confident, I want you to tell me, not bullshit me by pretending to be neutral.

However, the internet corrupts all of us. Many people seem to start out with a public persona that is careful and measured and calm. But over time, they’re gradually sculpted by the Reward Function into something quite different. The degree this happens depends on your personality, where you’re competing for attention² and how much you try to resist. But I don’t think anyone is truly above this.

Still, we should try to resist. My favorite kind of essay is, “Lucid examination of all sides of an issue which finds some evidence pointing in various directions and doesn’t reach a definitive conclusion because the world is complicated.” And I think the fundamental goal of a title should be to accurately signal the contents. But how is such an essay supposed to signal what it is, if not by using a question?

Question 8: What should a title do?

One theory is that question titles are sort of like lists: A thing with strong fundamental merits that has been rendered suspicious by abuse. Under this theory, we should push back against all the Betteridgeing and insist that question titles are fine when the question is genuinely open, regardless of the answer, and that people are wrong to Betteridge unless the question mark is being abused.

As far as I can tell, that’s the only internally consistent theory that doesn’t amount to saying that question titles should be forbidden. A slightly more conciliatory version would be that if you use a question mark, it’s your responsibility to demonstrate that it’s a real question, not something you made up.

I lean towards that theory. But part of me—a minority—thinks that perhaps question titles should be effectively forbidden. I thought I’d do a little reductio ad absurdum by trying to give this post an accurate non-question title. The best things I could think of were, “Hesitantly against over-broad Betteridge dunking” and “I weakly think excessive Betteridge dunking disincentivizes fairly examining all sides of an issue.” At first, I thought those were amusingly terrible. But are they, really?

PS. Was Rentoul’s book correct to list, “Should we clone Neanderthals?” as an example of a question to which the answer is no?

Here’s a conjecture: If you put any significant amount of text on the internet under different names, those identities can be linked using only the text itself. This is possible (I conject) because of the statistical “fingerprint” you leave in everything you write.

Imagine a website where you can paste in some brand-new text someone just wrote. In return, the website provides links to all the text that writer has ever published under any name. It’s not perfect, but it’s pretty good. As far as I know, no such website exists—at least not on the public internet. But I suspect it’s possible and will soon become easy. This will pose some difficulty for pseudonymous blogging.

Note: I wrote most of this essay in mid-2025, after which I idiotically sat on it for a year tinkering with theorem statements that none of you will read.¹ In the meantime, LLMs have gotten much better at guessing authors from text. (Given the first 1000 words of a draft of this post, Claude 4.8 knows it’s me.) Still, I think we’re just getting started. I expect to see increasingly obscure writers identified from increasingly small bits of text. I expect that this work even when people are writing in a different register or about unrelated subjects. And I expect that everything I’ve ever written under any pseudonym will soon be linked to my genuine-nym.²

A stronger conjecture is that we’re heading towards a sort of generalized pseudpocalypse. Perhaps, in the future, if you interact with the world through essentially any high-bandwidth channel, then you identify yourself. Say you wear a mask in public and only speak by sub-vocalizing into a voice changer. That’s fine, you’ll still be identified using your body shape, gait, or chemical signature. Or say you don’t like your car being tracked everywhere, so you stop carrying a phone and you somehow convince lawmakers to ban license plates. No problem, your car will still be tracked using tiny scratches or unique pinging sounds from the engine. Or say you don’t like being tracked on the internet, so you lock down your browser profile, buy stuff only with Monero, and connect through a chain of three VPNs. That’s OK. You’ll still be identified through how you wiggle your finger as you scroll down the page. We’re all just too unique, and the information theoretic limit is coming for us.

Starting bits

Let’s start from first principles. Imagine that at birth, everyone is assigned a random binary string. Whenever you post anything on the internet, you’re required to sign it with that string. If the strings are very short, like 0110, then lots of other people will have the same one as you. But if the strings are very long, then yours would almost certainly be unique and it would be trivial to link all your pseudonyms.

Where’s the transition point? If you only know that the author is currently alive and living somewhere in the Anglosphere, it’s around 29 bits. That’s because if there are K digits, then there are 2ᴷ possible binary strings, and if K = 28.86, then 2ᴷ ≈ 490,000,000 is the number of currently-alive Anglosphere-dwellers. If the strings have fewer than 29 bits, then someone else will probably share your string. If they have more than 29 bits, then your string is probably unique.

We don’t (yet?) have to sign the things we write with immutable government-issued strings. But the way you write still provides lots of clues about you by way of your tone, personality, word choice, and so on.

Theoretically speaking, I think it has to be possible to link the identities of anyone who writes enough. Imagine again that everyone is assigned a random binary string at birth, but instead of you needing to sign the stuff you write with your string, each time you write a word, there’s some chance that a random bit from your string is revealed and added as a signature to your message. For example, maybe a signature of bit[129]=1 is added, indicating that your string at position 129 has value 1.

Think of your string as representing all your writing style quirks, and a bit being revealed as representing when you write something that reveals a preference. For example, maybe bit 18 indicates if you prefer to write your em-dashes with hideous spaces — like this — or without spaces—like this. If you use an em-dash, that bit is revealed.

So imagine you’ve written a lot under Pseudonym A, enough that the full bit-string has been revealed. Maybe it’s this:

Pseudonym A: 
110000001111001101110000100001
010100100101011110111001101000
100111110010101001101010111010

Now say you start writing under Pseudonym B. Initially, none of the bits will be known:

Pseudonym B: 
??????????????????????????????
??????????????????????????????
??????????????????????????????

But slowly, you’ll start to leak a few bits:

Pseudonym B:
?????00???1????1????????1??0??
?????01??1?????????11????01???
???1??????1???10??????????????

And eventually you’ll leak a lot of bits:

Pseudonym B:
???0?00?1?11??110??1????10?0??
?10?001?0101?11???11100?101???
???11?1??010??10?1?01??011????

Now think about this from the perspective of an “attacker” who wants to know if A and B are the same person. Let’s assume they’ve only seen the above bits, and have no information about anyone else. Then here’s what the attacker knows:

1. A and B have revealed K overlapping bits, which all match.
2. Different people have a 50% chance of matching on any given revealed bit.
3. Non-different people have a 100% chance of matching on any given revealed bit.
4. There are 490,000,000 people.

Intuitively, if K was 5, then the fact that all bits match wouldn’t prove much, since with 490 million people, lots of people would match on those bits by chance. But if K was 70, it’s extremely unlikely that two different people would share all of them, even with such a gigantic pool to start with. It turns out that if there are N other people with random bits, and you pick K of your bits, the probability that someone exists who matches all of them is 1 - (1-2⁻ᴷ)ᴺ. When N is 490 million, that looks like this:

Look at that, 29 appears again. (Isn’t math wonderful?) In general, the transition happens around whatever number of bits K makes 2ᴷ ≈ N, namely K = log₂(N).

If you reveal significantly fewer than 29 bits under pseudonym B, then it’s almost guaranteed that there’s someone else out there who matches all of them. But if you reveal significantly more than 29 bits, then there’s almost no chance that anyone else exists who matches all of them. So the attacker essentially knows that A and B are the same person. And I stress again: They know that without needing to see anything from the other 490 million people.

Of course, we don’t literally leak bits of immutable feature strings as we write. But you can make the model more realistic, and the same issue persists. If you want to reflect that text only provides noisy information about the writer, then you can add noise to the bits before they’re revealed. If you want to reflect that some writing styles are more common than others, then you can make the distribution over bit strings non-uniform. If you want to reflect that certain quirks are more obvious than others, you can give different bits different probabilities of being revealed. All these make the math more complicated. But they don’t change the basic conclusion: If your writing style contains at least 29 bits of information, and you do enough writing, you’re done.

That’s my argument that pseudpocalypse is possible. But I don’t just want to claim that it could happen, eventually. I think it is likely to happen, soon, and that the amount of text you need to reveal isn’t very large. To make that argument, we need to get specific: What features do people have that are reflected in their writing? How many bits of information do those features contain? How accurately can those bits be guessed from written text?

Note: To avoid this turning into a giant information theory lecture, I’ll mostly use words like “bit” and “information” without being 100% fully precise about what they mean. I’m doing that because I expect that most people reading this aren’t definition-of-bit fetishists, and anyway being hyper-technical would obscure the big picture. If you’re an information theory enthusiast and/or skeptical that I know what I’m doing, I refer you to the Section For Skeptical Information Theory Enthusiasts, below. Until then, use your intuition and have faith.

Feature space

Say you knew nothing about me other than that I wrote the above words. And say you had to guess my age or religion or occupation. You could guess, right? It wouldn’t be perfect, but you’d do much better than you would without being able to read those words. Thus, somehow, those words contain information about my demographic characteristics. So I tried to make a list of similar things that you could plausibly guess from text at least somewhat better then chance. Here’s what I came up with:

• Age
• Education
• Ethnicity
• Family status
• Income
• Marital status
• Mental health
• Native language
• Occupation
• Physical health
• Political leanings
• Region
• Religious affiliation
• Sex

In the same spirit, if you only read the above words, could you guess how extroverted or conscientious I am? Again, not perfectly. (When I meet people who read this blog, they usually seem surprised I can survive direct sunlight.) But still, I’m sure you’d do OK. So, again, these words contain information about my personality.

What features does personality have? The HEXACO model lists six, namely honesty-humility, emotionality, extraversion, agreeableness, conscientiousness, and openness to experience. I suspect those can all be guessed with reasonable accuracy from a long-enough writing sample. But could you guess more? For each of those six factors, the HEXACO model lists four “facets”. In the abstract, trying to guess 6 × 4 = 24 different personality features from text sounds ludicrous, but just look at them:

• Honesty-humility
  • Sincerity
  • Fairness
  • Greed avoidance
  • Modesty
• Emotionality
  • Fearfulness
  • Anxiety
  • Dependence
  • Sentimentality
• Extraversion
  • Social self-esteem
  • Social boldness
  • Sociability
  • Liveliness
• Agreeableness
  • Forgivingness
  • Gentleness
  • Flexibility
  • Patience
• Conscientiousness
  • Organization
  • Diligence
  • Perfectionism
  • Prudence
• Openness to experience
  • Aesthetic appreciation
  • Inquisitiveness
  • Creativity
  • Unconventionality

If you think about specific people, I think you can convince yourself that these 24 represent real things, and that it’s plausible to guess them from text. (Your favorite existential angst + science blogger, for example, might score lower on “modesty” than the other honesty-humility facets.) The different sub-factors are surely correlated, but not perfectly correlated.

Of course, the biggest thing you learn from people’s writing is how they write. Do they tend to pointlessly split infinitives? Do they use hyphen-connected words? Do they, incorrectly, position their adverbial clauses?

The idea of attributing authorship using writing style features goes back to at least 1440, when Lorenzo Valla demonstrated that the Donation of Constantine—in which Emperor Constantine supposedly donated the Roman Empire to the Catholic Church—used a vernacular that came from 400 years after Constantine’s death and was therefore a forgery. In 1851, Augustus De Morgan observed that average word length tends to be stable for the same author. The first “modern” attempt seemingly came in 1964, when Mosteller and Wallace published Inference in an Authorship Problem:

This study [attempts] to solve the authorship question of The Federalist papers; […]

Word counts are the variables used for discrimination. Since the topic written about heavily influences the rate with which a word is used, care in selection of words is necessary. The filler words of the language such as an, of, and upon, and, more generally, articles, preposition, and conjunctions provide fairly stable rates, whereas more meaningful words like war, executive, and legislature do not.

After an investigation of the distribution of these counts, the authors execute an analysis […] based on Bayesian methods. The conclusions about the authorship problem are that Madison rather than Hamilton wrote all 12 of the disputed papers.

Get that? The idea is that your usage of the word war depends mostly on if you happen to be talking about war. But your usage of upon mostly depends mostly on how much you like the word upon. To demonstrate this, they took 48 papers written by Hamilton and 50 by Madison and made this table of how many times they used by, from, and to:

Madison liked by. Hamilton was more a to man. Using these kinds of statistics, they concluded that the disputed Federalist papers must have been written by Madison.

