At one point, shortly after discovering copy-and-paste, I decided to try sending the world's longest email. My strategy started with writing a sentence, "this is the world's longest email." I would copy this and paste it a few times:

"This is