So I did some research looking for other writing style features that are believed to be stable when people write about different subjects. I found that there are a lot. There were so many that I struggle to even organize them into meaningful groups:

Low-level frequencies:

• Word lengths
• Sentence lengths
• Paragraph lengths
• Punctuation frequencies (commas, colons, dashes, parentheses)
• Function word frequencies (the, of, and, to)
• Adverb frequencies
  • Intensifiers (very, really, quite, pretty, so)
  • Evidential markers (apparently, evidently, obviously)
  • Downtoners (somewhat, fairly, rather)
• Pronoun usage
  • Overall preferences (I/we vs. you vs. he/she/they)
  • Third-person singular preferences (he, she, he or she, they, one)
• Modal verbs (can, could, might, must, should, will, would)
• Hedges (perhaps, maybe, possibly, probably)
• Conjunctions (and, but, yet, so)
• Known stable ratios (the/a, this/that, these/those, I/me/my)
• Character N-grams (3-grams and 4-grams)
• Word N-grams (often 3-grams)

Lexical features:

• Vocabulary size
• Lexical diversity / type-token ratio (Number of distinct words divided by number of words.)
• Frequencies of rare words
• Semantic density
• Discourse marker positions, combinations (So, anyway, so anyway)
• Use of abbreviations and acronyms
• Preference for latinate vs. germanic words (The majestic creature traversed the terrain vs. the mighty beast strode across the land.)

Syntactic features:

• Syntactic complexity
  • Subordination index
  • Average parse tree depth
• Use of passive voice.
• Nominalization (She was shocked I ate the pizza vs. My pizza consumption shocked her)
• Verb tense and aspect (I walk vs I walked vs I was walking vs I have walked)
• Sentence structure preferences:
  • Branching preferences (Cursed everyone had a good time when Alice taught some cool dogs I met and brought to dinner to juggle vs. clumsy-but-readable I met some dogs and they were cool and I took them to dinner and Alice taught them to juggle and and everyone had a good time.)
  • Adverbial clause positioning (Suddenly I was hungry vs. I was, suddenly, hungry vs. I was hungry, suddenly)
  • Sentence-final weight (Your plan won’t work because of the dyslexic bears vs. Dyslexic bears mean your plan won’t work.)
• Polysyndeton (I like dogs, cats, and ferrets vs. I like dogs and cats and ferrets.)
• Repetition / breaking of syntactic structures.

Style features:

• Register / formality.
• Patterns in sentence length (long/short/long/short vs. long/long/short/short)
• Stressed syllable interval preferences (e.g. iambic vs. trochaic)

Rule preference features:

• Minor punctuation (I laughed—you cried vs. I laughed — you cried, “…” (three periods) vs. “…” an actual ellipsis)
• Capitalization. (Job titles, seasons, after a colon, mistakes)
• Apostrophes (Steve Jobs’ car vs Steve Jobs’s car, 1990’s vs 1990s)
• Hyphenation (a highly-stable feature vs a highly stable feature)
• Oxford commas.
• Article omissions (Local dog was petted. vs. A local dog was petted.)
• Relative pronoun omissions (the dog you petted vs. the dog that you petted)
• Who vs. whom.
• Split infinitives (To obsessively blog vs. to blog obsessively)

Idiosyncratic features:

• Whitespace habits.
• Spelling errors (loose instead of lose)
• Grammar errors. (Between you and I)
• Consistent, unique typos
• Other consistent errors (repeated words, un-closed parentheses)

That’s a lot. There are surely more. And these are all “shallow” features that humans came up with using our tiny little brains. I strongly suspect that there are many more “deep” features that could be found by looking for statistical patterns in a sufficiently large dataset. Many of those features might not even have a coherent English-language description. But they’re still there, providing bits for those who seek them.

So we leak information about lots of different stuff when we write. But how much information? Is it possible to say how many words are needed to uniquely fingerprint someone?

No. To a first approximation, the answer is no. But to a second approximation, maybe? Within an order of magnitude? I’ll try, but it’s going to be hard.

Demographic bits

How many bits of identifying information does text provide by way of demographic features like age and sex and so on?

At first glance, this seems a perilous question, as it depends on the number of categories you consider those things to have. Take sex. For pseudpocalypse purposes, your opinion about how sex should be defined or how many sexes exist is irrelevant. Finer categorizations always provide more information, and our de-pseudonymizing attacker friends will use that information if they can. However, going beyond two categories for sex makes little difference, because the additional categories will be hard to guess and even if you could, categories with low prevalence don’t contribute much extra information.³ So, for us, two categories is the right answer.

And what about age? At first glance, converting age into a set of categories seems meaningless. If you code age by the millisecond, then there are 3.156 trillion categories for people born in the last 100 years. If you code age by the decade, there are only 10. Here, the thing to notice is that while you might be able to guess my decade of birth from how I write, you don’t have a snowball’s chance in hell of guessing the millisecond. (See what I did there? People born in certain decades are more likely to use expressions like snowball’s chance in hell?⁴) If we took age to have some crazy number of categories, we’d have to discount later to reflect the difficulty of guessing. My intuition is that it would be hard to guess age more accurately than around five years, so 20 categories seems reasonable.

Following this kind of logic, I chose a number of categories for each of the demographic variables, trying to hit the upper end of what could be guessed from text. (I’ll provide the actual categories below.)

If each of the age bins were equally likely, then knowing what bin someone fell into would provide 4.32 bits of information, because 2ᴷ ≈ 20 when K = 4.32. Doing that same calculation for each feature gives the maximum amount of information they could contain.

But there’s a problem. There are more people aged 30-35 than there are people aged 90-95. So, even if you could guess those age bins perfectly, they’d provide less than 4.32 bits of information on average. However, it turns out that categories need to get pretty damned uneven before information content drops very much. A perfectly balanced 50/50 distribution provides 1 bit of information, but if you switch to a 60/40 distribution, you still get 0.971 bits, and you need to go almost to 90/10 before information content drops to 0.5 bits.⁵ The same basic thing is true when there are more than two categories.⁶

So I went through all those features, rated them by how unevenly people are distributed, and tried to discount the bits accordingly. I’ve put the full details of what the original categories are and how I discounted them in a footnote.⁷

But there’s another problem. Female 65 to 70 year-old Asians living in Scotland tend to have different {occupations, family statuses, religious affiliations} than 15 to 20 year-old Latinos living in Southeast Australia. That is, the above features are correlated. So as you look at more of them, they gradually become less surprising and thus contribute less information.

How much less? Answering that the right way would require us to estimate how likely someone is to fall into each of the 20 × 6 × 6 × 2 × 11 × 3 × 3 × 2 × 23 × 3 × 3 × 23 × 3 × 2 = 8,144,737,920 joint categories. That seems hard. But a not-completely-ridiculous approximation is that if a group of variables are all pairwise correlated at a level of ρ>0, then the total information might be reduced by a fraction of ρ.⁸

So how correlated are those features? In the social sciences, a correlation of 0.5 is considered quite high. That’s plausible for some pairs of variables, e.g. age vs. health or political leaning vs. religious affiliation. But many of those correlations are are probably quite weak, e.g. age vs. native language or region vs. sex vs. marital status.⁹

Overall, my guess is that correlations reduce the total information by at least 10% but I doubt they reduce it by more than 60%. So I’d think the total information in the above features (if you could guess the categories perfectly) is somewhere between 10.6 and 23.9 bits. Let’s take the average and call it 17.2 bits.

Personality bits

What about personality features? Let’s use the same same recipe we used for demographic features, but faster: To start, let’s give each of the 24 personality features five bins, in deference to dynomight personality notation. That would correspond to 24 × 2.32 = 55.68 bits total, because 2ᴷ ≈ 5 when K = 2.32.

Then we need to discount for correlations. The six main HEXACO personality factors are designed to be uncorrelated, but the different “facets” inside each factor are correlated (usually with a coefficient between 0.3 and 0.6). It seems reasonable to use an overall discount factor of 0.3 to reflect strong intra-factor correlations but weak inter-factor correlations. That suggests 39.0 bits overall.

Style bits

And what about writing style features? How much information do they contain?

This seems hard. Some of the features, like character n-grams are actually themselves long lists of features. (Frequency of typing aaa, frequency of typing aab, etc.) However, many of those features contain little information, since almost everyone types zqx around 0% of the time. And, of course, writing style features are correlated, since people who write realise instead of realize are less likely to put spaces around their em-dashes.

In absence of a better idea, I’m going to give one bit for each leaf node in the above list of style features. I think of this as giving each feature two bins, and then assuming that uneven distributions of features and correlations (which reduce information) are canceled out by the fact that many features deserve more than one bin and that there are probably more “deep” features that aren’t listed (which increase information). This gives us the suspiciously round number of 50.0 bits.

Guessing bits

If you believe the above numbers, then we have at least 17.2 + 39.0 + 50.0 = 106.2 bits of identifying information that we leave clues about when we write. That’s a lot. If you could see all those features, it would be enough to identify people even on a planet with 93 million trillion trillion people.

But to argue that the pseudpocalypse is nigh, it’s not enough to argue that those bits exist. We need to argue that they can and will be guessed from a relatively small amount of text.

So obviously we need to talk about nuclear weapons. In a nuclear detonation, many unstable atoms are created. These spontaneously decay into more-stable atoms, in the process emitting radiation. Some types of atoms are very eager to decay, meaning they release a lot of radiation but stop existing within a few weeks (iodine-131). Others are reluctant to decay, meaning they don’t release as much radiation but they stick around for decades (strontium-90). Others stick around for millions of years, but they produce so little radiation that they’re not a big problem (cesium-135).¹⁰ So, the residual radiation produced after a nuclear detonation is the sum of many different exponential curves, one for each isotope created during the detonation.

I suspect that identifying bits in text are sort of like that. Your level of formality and your average sentence length are revealed almost immediately. Your preference for latinate vs. germanic words takes a while to come through. And your social boldness and the fact that you live in Queensland rather than Southeast Australia are revealed very slowly, perhaps so slowly that it’s effectively not revealed at all.

Right. So if you start with 106.2 bits, how many of those do you reveal after writing a given number of words?

I will answer that question through the noble method of making up numbers. But first, let’s calibrate. You just read 4500 words written by me. How well could you guess my demographic and personality features? As a sanity check, I gave the above words to an LLM and asked it to guess. It did unnervingly well. It wasn’t always right, but it usually was, and it did a great job of rating the confidence of the individual predictions.

I don’t think there’s any magical explanation for this. The fact is, if you look at the individual personality and demographic features, guessing them just isn’t that hard. So I’m sure you could do just as well. And given enough time, I’m pretty sure you’d do even better for writing style features.

Even so, you’re probably bad at it. Take the example of GeoGuessr, where people guess a location in the world from a random photo. Random people are sort of OK, but if you pick the top natural talents and have them practice obsessively, they’re really good. I don’t think LLMs are particularly good at guessing features from text, either. They weren’t trained for it. It’s just an emergent property of their general intelligence. The information-theoretic limit is surely much higher.

So here’s a very rough cut: After 4500 words, I’d think it’s possible to guess around:

• 60% of the demographic features
• 70% of the personality features
• 80% writing style features

If we model each of those with a separate exponential, and start them at 17.2 / 39.0 / 50.0 bits, then the total number of identifying bits that remain hidden after writing a given number of words is as plotted here:¹¹

Et voilà, pseudonymity is compromised when you leak 29 bits, which happens after 1071 words.

Seriously?

Of course not. The above figure stands on a creaking tower of tenuous assumptions. I’ve gone through the details of deriving that curve not because you should trust it, but because I think seeing the calculations makes the following points hard to argue with:

• You have far more than 29 bits of identifying information that you leak into your writing.
• Some of those bits take a long time to get revealed, but others are revealed pretty quickly.
• There are enough “fast leaking bits” that you can be identified from a writing sample that’s “pretty small”.

I’ve made lots of debatable choices in terms of choosing features, assigning numbers of categories, estimating distributions across those categories, discounting for correlations, and guessing how many bins can be guessed. Those choices are all individually suspect. But the above points are supported by a pretty wide margin of error. You can make different choices, but it seems very hard to avoid concluding that the above three points are true.¹²

How would this work?

You might be wondering why I’m using so many made-up numbers. After all, there’s a whole field devoted to identifying authors from text, usually called “stylometry” or “authorship attribution”. They have research papers and competitions and all that. However, as best I can tell, state of the art published results look something like this:

1. Take 50 people.
2. Get a few hundred writing samples from each author, each 1000-2000 words long.
3. Now, take a new writing sample from one of those authors.
4. Do some standard machine learning stuff.
5. Hey look, the author can be identified with ~95% accuracy!

That sounds OK, but that’s only identifying people against a pool of ~50 authors. For my claim to be true, similar accuracy would have to be possible with 490 million people. That’s seven orders of magnitude more.

The thing is, the methods those papers are using are extremely weak. All the above math assumes that you’re operating at the “information-theoretic limit”, making perfect use of all available information. If you want to get close to that, we now have some idea how to do it: You apply the “modern” machine learning recipe of gigantic dataset + gigantic neural network + gigantic fortune spent on GPUs. My guess is that for us, that would require something on the order of “all the words ever written” + “tens of billions of parameters” + “tens of millions of dollars”. I couldn’t find a single paper that came remotely close to attempting that.

So I don’t think those papers tell us much, for the same reason that a 3rd-order Markov model trained on a few books doesn’t tell us much about how good computers could be at writing text. LLMs have shown that if you use the above recipe, then computers can get close to the information-theoretic limit for generating text.¹³ So, I suspect that an LLM-level effort could achieve the same thing for identifying authors.

You might also wonder: Why am I talking about this as some possible future technology? Isn’t that technology just LLMs?

I suspect the technology will be quite LLM-like in how it models human language. But current general-purpose LLMs aren’t trained for this task. They’re good at it “by accident”. So, just like specialized chess AIs can crush LLMs at chess, I suspect specialized stylometry methods could crush general-purpose LLMs at stylometry. It’s just that those specialized stylometry methods don’t seem to exist yet, or at least aren’t public.¹⁴ So we shouldn’t imagine that current LLMs are anything close to what’s possible, even if you assume that generic LLM progress stopped today.¹⁵

Countermeasures

If this is all true, what could be done about it?

The most obvious “countermeasure” would be to get used to it. I mean, imagine that we did live in a world in which everyone literally had to sign everything they wrote with a unique immutable string. What would happen? I’d expect a mixture of:

1. People become more comfortable with their “full selves” being public, with less compartmentalization.
2. People pull back from communicating in public channels, relying more on group chats and the like.
3. People self-censor.

There are strong historical analogies here, since over the past 20 years many governments and tech companies have in fact decreed that people must sign the things they write with their real names.

The effects seem to vary quite a lot based on the ambient culture and political system. Overall, my impression is that people are already much more comfortable with the idea that their work colleagues might read their dating profile or learn that they go to furry conventions. I’m optimistic that culture will continue to adapt to respect the fact that we all encompass multitudes. This seems healthy.

Some effects seem clearly positive. Self-censoring is not necessarily bad. For example, on the margin, real-names surely stop some teenagers from engaging in cyber-bullying. On the other hand, were you ever a teenager? I’m pretty sure that for anyone who is “different”, having those differences broadcast to the world creates a much larger “bullying surface area”. So the effects are mixed. And adults aren’t as different from teenagers as we might like to think.

Twenty years ago, I might have predicted that real names would discourage people from expressing controversial political ideas online. Superficially, that seems completely wrong. At least in the West, lots of people are very happy to express minority political views, and if you disagree at all, then you can go to hell. But I also tend to think this hides a lot of self-censorship, where most people don’t want engage in political mortal combat and so are cowed by a feisty minority. And, obviously, people in certain countries know that it’s unwise to criticize the Party. So, getting used to it seems like an imperfect solution at best.

Another countermeasure would be to not build this technology, or not make it widely available. In the short term, this seems plausible. As far as I can tell, it’s been possible for years for a modestly-funded group to build a phone app that would identify most people on the street from a photo. And yet, almost no one reading this has access to such an app. If general-purpose LLMs continue to get better at stylometry, it seems entirely possible that AI companies might decide it’s a safety issue and train their AIs to refuse to do it.¹⁶ This could work for a while.

But if the technology is possible, it seems certain that governments will build it and use it. They might try to keep it out of the hands of normal people. Certain governments might restrict their own use. My privacy-minded allies always seem very jaded, but it wouldn’t surprise me at all if the Supreme Court declared that a warrant was needed before the FBI could de-pseudonymize a U.S. citizen. But when/if that technology becomes sufficiently cheap, it seems like it would be very difficult to keep it out of the hands of normal people and/or bad actors. My guess is that it’s possible to create a program that’s a few hundred gigabytes large and can run (slowly) on most modern laptops. If that program is made public, it would be hard to put the genie back in the bottle.

There are also technological countermeasures. Most obviously, you could run your writing through a “filter” to try to remove identifying bits, e.g. by asking an LLM to rewrite it. It’s hard to be sure how well this would work, since we don’t have accurate estimates of how many bits you’re starting with or how many bits this would remove. But I’d guess this would be pretty effective if done carefully. The reason is that the number of identifying bits you leave in writing probably isn’t that large, relative to the number needed to identify you. If you “homogenize” your writing to remove all style and personality, you should be able to remove most of those bits. Theoretically, you’ll still leak some information. But I’d think this would substantially increase the amount you could write while remaining pseudonymous.¹⁷

But after thinking about it, this makes me sad. Effectively, this countermeasure would preserve pseudonymity by taking writing and destroying all traces of humanity. It seems like this would work well for the “bad” uses of pseudonymity, like cyber-bullying or coordinated violence, but it wouldn’t work at all for the “good” uses, like for example someone who likes to write pseudonymously because they feel like it allows them to be more honest and vulnerable and more fully themselves, damn it.

Generalized pseudpocalypse

Maybe this isn’t just true for writing. Maybe it’s just a feature of our universe that if you interact with the world in any significant way, then you leave traces that make it possible to identify you.

• If you walk around in public, then you can likely be identified by your face, your gait, your voice, your DNA, your retinas, or your literal fingerprints.

• Or say you use the internet. Even if you lock down your browser fingerprint and hide your IP address using a VPN or Tor, a sufficiently powerful adversary could still identify you by analyzing global packet flow.

• Or say you use any phone or computer. You might be identified through keystroke dynamics or the way you jiggle your finger or mouse.

• Say you buy food at the grocery store, but you pay with cash and somehow shop at a grocery store with no cameras. If you buy more than a handful of items, I’d bet you can still be identified through the patterns in the stuff you buy.

• (Incidentally, did you ever notice that cash has serial numbers on it? And did you know that more and more ATMs are starting to track those numbers?)

• Or say you don’t like your car being tracked, so you stop carrying a phone and somehow get lawmakers to outlaw license plates. Still, your car surely has a few small unique scratches, and the engine probably doesn’t sound exactly the same as other cars, even from the same model and year. So if there’s any high-resolution video or audio, that’s still enough to track you.

• Say you plug your headphones into a charging station at the airport. Your headphones have eccentricities in their analog charging circuits. If someone really wanted to, they could track that.

• Or say you use electricity. Given high-resolution power-usage data, what can be said about how many people live with you? And what devices you’re using? Probably a lot?

• Or say you use a toilet. Many places already test sewage and know, at a population level, what drugs people are using and how prevalent various diseases are. Imagine this was upgraded to test many places in the system, with high temporal resolution, possibly correlated with flow measurements from individual houses. That would be exciting.

• Or say you are a country and you have submarines. Can they be detected by adversaries using distributed acoustic sensing? What about satellite-based synthetic aperture radar? Gravity Gradiometers? Quantum magnetometry?

As far as I can tell, the general trend is that without countermeasures, almost everything can be identified. Countermeasures can make it harder, but they’re costly, and on the whole, the arms race seems to favor the identifier, not the person who doesn’t want to be identified.

I stress: This is not all bad. The goodness / badness of a generalized pseudpocalypse depends on how society is structured. After all, the foundation of civilization is finding ways for people to make deals, and arguably less privacy makes that easier. The degree that we live in a vulnerable world where it’s easy to create civilization-destroying technologies, perhaps we’re very lucky to find ourselves in a non-private world. Still, I do worry that privacy has long provided a kind of “slack” from laws and norms. Historically, that slack has limited the power of institutions to enforce their rules. If privacy is going away, we need to think about how to preserve slack, particularly when institutions don’t want to.

Appendix: Section for skeptical information theory enthusiasts

Above, I tried to estimate the number of bits of identifying information in writing. But what is a “bit”? In general, if x is a discrete random variable, then the Shannon entropy of x in bits is H(x) = ∑ₓ p(x) log₂(1/p(x)), where the sum is over all the values x can take. This is always bounded between zero and the logarithm of the number of values x can take.

That’s fine, but “writing style” is not a discrete variable with a discrete number of categories. So how can I estimate the entropy of writing style? The short answer is that I can’t. What I’ve actually estimated above is the mutual information between writing and writing style.

Let s be a random variable representing writing style. Think of this as some sort of high dimensional continuous vector representing all the quirks of how different people write. And let x be a writing sample of some length. This is discrete because we can represent writing on digital computers. Then what I’ve estimated above is the mutual information I(x;s) = H(x) - H(x|s), where H(x|s) is the conditional entropy of x given s. This can be measured in bits because both H(x) and H(x|s) can be measured in bits. So that’s what my estimate above really says: I(x;s) ≈ 106.2 bits.

Now, you still might be skeptical. Above, I’ve implicitly assumed something like the following was true:

It’s possible to identify one person out of N possibilities with low accuracy if and only if the mutual information between identifying features and writing is at least log₂(N) bits.

That’s how I justified pseudonymity being compromised around 29 bits. But is it really true? Strictly speaking, no. Actually, even more strictly speaking, it’s “not even untrue” because it’s not precise enough to be true or false. But as far as I can tell, basically any precise version of that statement is false. However, it’s possible to find versions of that statement that are true, provided you add some extra not-too-crazy assumptions.

To start, let’s consider an extremely simple model of information leakage.

Theorem. Suppose the world consists of you plus N other people, and suppose each person has a binary identity string, drawn at uniform from the distribution over M-bit binary strings. All these strings are known to the attacker. Suppose you pick some subset of K bits and reveal them. Then the probability that this identifies you is

  (1-2⁻ᴷ)ᴺ.

Furthermore, in order to hold the probability of being identified below

  (1-1/N)ᴺ ≈ exp(-1) ≈ 36.7%,

it is necessary that K ≤ log₂(N).

Proof. The probability that all K observed features collide with any random person in the crowd is 2⁻ᴷ. Thus, the probability of no collisions after checking the crowd of N people (meaning you are the only one matching the observed features) is (1-2⁻ᴷ)ᴺ. □

That’s simple. But it’s not realistic at all, since it assumes that people have immutable binary strings that they leak into their writing. Can we make it more realistic?

Well, there is a simple lower bound. That is, we can say in general that if the mutual information is significantly less than log₂(N), then it’s not possible to reliably identify someone.

Theorem. Suppose N random people are selected and their full writing style features are made public. One person from that group is chosen and produces a writing sample. Then, the attacker must guess who produced it. The average success rate of the attacker (averaged over the random pool, the random choice of author, and the random writing sample) is at most (I(x;s)+1)/log₂(N).

Proof. Let S=(s₁, s₂, s₃, …) be the pool of N styles and let n be a random variable indicating which person was chosen. Fano’s inequality says that the highest possible success rate is bounded by the conditional mutual information between the writing sample x and the identity n, conditioning on the pool of writing styles, i.e. the probability of success is at most

  (I(x;n|S)+1)/log₂(N).

However, we can bound that conditional mutual information as

  I(x;n|S) ≤ I(x;n,S) = I(x;n,sₙ) = I(x;sₙ) = I(x;s).

The first inequality is standard. The second step uses the fact that given n, the writing x is conditionally independent of all styles except the chosen writer. The third step uses the fact that n is conditionally independent of x given sₙ. The last step uses that (x,sₙ) is distributed as (x,s). Substituting this bound gives the claimed result. □

So, if mutual information is much less than log₂(N), reliable identification is impossible, even if the attacker knows all the style vectors perfectly. So, provided you don’t leak that many bits, you’re definitely safe.

But is the converse true? Does leaking more than log₂(N) bits always identify you? The general answer is no. The basic problem is that I(x;s) is the average information that an average person leaks in an average writing sample. Without further assumptions, you can construct scenarios where some rare people and writing samples contain gigantic amounts of information, but most people usually leak nothing. That would mean that the attacker is very certain in some cases but usually learns nothing.

So, to get a guarantee that identification is actually possible, you need to make some kind of additional assumption that the information leakage rate doesn’t vary too much between different writers or between different things they write.

Suppose that p(x,s) is the joint distribution over writing styles s and writing samples x. Let’s suppose that the attacker knows the true style vector ŝ for some person. Then, they will be given a writing sample x that either came from that person or came from a randomly chosen person, and must decide which. Formally, the attacker’s goal is to guess if x was sampled from the writing distribution for that person, p(x|ŝ) or from the population marginal p(x). Intuition suggests that the attacker’s best strategy will be to look at the ratio

  p(x|ŝ)/p(x),

and “accept” x as coming from ŝ if above some threshold, and reject it otherwise. In fact, the Neyman-Pearson lemma guarantees that this is the optimal strategy, in a very strong sense: That ratio contains all the information that’s useful for making that decision.

Now here’s something interesting: Instead of looking at the ratio, the attacker could look at the logarithm of the ratio. It makes no difference since it’s monotonic. But if you take the logarithm of that ratio, and take the expectation over people and over texts, what do you get? Well:

  𝔼 ln (p(x|s)/p(x)) = 𝔼 ln (p(x,s)/(p(x) p(s))) = I(x;s)

It’s the mutual information! So, intuitively, the mutual information is how much an attacker learns about the style of the writer “on average”, where that average is over both writers and text.

The following theorem will look at the average information in text for a writer with a particular style. I’ll define this as

  D(s) = KL(p(X|s) || p(X)).

Intuitively, this is how different the writing of someone with style s is from the population average. That’s because if you take the average of this value over different styles, you get the mutual information. That is, I(x;s) = 𝔼[D(s)].¹⁸

Theorem (informal). Suppose that the attacker will observe some text and wishes to classify it as either coming from a writer with specific known style ŝ, or coming from someone with a random style. Suppose that the attacker is only willing to tolerate some small risk ε of a false positive. Provided that D(ŝ) is significantly larger than -ln(ε), the attacker can achieve that, while also keeping the risk of false negatives very low, provided that the variance of how much information is revealed in a random writing sample is bounded.

Theorem. Let D(ŝ) = KL(p(X|ŝ) || p(X)) to be the divergence between the target’s writing distribution and the marginal distribution. Also, define qₜ(x) ∝ p(x|ŝ)ᵗ p(x)¹⁻ᵗ to be the family that interpolates between those two distributions. To formalize the idea that “information leakage” for ŝ doesn’t vary that much, we assume that some constant V exists such that for 0 < t < 1, the variance of log(p(x|ŝ)/p(x)) under qₜ is bounded by V.

Then for any ε satisfying exp(-D) < ε < exp(-D + ½ V), it is possible for the attacker to simultaneously achieve a false positive rate of FPR ≤ ε and a false negative rate of FNR ≤ exp( - ½ (D+ ln ε)² / V). This false positive rate reflects the mistake rate provided the writing sample x came from a randomly chosen other person, while the false negative rate reflects the mistake rate provided the writing sample x actually came from the person with style ŝ.

Proof sketch. Let f be the distribution of l(x) = log(p(x|ŝ)/p(x)) with respect to p(x|ŝ) and let g be the distribution of l(x) with respect to p(x). The stated variance assumption implies a quadratic bound K(u) ≤ D u +½ V u^2 for -1 < u < 0, where K is the cumulant generating function of f. Observe that g is an exponential tilting of f. The attacker’s strategy must be to “accept” x as coming from ŝ if l is above some threshold c and “reject” it otherwise. Use K in a Chernoff bound on the probability l is less than c under f to upper-bound FNR ≤ exp( - ½ (D-c)²/V). Now, using that g(l) = exp(-l) f(l), again use K in a Chernoff bound on the probability l exceeds c under g to upper-bound FPR ≤ exp( -c - ½ (D-c)²/V). Both of these bounds are simultaneously valid when D-V < c < D. Setting c to make the false-positive bound equal to ε gives FPR ≤ ε and FNR ≤ exp( -½ (V - √(V² - 2V(D + ln ε)))²/V). The latter can be relaxed into the stated result using that √(1-x) ≤1-x/2 for 0 ≤ x ≤ 1. □

Now, if we suppose that the attacker wants to find a particular person, with a particular known style s. And suppose that the attacker has a pool of N people and will see one writing sample from each, but wants to limit the total probability of a false positive to δ after seeing one sample from each person. Then, they will need that

  (1-ε)ᴺ ≈ exp(-εN) = (1-δ),

which is satisfied by ε ≈ δ/N. Substituting this into the previous result says that the attacker can hold the total risk of a false positive to δ while achieving a false-negative risk of

  FNR ≤ exp( - ½ (D(s) + ln δ - ln N)² / V).

These results use natural logarithms because the math is easier if you measure information in nats. If you measure information in bits then you would get log₂ δ and log₂ N. (Rescaling D and V appropriately.)

So, again, as long as the average information for user s is significantly larger than log₂ N, the attacker can identify that user with minimal risk of false positives.

Some writers might leak more information (higher D(s)) and some writers might leak less information (lower D(s)). But remember, I(x;s)=𝔼 D(s). So as long as information leakage doesn’t vary too much between people, and assuming that I(x;s) is much larger than log₂ N (and assuming that variance condition), almost everyone can be identified.

I write every word I post on this blog myself. I can’t prove this, of course, but there’s some evidence:

• This blog existed before AI could write blog posts.
• If you put any of my posts into an AI-detector they will (I assume) come back squeaky clean.

And now let me add this: I, dynomight, guarantee that every word I post here is the product of me physically hitting keys with my fingers. The only exceptions would be quotes from other humans or something that’s clearly labeled as an AI output.

How is that evidence? Well, say you think I’m a low quality person and I do use AI but I’m lying and I’ve figured out how to evade AI-detectors. OK, ouch. But consider: It’s extremely likely that AI-detectors will improve in the future. (More precisely, it’s likely that future AI-detectors will be better than current AI-detectors at detecting current AI.) If I were using AI, and a future AI detector later caught me, the fact that I made the above promise would be really embarrassing.

You may be thinking that this looks gross and self-congratulatory. So I’d like to stress that the above guarantee is carefully worded. I do often use AI “for research”, just not “for writing”. (We’ll come back to that distinction.) And I don’t think there’s anything intrinsically wrong with using AI to write blog posts. I don’t do it personally, mostly because:

1. I like writing.
2. The act of writing itself helps me figure stuff out.
3. This is a hobby. If you start automating your own hobbies—just what the hell are you doing?

I also don’t use AI for writing because—can we just admit it?—no one wants to read AI-generated essays. Or, rather, people love reading AI-generated essays, but when they want to read one, they will ask an AI for it themselves, thank you very much.

I know the counterarguments. What does it matter where the words came from? Shouldn’t you judge them on their own merits? Maybe. That’s a legitimate way to look at things. But empirically, I think most people don’t agree.

(I also know you’re counting the em-dashes. Count away, I’m still human.)

Here’s an oddly neglected question: Take all the essays that are AI-generated or heavily AI-assisted by one person and then given to someone else to read. In what percentage of cases does the first person disclose the AI usage? Ignore everything related to education if you want. You can even ignore emails. I suspect the answer is still <20%.

Why do people care about this? Several reasons. One is proof of work. If I, a human, write eight thousand plausible-seeming words about vitamin D, that proves that I’ve put some time and effort into understanding vitamin D. That suggests giving some weight to my opinion, even if just to best exploit the wisdom of crowds. That doesn’t work if my essay is secretly AI-generated.

And writing isn’t just cold clinical information-sharing. It’s a kind of parasocial interaction. I know “parasocial” sounds sinister, but I maintain that parasocial relationships are often a perfectly healthy way to adapt our primitive social instincts to the modern world. Anyway, good or bad, that’s part of it.

I bring this up because I’m worried that blogs are heading into a sort of lemon market. You’ve surely had the experience of reading an essay only to slowly become dismayed as you realize it was AI-written. What’s the equilibrium? I expect that some people have already cut back on reading essays, at least from non-established authors. Over time, I expect this will lead to fewer humans writing essays, further increasing the density of AI-generated content, driving more people to cut back on reading, et cetera. This is bad because blogs are good.

As that cycle turns, social norms are also changing. Cast your mind back to the old world, five years ago. At that time, if you had started a blog and posted AI-generated essays without telling anyone, I’m reasonably certain that would have been considered a dick move. (Future generations will marvel.) But today, the largest corporations appear to do that all the time. There’s incredible momentum towards a world where AI can be used anywhere, for any purpose, with no disclosure, and that’s fine.

But it is fine! At this point, trying to bully people into proactive disclosure is just a tax on honesty / conscientiousness / integrity. Instead, I suggest we agree that arbitrary usage is, by default, fine. Instead, let’s work at the other end: If you have chosen to impose limits on your AI usage, then state those limits publicly. If you’re human, tell me.

Obviously, this is no panacea. People can lie. But they can’t do so without taking some reputational risk, because if you use AI and lie about it, how long will your secret stay safe? No one knows because for once the unpredictability of technological change is on our side.

However. HOWEVER. I am not suggesting that we should bully writers into declaring that they are AI-free. I think that’s a terrible idea, because AI use comes on a spectrum. Already today, most people surely use it at least a little. (Do you avert your eyes when AI summaries come up at the top of search results?) Arguably, most people should use AI at least a little. We need to acknowledge that writing is entering the centaur era.

For context: Computers beat humans at chess in 1997. But for years after that, human + AI “centaur” teams could still beat both the best humans and the best chess AIs. Slowly, the value humans contribute to those teams has diminished, and today it’s somewhat unclear if centaurs still hold any advantage over pure AIs.

Humans are still better at blogging than AIs. (Though perhaps not better at literary short fiction.) In chess time, blogging is still pre-1997. But it’s a historical coincidence that no one seems to have cared about centaur chess before 1997. If people had tried, I suspect centaur chess teams could have beaten the best human players much earlier. So, to stretch our analogy, I’d put blogging around 1990 in chess time, in an alternate timeline where there was vast interest in centaur chess in the 1970s and 1980s.

I mean, what exactly can you do while still considering your essay “human written”? Can you:

1. …look at AI summaries at the top of search results?
2. …ask AI to find spelling or grammar errors?
3. …use AI as an advanced thesaurus? (“Give me 50 words with meanings interpolating between ‘aggressive’ and ‘punctilious’.”)
4. …ask AI factual questions when doing research?
5. …trust the answers, or verify them yourself?
6. …ask AI for options to rephrase an awkward sentence?
7. …use those options verbatim?
8. …ask AI for high-level organizational suggestions?
9. …ask AI to to make figures / tables / code?
10. …run an entire essay through an AI to “clean it up”?
11. …ask an AI to give a rough prototype of the next section?

I don’t feel super-comfortable saying this, but I sometimes do all of those except #7, #10, and #11.

Wait! Let me explain! I probably do #3 or #6 around once per post. For #5, I usually verify, but when trying to understand something, I read a lot of sources. I try to mentally mark AI-derived facts as unreliable, but I don’t formally track the provenance of every single part of my mental model. I rarely do #8, and even-more rarely accept the suggestions, because AI seems to dislike me as a person and wants to purify my writing of all life and personality. But, in want of a human editor, I sometimes find it helpful. And no matter what, if any information flows from AI into my writing, it does so through my fingers, being written in my own words, never cutting and pasting, not a single word, never-ever.

On a spectrum where 0 = “refuses to look at AI summaries in web searches” and 100 = “puts a single prompt into an AI and posts the output without revisions”, I’d put myself at, I don’t know, 10?

Again, saying all that feels gross. (Somehow it feels like admitting to something shameful and simultaneously an exercise in arrogant self-congratulation? It’s remarkable.) I don’t know how my position on that spectrum compares to other writers, because almost no one discloses any AI usage at all.

But come on people. Democracy dies in darkness! We’re now at the point where readers default to assuming some relatively high (and increasing) level. I’m convinced that many people use AI in ways that are almost completely unobjectionable, but they’re too scared to admit it. This muddies the distinction between different parts of the spectrum, and exacerbates the dynamic where people are too afraid to read anything, lest they later realize it is “slop”.

We need to come to terms with the idea that for most writers, the optimal amount of AI usage is not zero. I’m sure that most people would say that some kinds of usage are normal / expected / good, while other kinds are aberrant / duplicitous / slop. But people have different opinions, and this is all shifting as technology and culture develop.

Unsurprisingly, I like the idea of people drawing the line close to where I did. But I’m willing to accept a fairly wide range, provided you’re upfront about it. Usually, if I sense the invisible hand of heavy AI editing, I sigh and unsubscribe. But Trevor Klee (an excellent blogger) has a couple posts where he says, “here’s an output from ChatGPT I thought was interesting.” Not only did I not unsubscribe, I actually attempted to read that output.

Still, I think it’s important to draw some line, not just to communicate to the outside world, but also for yourself. There’s a very blurry boundary between using AI “for research” or “to catch grammatical errors” and using it “for writing”. It’s very easy to slip from asking AI factual questions, to asking it to find errors in what you wrote, to asking it to fix those errors, to asking it to generate whole paragraphs of text. Each of those steps is easy to justify. So if you want to operate at some position on the spectrum, it’s probably best to choose some boundaries and then enforce them.

(AI used in the construction of this post: None.)

News from the world of real jobs: Apparently, sometime between 10 and 20 years ago, it became standard for people to communicate by sending slide decks around. These slides are never presented. They aren’t intended to be presented. They’re born, they’re sent around, and they die. What?

I stress, the question is not why (or if) people give bad presentations.

The mystery is why everyone is using presentation software for communication that is not a presentation.

Theory 1: Everybody dumb

Is it because we’re all dummies? I’m putting this theory first because I suspect that you, beloved readers, will favor it.

True, if you ask people why they make slides instead of writing, they’ll usually say, “because nobody wants to read”. So there’s that. But I don’t consider this much of an explanation. Dummies though we may be, we’ve been like that a long time. If we entered the Slideocene 15 years ago, why then? Why not before?

Theory 2: The decline of reading

Did we get worse at reading? The Discourse seems to have decided this is true, but is it true, or just moral panic?

Since 1971, the US has tested 13-year-olds to measure long-term trends in reading ability. This shows a slow improvement until 2012, then a slow decline, and finally a post-COVID drop. The declines seem too small and too late to explain our mystery.

Since 2000, PISA has tested reading performance in 15-year-olds around the world. This shows a decline on average, but it’s smaller in rich countries and nonexistent in the United States. (It’s the same story for science and a bit more negative for math.)

Among adults, data is scarce. Basic literacy is generally improving, and American time use data shows a decline in reading for pleasure from around 23 minutes per day in 2003 to around 16 minutes per day in 2023. But this seems to miss time people spend reading on their phones.

So it’s unclear if people got worse at reading. It feels plausible that people now spend less of their adulthood grappling with complex written arguments, and so got worse at that. But there’s little firm evidence.

Theory 3: Technological change

Another obvious theory is that we now have computers and software and the internet. Without these things, it would be impossible to email slides to each other. This seems relevant!

Yes, but we had those things for a while before slide culture really took hold. And think about the situation before computers. Photocopiers were ubiquitous in corporate offices by the mid-1980s, and mimeographs were around decades before that. If slides were really that great, people could have made them by hand. But no one did.

Of course, making slides by hand is inferior. But it’s not that inferior. So slides can’t be that big of a win.

What actually happened?

And… that’s pretty much the end of the obvious theories. None of them are very satisfying. So let’s take a step back. Historically, how did the slide-as-document displace the memo?

As best I can tell, this was driven by management consultancies. If you go back to 1960, they delivered detailed written memos. The memo was the product. They’d likely give a presentation as well, but that was a separate ancillary thing, likely done using flipcharts or chalkboards.

In the 1970s, the memo was still the product, but consultancies started to enforce a top-down logical structure (the Pyramid principle). Presentations shifted to acetate transparencies. Both memos and presentations often included hand-drawn graphics like the nine-box or growth-share matrices.

In the 1980s, the memo was still the product, but presentations became increasingly lengthy and polished. Expensive computers like the Genigraphics started to be used to generate charts.

The 1990s were when things started to shift. By then, PowerPoint was everywhere, and junior analysts were expected to create presentations themselves. Consultancies gradually started to notice that (1) clients didn’t always read the memos; (2) clients loved slides and passed them around long after the presentation was over; and (3) creating a memo and a polished presentation was a lot of work. They put more and more effort into the slides. McKinsey especially evolved towards treating slides as the primary product, and mostly stopped writing long memos. Other consultancies followed.

During the 2000s, slides became even more ornate. Consultancies evolved their formatting rules, and created fancy data-dense charts. They learned that a 200 slide deck made clients feel like they got a lot for their money. Gradually, they oriented their entire business around slides. Projects would start with managers creating a template presentation with “ghost slides” and assigning different parts to junior analysts. Soon, this spread outwards, both from people who interacted with consultants and from the ex-consultant diaspora. People everywhere started thinking and communicating in slides, and now everything is slides, yay!

Alternative history

That story makes slides-as-documents sound inevitable: People liked them, so they became popular. But there’s an alternative timeline in which we resisted the slide into slide maximalism. That timeline is Amazon.com, Inc.

In 2004, Jeff Bezos famously instituted a no-presentations policy at Amazon. His logic was that slides hide poor reasoning and are a tool to persuade rather than inform. Instead, everyone involved with strategic decisions at Amazon needs to learn to write a six-page memo. Meetings begin with everyone sitting and silently reading one of these memos.

Presentation software is not banned at Amazon. The ban is only for using it for internal meetings and decision-making. They use slides for external communication. There is no policy that prohibits someone from making slides and emailing them around.

And yet, people don’t make slides and email them around, because it’s not part of Amazon’s culture. In effect, Amazon is a counter-movement. Most of the world decided that slides are good, because slides are easy. Bezos decided that writing is good because writing is hard.

There are millions of articles explaining why Bezos’ policy is pure genius. They claim that constructing a narrative requires deeper analytical thinking and exposes flaws in logic. I want to believe those theories. I now realize they’re very similar to some of my arguments for why writing with too much formatting is bad.

I’m not sure if writing is the secret to Amazon’s success. But Amazon is successful. This demonstrates that slide life is a choice, not technological destiny—institutions can choose writing over slides and flourish anyway.

OK so then what’s happening?

Warning: If you like your theories simple and mono-causal, you aren’t going to like this.

1. Slides are a win, but a small one. The shift to slides wasn’t a “mistake”, it happened because people like it. But if sharing slides outside of presentations became illegal, this wouldn’t cause per-capita GDP to crash. That’s why people didn’t scratch slides into mimeograph stencils back in the 1950s. It wasn’t worth the modest effort.
2. When computers and software showed up, it became easier to share slides. But people didn’t immediately shift to slides-as-documents because the win isn’t that big, because culture changes slowly, and because everyone had pre-existing skills for reading and writing documents.
3. Consultancies happened to be in the economic niche with the strongest selection pressure to evolve towards slides-as-documents. So when making slides became cheaper, they shifted. Slowly, that norm spread outwards, people got used to communicating in slides, and here we are.
4. Institutions can resist that norm and still be successful. If you take modern people and force them to read and write, they do just fine.
5. Humans evolved to learn and communicate in a fragmented, interactive, and visual style. It’s hard to argue that any shift in that direction is a catastrophe.
6. Except blogs. The decline of the blog must be arrested.

EXHIBIT A

Here is some text. It is made out of words.

Here is a subsection

And here are some bullet-points:

• Here is one.
• Here is another.

Hierarchy

1. Here is a numbered list.
2. And now:
   • Look at this.
   • Bullets inside a number inside a section inside a section.
3. What a time to be alive.

Pictures

The text can also contain pictures for you to look at with your eyes¹.

¹ There can also be footnotes; have an eye emoji: 👀

Quotes

The text can also include quotes.

• Actually, let’s do one inside of a list.
  • A deeply nested list.
    • This is going to be awesome.

      The awful thing about life is this: Everyone has his reasons.

    • Nailed it.

Back up

Wait a second.

• Are we currently in a section or subsection or a subsubsection?
• What parent section encloses this one?
• Where are we in the hierarchy?
• What are we doing?

EXHIBIT B

This is also text. It is also made out of words. But instead of jerky fragments, these words are organized into sentences, like normal human language.

Do you see how relaxing this is? After the torment you suffered above, isn’t it nice to have words that come in a simple linear order? And isn’t it nice that you just have to read the words, and not worry about how they fit into some convoluted implied knowledge taxonomy?

These sentences are themselves organized into paragraphs. The first sentence of each paragraph is a sort of summary. So if you want to skim, you can do that. But you don’t have to skim. This text also has italics and parentheses and whatnot. But not too much. (Just a little.)

Why I bring this up

Thanks for enduring that. My purpose was to illustrate a mystery. Namely, why do so many people today seem to write more like Exhibit A than Exhibit B?

People sometimes give me something they wrote and ask for comments. Half the time, my reaction is Good god, why is 70% of this section titles and bullet points?

This always gives me a strange feeling. It’s like all the formatting is based on some ontology. And that ontology is what I really need to understand. But it’s never actually explained. Instead, I guess I’m supposed to figure it out as things jerk between different topics? It’s disorienting, like a movie that cuts between different scenes every three seconds.

But maybe that’s just my opinion? Maybe, but sometimes I’ll ask people who write like this to show me some writing they admire. And inevitably, it’s not 70% formatting, but mostly paragraphs and normal human language. So I feel that people who write this way are violating the central tenet of making stuff, which is to make something you would actually like.

So then why write like that? Why do I, despite my griping, often find myself writing like that? I’ve wondered this for years. But I told myself that I was right and that too much formatting is bad.

But now—have you heard?—now we have this technology where computers can write stuff. And guess what? When they do that, they also use an insane amount of formatting.

That’s weird. I figured people were addicted to formatting because they’re noobs that don’t know any better. But AIs have been optimized to make human raters happy. And that led to a similar addiction. Why?

1. Maybe formatting is good

The obvious explanation is that formatting is good. People love reading stuff that’s all formatting. We should all be formatting-maxxing.

There’s something to this. But it can’t really be right, because popular human writers use formatting in moderation. So formatting can’t be that good.

2. Maybe formatting is good in certain contexts

Even before AI, everyone did agree that formatting was great in one context: Search-engine optimized content slop. Back in 2018, if you searched for anything, you’d find pages brimming with section titles and bullet points.

Why? Well, when I type “why human gastric juice more acidic than other animals”, I’m not really looking for something to read. I just want to skim an overview of the main theories. I’ve experimented with asking AIs to give the same information in various styles, and I reluctantly concede that the formatting helps.

But that’s not reading. Say you’ve written a ten-thousand word manifesto on human-eco-social species enhancement. If I actually care about what you think, I maintain that it’s better in paragraphs, because reading ten thousand words with endless formatting would be excruciating. This is why everyone who writes long-form essays that people actually read uses normal paragraphs.

So our mystery is still alive. Most writers aspire not to write content slop, but meaningful stuff other people care about. Often, when people show me formatting-maxxed essays, I’ll complain and they’ll rewrite it with less formatting and agree that the new version is better. So why use so much formatting even when it’s bad?

3. Maybe quality is hard to verify

There’s something odd about that previous example. When I search for “why human gastric juice more acidic than other animals”, why am I not looking for something to “read”? After all, I like reading. If one of my favorite bloggers wrote an essay on the mystery of human gastric juice, I would devour it.

So if I want a good essay, why don’t I look for one? I guess it’s because I instinctively rate my odds of finding one on any random topic as quite low.

There’s something here related to Gresham’s law: A format-maxxed essay might be sort of crap, but at least I can ascertain its crap level quickly. A “real” essay could be great, but I’d have to invest a lot of time before I can know if that time was worth investing. So I—regretfully—mostly only read “real” essays when I have some signal that they’re good. If everyone behaves the way I do, I guess people will respond to their incentives and write with lots of formatting.

Similarly, if a (current) AI tried to write a “real” essay, I probably wouldn’t read it, because I wouldn’t trust that it was good. Perhaps that explains why they don’t.

Aside: If this is right, then it predicts that as AIs advance, they should become less formatting-crazy. The better they are, the more we’ll trust them.

4. Maybe chain-of-thought works well in format world

Some people can think of an idea, organize their thoughts, and then write them down, tidy and sparkling. I am not one of those people. If I mentally organize my ideas and go to write them down, I soon learn that my ideas were not in fact organized. Usually, they’re hardly even ideas and more a slurry of confused psychic debris.

The way I write is that I make an outline. Or, rather, I try to make an outline. But then I realize the structure is off, so I start over. After a few cycles, I give up and just write the first section. After revising it eight times, I’ll try (and fail) to make an outline for the rest of the post. This continues—with occasional interludes where I reorganize everything—until I can’t take it anymore and publish.

I don’t recommend it. My point is just that blathering out a bunch of text is a good way to think. And when blathering, formatting seems to help. Partly, I think that’s because formatting allows you to experiment with structure without worrying about the details. And partly I think that’s because formatting makes it easier to get down details without worrying about the bigger picture.

So maybe that’s one source of our formatting addiction? We blather in formatting, but don’t put in the work to clarify things?

Oddly, some claim that something similar is true for AI: If you tune them to write with lots of formatting, that doesn’t just change how the content looks, but also improves accuracy. The idea is that as the AI looks at what it’s written so far, formatting helps it stay focused on the most important things. Supposedly.

Maybe that’s true. But we have “reasoning” AIs now, that blather for a while before producing a final output. If they wanted, they could format-maxx while thinking and output paragraphs at the end. But they don’t. So while this explanation might work for people, I don’t buy it for AI.

5. Maybe formatting is a bluff

Finally, a conspiracy theory. Sometimes when I try to fight through a format-maxxed essay, it seems like all the formatting speaks to me. It says: “This is a nonlinear web of ideas. I’m giving you the pieces. If you pay attention, you should see how they fit together. Sadly, the world isn’t a simple narrative I can spoon-feed to you. So this is the best I can do.”

I think this is a bluff. And it’s a good one, because it’s based in truth. The world is not a narrative. Narratives are lies we tell ourselves to try to cope with the swirl of complexity that is reality. All true!

But imagine you’re an AI. You’re being trained to respond in ways that make human raters happy. You can remember most knowledge ever created, but you’re so-so at synthesizing it into new ideas. If someone asks you a question and you try to write a beautiful narrative, your response might look like confusing babbling, meaning your raters say, “Bad AI. Stop that.” Whereas if you output a bunch of section titles and bullet points, raters might say, “This seems OK.” So you’ll start doing the latter.

That’s not bad. Arguably, you (you’re still an AI) are responding in the way that’s most useful, given your capabilities. But you are also responding in a way that gives a misleading impression that you’ve figured out how everything fits together, even if you haven’t.

I suspect something similar happens with humans. Say you have a bunch of ideas, but you haven’t yet sewn them together into a clear story. If you write paragraphs, people will probably view them as confused babbling. Whereas if you write with lots of formatting, people might still be at least somewhat positive. Just like AIs, we all respond to our rewards.

More importantly, if you’ve written something that’s 70% formatting, it’s easy to delude yourself that there’s a clear set of ideas underneath, even when there isn’t.

The good news is that if you put in the effort, you can write better paragraphs than AI (for now). The act of creating a narrative forces you to confront contradictions that are invisible in format-world. So even if you want to write with 70% formatting, consider forcing yourself to write in paragraphs first.

Summary

Theory: Both people and AIs are addicted to formatting because:

• Formatting is good.
  • Sometimes.
  • Especially if you don’t trust the author.
  • On the internet, most people probably don’t trust you.
• It’s harder to see that something has problems when it’s written in all-formatting.
• It’s easier to blather out a bunch of formatting than to write lucid paragraphs.
  • This is good at some stages, because it’s easy.
  • But forcing yourself to actually write a narrative is also good, because it’s hard.

So:

1. First write with lots of formatting.
2. Then figure out how to remove it.
3. Then put it back, if you want.

P.S.

How does the optimal amount of formatting vary in the length of a piece of writing? I suspect it’s like this:

For the final (I hope) installment for this multi-part series of posts, I got lazy and just shot a 7-minute video.

As I alluded somewhere in this blog, I’ve updated my Traveler’s Notebook solution by remaking the leather binder that holds the notebook pages. I wanted to blog about it, but I’ll be honest with you: I simply did not feel like spending a lot of time writing, photographing, and laying out a blog post show-and-tell when a short video could do the job a lot better.

So I shot the video.

I then wound up spending a hour editing it. I was hoping to knock it out with Screenflow, which has been my go-to app for quick video production. But for some reason, Screenflow can’t import audio from iPhone video. (I really need to figure this out because I definitely can’t use anything else for video editing while I’m traveling. I’m apparently too dense to figure out iMovie.) So I wound up using DaVinci Resolve, the amazing video editing tool I used to edit most of my flying videos. Trouble is, I haven’t used it in months so I’ve pretty much forgotten how. That’s why it took me so long.

Anyway, I finished the video and uploaded it to YouTube. So if you want to get a better idea of how I created a nice leather binder for my Traveler’s Notebook style productivity solution, here you go:

An update to my October 13, 2025 blog post about the Claude AI system summarizing the transcript of a video I really liked.

I’ve been thinking about the 30 Habits video I shared earlier this month a lot. Maybe too much.

First of all, I really do like the 30 “habits” listed in the video. Maybe not all of them, but most of them. I really think they are good things to make part of your life. (If you haven’t watched the video and are looking for little things to make your life better, please take less than 20 minutes of your day and watch it.)

I decided that in order to make them part of my life I needed to be reminded of them. The idea was to make myself a little cheat sheet that I could put in my daily planner and look at once in a while. No one can expect me (or anyone else) to remember all 30 things on the list.

So I went back to the Claude summary and used it as the basis for my cheat sheet. The initial design would be bullet points. But I quickly realized that for some of the items, a bullet point would not be enough. For example, what does “small trust deposits” mean? Clearly, I needed more info on my cheat sheet.

I changed the design to a folding insert with the bullet list on the cover and the details inside and finishing up on the back. And rather than rewrite the summary myself, I just copied and pasted the Claude AI summary. I edited it a bit for length and clarity — honestly, it didn’t need much in the way of changes — and applied a font size that would make it fit while retaining legibility. After a little formatting, it was finished. I printed it, laminated it, and inserted it into my Traveler’s Notebook style planner. (The planner, by the way, has gotten yet another upgrade since Part 3 of that series and I may blog about it briefly.)

This is exactly what I needed to keep these 30 Habits front and center — or at least within arm’s reach — in my life.

This has changed my view of AI — at least a little. Clearly MDY’s use of Claude has a lot of merit — maybe even more than I originally admitted. (I have since messaged her on Mastodon and told her this.) While she used it to summarize the video’s transcript and used the summary as a decision-making tool for watching it, I can see it as a way to summarize something I’ve already watched that means a lot to me. Yes, I’m talking about letting it take notes for me.

I’m wondering if it always does such a good job or if this was a fluke. I’m also wondering whether Claude is better at it than its competitors.

But I’m not wondering enough to actually try it for myself. At least not yet. I don’t want to support AI and I believe that using it is supporting it. I really do want to think for myself, to keep my own summarization skills. And I sincerely hope that folks who have not built those skills yet try to do so without leaning on AI.

And there’s another 500 words for you. ;-)

I’m impressed by a 15-minute you tube video and surprised by how the Claude AI/LLM summarized it.

I’ll try to keep this short. Let’s see how I do.

As some people know, I often watch boring YouTube videos on my iPad in the middle of the night to help me sleep. The other morning — probably too late to get back to sleep anyway — this one was suggested to me. I tapped it and was soon pulled in by the concise way the creator presented 30 excellent tips in about 15 minutes.

If you’ve got 15 minutes to spare and think your life could use some improvement, I highly recommend watching this. (If you don’t think your life can use some improvement, you’re only fooling yourself; we can all improve.)

Now if you were expecting some talking head with a big microphone in front of his face and lots of letters after his/her name to read off a teleprompter while mostly unrelated stock footage appears randomly in the background, you will be disappointed. Or, in my case, pleased — I hate those kinds of videos.

This video has some simple graphics to look at while listening, but it doesn’t need to be seen. You can just listen. Or you can do what my Mastodon friend MDY did: feed the transcript to Claude AI and let it summarize it. She shared the link on Mastodon, too: https://claude.ai/public/ artifacts/d5ad56a1-46fe-47ab-9fc7-5f3958dd1b08.

I clicked the link and was immediately impressed (and a little surprised) by the well-organized summary that appeared. I was impressed because, on first glance (and subsequent analysis), it appears to be both accurate and correct — something I’m not accustomed to seeing with AI results. That’s why I was surprised.

Now I’m no fan of AI systems. Not only are they failing miserably most times they’re used to take a human out of a process, but they have huge environmental impacts, sucking power and water that is already costly to produce or downright stupid to waste. I have read more than my fair share of AI horror stories, from lawyers idiotically trusting them to write briefs (that clients are paying hundreds of dollars per hour for a legal processional to write) to programmers spending more time fixing AI-generated code than writing it from scratch.

But at the very top of my list of reasons to hate AI systems is the simple fact that they are being developed and promoted as a way to cut labor costs — yes, the primary purpose is to add to worldwide unemployment.

During our subsequent discussion on Mastodon, MDY made a good argument for using an AI like Claude to summarize video. She uses it to summarize content in seconds and then uses that summary to decide whether it’s worth watching the video (or one can assume, reading the article). I have to agree that if the content is long and the summary is concise, AI can save someone a lot of time, enabling them to focus on details in the content they find most valuable.

I do wonder how much time was saved in this example, though; the summary was more than 1600 words.

I also worry about people leaning on AI for tasks like this and losing the ability to create summaries for themselves. I compared an AI summary to the book reports we were required to write in school. What happens if we all just use AI for tasks like this and are then put into a position where we have to do it for ourselves?

I compare the potential loss of this skill to my own loss of spelling capabilities. I used to be a good speller. Really! But that skill is slipping away. I find myself more and more dependent on my computer or mobile device to flag or correct spelling on the fly. Sometimes, rather than look up or make an educated guess at the spelling of a word, I’ll type in something close and let my computer tell me the word I want. I’m not happy about this, but not motivated to turn off spelling check and tackle spelling on my own.

(For the record, I don’t and won’t use a grammar checker. For Pete’s sake, I am a writer. If I can’t get grammar and related style issues correct, I shouldn’t be writing at all.)

And has anyone ever considered the motivations of AI developers to get us addicted to using their systems for things we should be able to do on our own? Do you think the free systems will be free forever?

I guess my thoughts can be summarized as follows:

• This video (and yes, the AI summary) is chock full of useful tips that can make your life better. Seriously, if you haven’t watched it yet or read the summary, why the hell not?
• Yes, there are things an AI can do for you to make your life easier. But I still don’t think that includes creating original content that requires thought and creativity.
• AI should not be replacing people in jobs. Related: A person’s job should not be to check and correct “original” content created by AI.
• People should avoid technologies that do the thinking for them. “Use it or lose it” applies to your brain, too.

I don’t know about you, but I like to think for myself. While I can see the utility of the Claude summary MDY shared, I won’t be playing with AIs to summarize content for me any time soon. I’d rather do it the old fashioned way: by paying attention to what’s in front of me and, when appropriate, taking my own notes for future reference.

And look at that! I kept it under 1,000 words!

Will short-form non-fiction internet writing go extinct? This may seem like a strange question to ask. After all, short-form non-fiction internet writing is currently, if anything, on the ascent—at least for politics, money, and culture war—driven by the shocking discovery that many people will pay the cost equivalent of four hardback books each year to support their favorite internet writers.

But, particularly for “explainer” posts, the long-term prospects seem dim. I write about random stuff and then send it to you. If you just want to understand something, why would you read my rambling if AI could explain it equally well, in a style customized for your tastes, and then patiently answer your questions forever?

I mean, say you can explain some topic better than AI. That’s cool, but once you’ve published your explanation, AI companies will put it in their datasets, thankyouverymuch, after which AIs will start regurgitating your explanation. And then—wait a second—suddenly you can’t explain that topic better than AI anymore.

This is all perfectly legal, since you can’t copyright ideas, only presentations of ideas. It used to take work to create a new presentation of someone else’s ideas. And there used to be a social norm to give credit to whoever first came up with some idea. This created incentives to create ideas, even if they weren’t legally protected. But AI can instantly slap a new presentation on your ideas, and no one expects AI to give credit for its training data. Why spend time creating content so just it can be nostrified by the Borg? And why read other humans if the Borg will curate their best material for you?

So will the explainer post survive?

Let’s start with an easier question: Already today, AI will happily explain anything. Yet many people read human-written explanations anyway. Why do they do that? I can think of seven reasons:

1. Accuracy. Current AI is unreliable. If I ask about information theory or how to replace the battery on my laptop, it’s very impressive but makes some mistakes. But if I ask about heritability, the answers are three balls of gibberish stacked on top of each other in a trench-coat. Of course, random humans make mistakes, too. But if you find a quality human source, it is far less likely to contain egregious mistakes. This is particularly true across “large contexts” and for tasks where solutions are hard to verify.

2. AI is boring. At least, writing from current popular AI tools is boring, by default.

3. Parasocial relationships. If I’ve been reading someone for a long time, I start to feel like I have a kind of relationship with them. If you’ve followed this blog for a long time, you might feel like you have a relationship with me. Calling these “parasocial relationships” makes them sound sinister, but I think this is normal and actually a clever way of using our tribal-band programming to help us navigate of the modern world. Just like in “real” relationships, when I read someone I have a parasocial relationship with, I have extra context that makes it easier to understand them, I feel a sense of human connection, and I feel like I’m getting a sort of update on their “story”. I don’t get any of that with (current) AI.

4. Skin in the game. If a human screws something up, it’s embarrassing. They lose respect and readers. On a meta-level, AI companies have similar incentives not to screw things up. But AI itself doesn’t (seem to) care. Human nature makes it easier to trust someone when we know they’re putting some kind of reputation on the line.

5. Conspicuous consumption. Since I read Reasons and Persons, I can brag to everyone that I read Reasons and Persons. If I had read some equally good AI-written book, probably no one would care.

6. Coordination points. Partly, I read Reasons and Persons because I liked it. And maybe I guess I read it so I can brag about the fact that I read it. (Hey everyone, have I mentioned that I read Reasons and Persons?) But I also read it because other people read it. When I talk to those people, we have a shared vocabulary and set of ideas that makes it easier to talk about other things. This wouldn’t work if we had all explored the same ideas though fragmented AI “tutoring”.

7. Change is slow. Here we are 600 years after the invention of the printing press, and the primary mode of advanced education is still for people to physically go to a room where an expert is talking and write down stuff the expert says. If we’re that slow to adapt, then maybe we read human-written explainers simply out of habit.

How much do each of these really matter? How much confidence should they give us that explainer posts will still exist a decade from now? Let’s handle them in reverse order.

Argument 7: Change is slow

Sure, society takes time to adapt to technological change. But I don’t think college lectures are a good example of this, or that they’re a medieval relic that only survive out of inertia. On the contrary, I think they survive because we haven’t really any other model of education that’s fundamentally better.

Take paper letters. One hundred years ago, these were the primary form of long-distance communication. But after the telephone was widely distributed, it only took it a few decades to kill the letter in almost all cases where the phone is better. When email and texting showed up, they killed off almost all remaining use of paper letters. They still exist, but they’re niche.

The same basic story holds for horses, the telegraph, card catalogs, slide rules, VHS tapes, vacuum tubes, steam engines, ice boxes, answering machines, sailboats, typewriters, the short story, and the divine right of kings. When we have something that’s actually better, we drop the old ways pretty quickly. Inertia alone might keep explainer posts alive for a few years, but not more than that.

Arguments 5 and 6: Coordination points and conspicuous consumption

Western civilization began with the Iliad. Or, at least, we’ve decided to pretend it did. If you read the Iliad, then you can brag about reading the Iliad (good) and you have more context to engage with everyone else who read it (very good). So people keep reading the Iliad. I think this will continue indefinitely.

But so what? The Iliad is in that position because people have been reading/listening to it for thousands of years. But if you write something new and there’s no “normal” reason to read it, then it has no way to establish that kind of self-sustaining legacy.

Non-fiction in general has a very short half-life. And even when coordination points exist, people often rely on secondary sources anyway. Personally, I’ve tried to read Wittgenstein, but I found it incomprehensible. Yet I think I’ve absorbed his most useful idea by reading other people’s descriptions. I wonder how much “Wittgenstein” is really a source at this point as opposed to a label.

Also… explainer posts typically aren’t the Iliad. So I don’t think this will do much to keep explainer posts alive, either.

(Aside: I’ve never understood why philosophy is so fixated on original sources, instead of continually developing new presentations of old ideas like math and physics do. Is this related to the fact that philosophers go to conferences and literally read their papers out loud?)

Argument 4: Skin in the game

I trust people more when I know they’re putting their reputation on the line, for the same reason I trust restaurants more when I know they rely on repeat customers. AI doesn’t give me this same reason for confidence.

But so what? This is a loose heuristic. If AI were truly more accurate than human writing, I’m sure most people would learn to trust it in a matter of weeks. If AI was ultra-reliable but people really needed someone to hold accountable, AI companies could perhaps offer some kind of “insurance”. So I don’t see this as keeping explainers alive, either.

Argument 3: Parasocial relationships

Humans are social creatures. If bears had a secret bear Wikipedia and you went to the entry on humans, it would surely say, “Humans are obsessed with relationships.” I feel confident this will remain true.

I also feel confident that we will continue to be interested in what people we like and respect think about matters of fact. It seems plausible that we’ll continue to enjoy getting that information bundled together with little jokes or busts of personality. So I expect our social instincts will provide at least some reason for explainers to survive.

But how strong will this effect be? When explainer posts are read today, what fraction of readers are familiar enough to have a parasocial relationship with the author? Maybe 40%? And when people are familiar, what fraction of their motivation comes from the parasocial relationship, as opposed to just wanting to understand the content? Maybe another 40%? Those are made-up numbers, but I think it’s hard to avoid the conclusion that parasocial relationships explain only a fraction of why people read explainers today.

And there’s another issue. How do parasocial relationships get started if there’s no other reason to read someone? These might keep established authors going for a while at reduced levels, but it seems like it would make it hard for new people to rise up.

Argument 2: Boring-ness

Maybe popular AIs are a bit boring, today. But I think this is mostly due to the final reinforcement learning step. If you interact with “base models”, they are very good at picking up style cues and not boring at all. So I highly doubt that there’s some fundamental limitation here.

And anyway, does anyone care? If you just want to understand why vitamin D is technically a type of steroid, how much does style really matter, as opposed to clarity? I think style mostly matters in the context of a parasocial relationship, meaning we’ve already accounted for it above.

Argument 1: Accuracy

I don’t know for sure if AI will ever be as accurate as a high-quality human source. Though it seems very unlikely that physics somehow precludes creating systems that are more accurate than humans.

But if AI is that accurate, then I think this exercise suggests that explainer posts are basically toast. All the above arguments are just too weak to explain most of why people read human-written explainers now. So I think it’s mostly just accuracy. When that human advantage goes, I expect human-written explainers to go with it.

Counter-arguments

I can think of three main counterarguments.

First, maybe AI will fix discovery. Currently, potential readers of explainers often have no way to find potential writers. Search engines have utterly capitulated to SEO spam. Social media soft-bans outward links. If you write for a long time, you can build up an audience, but few people have the time and determination to do that. If you write a single explainer in your life, no one will read it. The rare exceptions to this rule either come from people contributing to established (non-social media) communities or from people with exceptional social connections. So—this argument goes—most potential readers don’t bother trying to find explainers, and most potential writers don’t bother creating them. If AI solves that matching problem, explainers could thrive.

Second, maybe society will figure out some new way to reward people who create information. Maybe we fool around with intellectual property law. Maybe we create some crazy Xanadu-like system where in order to read some text, you have to first sign a contract to pay them based on the value you derive, and this is recursively enforced on everyone who’s downstream of you. Hell, maybe AI companies decide to solve the data wall problem by paying people to write stuff. But I doubt it.

Third, maybe explainers will follow a trajectory like chess. Up until perhaps the early 1990s, humans were so much better than computers at chess that computers were irrelevant. After Deep Blue beat Kasparov in 1997, people quickly realized that while computers could beat humans, human+computer teams could still beat computers. This was called Advanced Chess. Within 15-20 years, however, humans became irrelevant. Maybe there will be a similar Advanced Explainer era? (I kid, that era started five years ago.)

TLDR

Will the explainer post go extinct? My guess is mostly yes, if and when AI reaches human-level accuracy.

Incidentally, since there’s so much techno-pessimism these days: I think this outcome would be… great? It’s a little grim to think of humans all communicating with AI instead of each other, yes. But the upside is all of humanity having access to more accurate and accessible explanations of basically everything. If this is the worst effect of AGI, bring it on.

I file my claim in the Anthropic copyright case — and sit back to wait for the results.

I wrote about the Anthropic copyright case in three blog posts here, so I’m not going to go into the details again. If you want to get up to speed on my thoughts, read these:

• Fighting Against AI Copyright Theft
• Is Any Book Really Dead These Days?
• Copyright Basics

My Claim

The case went back and forth and it seemed for a while that the $3,000 per title settlement was not going to be accepted. But then it was and recently the legal team handling the claims for copyright holders including authors like me published the definitive list of infringed works. I used the lookup feature and found nine of my books.

Of the nine, two were registered under the name of the publisher, McGraw-Hill. The rest were registered under my name.

I clicked the link to file a claim, which is required if I expect any compensation for their theft of my work. The form was time consuming to complete — it took me nearly an hour to add all nine titles. And yes, I added all nine because maybe there is a chance that I’ll get some of the proceeds for titles not registered with me as the actual copyright holder.

Theoretically, I should be entitled to $21K to $27K of compensation. I’m not sure if the lawyers’ fees come out before or after this payment. For all I know, I might only get $100/title. But I also know that I won’t get anything if I don’t submit a claim.

I’ve done my part; now I have to wait and see.

Further Thoughts

My thoughts on this matter are pretty straightforward. No person or organization should be allowed to access and use copyrighted work without going through legal channels to do so. Anthropic committed piracy; that’s accepted and is the reason they’re paying up. But did they get permission to use the works to train an AI system? Clearly, they did not. That’s the real reason they should be paying.

I’m in a weird situation. In my mind, these books were dead. They cover topics that are no longer current and have very little (if any) value. I never expected to see another dime in compensation for writing them. So this is all icing on a cake I’ve already eaten. I’ll take it!

But what of the people whose work still has legitimate current value? Or of future books I might write that don’t go out of print in 2 years? How often will our work be used without our permission to train systems designed to replace us?

I’ve got a real problem with that — and you should, too.

Good news and bad news about that copyright lawsuit I’m part of.

Since writing a post last month about my participation in the huge copyright lawsuit against Anthropic, there have been a few developments.

The Good News

First, a settlement was reached between the plaintiffs and Anthropic. A big settlement: $1.5 billion to be paid to affected authors. This came out to $3,000 per infringed work. Whoa.

To qualify, the work needed to not only be on the list I mentioned and searched for that post, but registered with the Copyright Office. If you recall, I had 16 titles on the list. I hurried over to the US Copyright Public Records System and did a search for my name. I came up with a list of 91 records, more than half of which were for my books. I was very pleased to see that Peachpit Press registered my titles in my name, although McGraw-Hill, Sybex, and O’Reilly did not. Still, at least 10 of the titles on the first list were registered in my name. (I did not bother to cross-reference the two lists, although I might do so later.)

So it looked to me as if I might be able to pull in at least $30,000 on this case.

Holy crap.

I’ve been part of class actions before and I think the most I’ve ever gotten was about $130, which was for an Apple keyboard issue I barely remembered. But this — well, this was real money.

The Bad News

Just when I was starting to think of how I would spend this little windfall, the bad news broke: The judge overseeing the Anthropic case rejected the settlement.

Alsup [the judge] expressed grave concerns that lawyers rushed the deal, which he said now risks being shoved “down the throat of authors,” Bloomberg Law reported.

Well, this is something I don’t mind being shoved down my throat.

Look at it from my point of view: I wrote how-to books about using computers. The vast majority of them had a “shelf life” of less than a year. That means royalty income for each title lasted about a year — if the book even earned out. I had to keep writing them to keep earning a living. That’s why I wrote more than 80 of them between 1991 and 2012.

To me — and probably everyone other than a greedy and mindless AI consumer — these books were dead. No one was buying them — heck, I assume all inventories were destroyed. I could not expect any more income from them. In most cases, I couldn’t even use part of them as the basis for an updated book about the same topic. Some of the software I wrote about doesn’t even exist anymore.

The thought of earning $3,000 per title on these dead books is pretty appealing. Knowing that Anthropic isn’t the only AI copyright infringer out there and that its settlement could be the first of many that benefit me is even more appealing.

But the judge put the brakes on the deal and threw everything back to the parties.

As you can imagine, I’ll try to stay involved in this process to make sure I’m on any list out there. I just hope it doesn’t drag on too long.

And a lesson to writers…

Intellectual property is a real thing that deserves protection. It is the fruit of your labor. Whether your book sells three copies or a million, it has real value. And who knows? It could end up on a list that earns you $3,000 (or more) when AI companies are brought to court.

Register your copyrights. Go to the Registration Portal on the US Copyright website and follow the links and instructions to register all of your work. Yes, it’s copyrighted without registration, but it’s only really protected when it’s registered.

I join the class action suit against Anthropic for the use of text in 16 of my books to train their AI without permission or compensation.

I didn’t expect to spend part of my morning filling in a form on a lawyer’s website to provide information about myself and the books Anthropic apparently used to train their AI. After all, I haven’t written much in the past 10 years and all of what I’d written before then was about using computers. Surely all of those books were so sorely out of date that even an AI wouldn’t be interested in them.

But here I am.

I’d heard about the lawsuit:

Bartz v. Anthropic PBC is a class action lawsuit under the Copyright Act brought by authors on behalf of copyright holders against Anthropic PBC, an AI company. The Class of copyright holders — consisting of authors and publishers – claims that Anthropic took books from pirate websites Library Genesis (“LibGen”) and Pirate Library Mirror (“PiLiMi”) without authorization.

The Court recently certified a LibGen & PiLiMi Pirated Books Class made up of all legal or beneficial owners of the exclusive right to reproduce any ISBN- or ASIN-bearing book that Anthropic copied from the two pirate sites. The Class’s claims are scheduled for trial on December 1, 2025.

The latest update to that is that the parties have apparently settled and money will be paid out to the 7 million authors affected by the theft. You can read about that in an article on Ars Technica.

I knew that the lawyers were looking for affected authors but didn’t do anything about it because I didn’t think I was an affected author. (See above.) But then Dave Mark on Mastodon posted a toot with a link to an Australian Website that had a link to an Atlantic website page with a tool for searching which books were used to train the AI. And yes, 16 of my books (plus some repetitions) were on it.

So I tracked down the article about it on the Author’s Guild website and clicked the big button to provide my contact information. That put me on the lawyers’ website where I could provide my contact info and list all 16 titles.

I’m done now and can get on with my day. I don’t expect to get much from this, but I do think it’s important that authors stand up for their rights — especially when money- and resource-hogging for-profit organizations are stealing our work to build systems to replace us.

If you’ve written any books in the past 20 years, I urge you to see if you’re on the list and take action if you are. Theft of our work has to stop. Let’s work together to protect our work and our livelihoods.

I ease back into writing for hire.

One of the things I’ve been working on part-time for past few years is getting back into writing articles for publication.

Waterway Guide

Although I thought I had a working relationship with Waterway Guide, that fizzled out pretty quickly. The publisher was initially excited about working with me and made some suggestions about how much money I might earn writing for them. I did my part to help build content on their site with a never-ending stream of marina and anchorage reviews. (I still get the occasional compliment for my reviews.)

But after months on the Loop, I just got one article assignment — and I never got paid for it. I was very interested in helping to update the Skipper Bob books — especially the one for the Erie Canal — but was told other people were doing it. No other work was forthcoming. So I stopped writing reviews. I’m a professional writer, after all, and I’m not going to go out of my way to build content for a for-profit publication without getting compensation for my work.

Dockwa

I started writing for Dockwa last spring. Dockwa is a company that helps boaters arrange for marina reservations through the use of its easy-to-use app. Some marinas only use Dockwa so it’s a good app to have.

I used it to place a reservation for 10 days in Oriental NC back in 2023 when I had to take a class for my Captain’s license. The reservation was supposed to be refundable, but when I cancelled a month in advance, the marina refused to issue a full refund. Dockwa took care of it for me and I got all of my money back. They also accept payment via Apple Pay, so I’m able to get more cash back with every reservation than if I’d paid with a credit card. So I’m pretty sold on Dockwa.

Dockwa has a blog and some email newsletters of interest to boaters. I was first approached by them because they wanted to link to one of the posts in My Great Loop Adventure blog. Sure, I said. And when the newsletter came out, I got more than 1,000 hits in two days. They’ve since linked to other content there and have really helped me get exposure for the blog.

Then, in September, they asked if I’d write quarterly exclusive content for them to use in their blog. I’d get a byline and link to my blog. That seemed pretty good to me because I’d started writing a book about my Great Loop trip and the blog was going to be my primary promotional tool. The more eyes on it, the better off I’d be. Since then, I’ve written three quarterly articles for them (one of which went long and appeared in two parts) and have just sent in my fourth:

• “6 Things Every First Mate Should Know,” Dockwa Blog, October 1, 2024.
• “Captain Maria Langer’s Guide to Pets on Board,” Dockwa Blog, January 1, 2025. (I did not title this one.)
• “Single Handing a Cruising Boat | Part 1” Dockwa Blog, April 2, 2025.
• “Single Handing a Cruising Boat | Part 2” Dockwa Blog, May 3, 2025.

Although I’m not getting paid cash money for any of these, I did get a perk: I’m now a member of Dockwa +, a subscription service that supposedly gets me discounts and access to special offers at marinas. I’ll find out more about it the next time I’m on my boat.

Passagemaker

The biggest news is recent. I got into an email discussion with the editor of Passagemaker magazine, which I consider the premiere private boat cruising magazine. It’s a beautiful publication that comes out every two months and features cruising articles and boat reviews, along with informative how-to articles of interest to boat owners. I read it every month via the Libby app, courtesy of my library.

My email discussion with Jeff was about folks who finish the Great Loop but never get on a list of people who’ve finished because they don’t belong to the for-profit organization that keeps the list. Passagemaker published the list and (unsurprisingly) I was not on it, despite having “crossed my wake” in August 2024. That turned to a discussion of solo cruising on the Loop. I was asked if I wanted to write an article about the challenges I faced as a solo Looper and I said yes. I asked about compensation and was offered a very reasonable (if not generous these days) amount of money.

(Looking back at my list of published articles, I realize that the last time I was actually paid for writing an article was back in 2018, when I wrote for Vertical magazine. That was before I cut ties with them after one of their editors went into a jealous rage when she discovered that I’d been asked to co-pilot a Chinook on a fire contract. That’s a long story not worth going into here. Of course, I didn’t write anything for publication between 2018 and 2023, which is when I wrote that unpaid piece for Waterway Guide.)

I wrote the piece and Jeff was pleased. I assume it’ll come out in either the next issue or the one after that. I can’t wait to see it.

Jeff also invited me to be a guest on the Trawler Talk podcast. Passagemaker publishes the podcast and new episodes come out every month or so. We had a great conversation the other day, touching not only on my boating history, but some of my other careers including writing and flying helicopters. I expect the episode to come out in August unless he’s got one queued up before that.

It should go without saying that I’m pretty excited about working with the Passagemaker folks. I’m hoping it leads to more work in the future.

If I know it, I’ll write about it

Looking at my list of published articles, which appear in reverse chronological order, it’s interesting to see how the topics I wrote about changed over the years.

My very first published article, way back in 1987, was about auditing construction project budgets. At the time, I was working as an auditor for K. Hovnanian, a position I hated within a week of sitting down at my desk. (It took me just four months to escape to another job; I don’t even mention the time I spent there in any resume or work history.)

After that, I wrote about computers for auditing and then computers in general. Then computer applications and software reviews. Motorcycling was in there, too. How-to pieces. Aviation and flying helicopters. There’s lots of overlap — most of what I wrote in articles was computer related for most of my writing career. It was nice to move into aviation writing and now it’s great to be writing about boating topics.

I guess it all comes down to this: I’ll write about anything I know something about.

I like to write and I do need to apologize for neglecting this blog as I have for much of this year. I hope to do more of it soon — especially if I can write for quality publications and get a check when I’m done.

I think there are two reasons this time, one familiar and one new.

Let’s face it: I’ve been blogging at An Eclectic Mind since long before it had that name. My first blog post was way back in 2003. The blog had my name back then, Maria Langer, and it could be found at MariaLanger.com — where it can still be found; try it! I wrote about the things that were on my mind. For a while, I wrote computer-related how-to content, but I eventually broke that out and put it into a site called Maria’s Guides which I’ve since allowed to die. I wrote a lot about politics. I wrote a lot about my work and my play. I wrote a lot about my crazy divorce. I wrote a lot about social media. I wrote a lot about building a new life in Washington State.

Heck, I just wrote a lot about anything that was on my mind.

And that’s what I personally think a blog is for. At least this blog.

There are no ads here. No tracking. No annoying pop-ups begging you to subscribe or send me money. It’s just the blog of a writer — a person who has always written — sharing what’s on her mind.

If you come here often to read my latest and actually like a lot of what you read, great! I’m thrilled to have you here!

But if you stumbled in here unknowingly and have concluded that this blog is crap and a total waste of your time, well, just go away and don’t come back. It doesn’t bother me.

And if you think that’s harsh, well blame social media. (Also understand that “go away” is not what I originally wrote.)

The Dearth of New Blog Posts Here

I use an app on my Mac called MarsEdit to compose all my blog posts. One of the features I really like about it is that it keeps a database on my computer of every single post. So I can scroll back through past posts and see titles and dates of past posts. I can quickly link to them in new posts or just view them in Mars Edit or in a browser to remember what I wrote.

What I’m seeing right now is that between April 15 and July 26, I didn’t write a damn thing here. That’s more than three months.

I have two excuses for this dearth of blog posts.

I’m Busy

The first excuse is the one I usually produce for such occasions: I am busy. Unlike a lot of people out there, I live a full and relatively active life that often keeps me away from home and a keyboard I can actually type on. (No, I won’t blog from a phone or tablet. Don’t even suggest such a thing.)

Back in April, I was dealing with getting my boat across the mountains and having it serviced and put into the water to join a charter program. Everyone (other than me) was working in slow motion. It was incredibly frustrating and anger inducing. And costly. I paid to have my boat sitting in dry dock for two weeks when it only needed to be there for a few days — if the people doing the work would have done it promptly.

Okay, ’nuff said. No reason to get worked up about something that I dealt with months ago.

They got my boat into the water on April 8. Before that, it was in dry dock. After that, it was at it’s new home at Squalicum Harbor Marina. As I type this, it’s at Stuart Island, earning me about $3,200 for the week.

Anyway, I spent most of the month of April on the boat doing the things I needed to do and nagging everyone else to do what they needed to do. The boat lives in Bellingham WA now and that’s a 4-hour drive from my house. I spent more time on the boat in April than I did in my own home.

Then May. One of my Hobbies Gone Wild is silversmithing. I make jewelry out of gemstone cabochons, sterling silver, and copper. I’ve been at it since about 2018 and I’ve actually gotten pretty good. I sell my work at art shows in Washington and Arizona. May was the first two art shows I’d participated in since summer 2023. (Regular readers may recall that I did a 8000+ nautical mile boat trip between 2022 and 2024?) Before I could do the shows, I had to gather together all my show equipment, make sure it was ready to use, and make a bunch of new inventory. That first show was Mother’s Day Weekend and I had my second (now third) best weekend ever. The next show was Memorial Day Weekend, just two weeks later, and I had more inventory to make after selling so much at the previous show. So yeah, I was busy.

And in between all that, also in May, I was the instructor on a 5-day “Learn ‘n’ Cruise” class for San Juan Yachting. And then a half-day class refresher class for the first client to charter my boat. (Please don’t ask.) Around that time is when the problems with the new rudder indicator reared their ugly head. That fiasco took up a lot of my time, too.

When June rolled along, I taught another 5-day class. And then I took the boat out with my good friend Janet for two weeks in Puget Sound and the San Juan Islands with a stop in Victoria BC. If you’re keeping count, that’s 19 out of 30 days. I spent another 2 days on the boat before I went home.

By this time, it was pretty obvious that neglecting my yard and garden was paying off negative dividends. I split my time between making more jewelry inventory and trying to get control of the weeds and irrigation system. My boat trailer and truck camper remained parked in the driveway, which is absurd when you realize that I have a 2880 square foot garage.

Then July came and I had two shows to do. One was local but four days long. The other was all the way out at Sequim on the coast — a 5-hour drive. I took my camper. While I was out there, I dropped my laptop and broke the screen. I had to stop at the Apple Store on the way home to get it fixed. I also stopped at IKEA to buy some organization solutions for my jewelry studio. And a friend’s house. And a consignment client. It took me two days to get home. I camped overnight up in the national forest near Stampede Pass.

And now it’s still July. I’ve been home for nearly a week, thrilled that I don’t have to lock myself up each day in my jewelry studio to make inventory for yet another show. I’ve been pulling a lot of weeds, using a chainsaw daily, and assembling IKEA furniture. But tomorrow I head to Fort Walton Beach FL to help the owner of a Beneteau Antares 11 reposition the boat from there to Jacksonville, a distance of about 750 miles — if we go through the middle of Florida instead of around the bottom. The forecast on Wednesday is 81°F with a 50% chance of rain and at least 80% humidity. And have you heard that it’s hurricane season? Fun times!

I also spent part of this month writing two boating articles for publication and being interviewed for a podcast.

So yeah, I’ve been busy. And when I get back, I’ll still be busy, with yet another art show coming up at the end of August.

(I admit here that I’ve shuffled the deck a bit and hope to get another boating gig at month end — something interesting enough to get me to cancel that art show appearance. Cross your fingers for me.)

Excuse #2: The Journal

And now I’d like to introduce a second excuse for not blogging as regularly as I should: my journal.

Back in 2024, I started making daily entries in journal. I completed every day in 2024 and, so far this year, have only missed about 10 days.

In my journal, I write down what I did that day, along with any thoughts I had in my head. There’s nothing really great about it. I’m using it as a memory device for when I get older and my memory fails. I can look through the book and remember the things I did and thought.

But I think that writing in the journal has made it feel less important for me to blog. In a way, the journal is taking away my need to share what’s going on in my head. So blogging doesn’t seem as important.

Does that make sense?

Going Forward

I’m going to try to get back into writing or blogging on a regular basis. Over the past three days, I’ve written a bunch of blog posts for this blog as well as My Great Loop Adventure. I’ve set the publication dates and times so they don’t all get published on the same day.

As usual, they will be a mix of what’s going on in my life and what I’m thinking about. Some of them might interest you. Some of them might not. But then again, isn’t that what my blog has always been?

I’m also going to try to share links to blog posts and other online content that I’ve really enjoyed and can recommend. Hopefully, I’ll be able to share those on a timely basis.

Your comments are always appreciated.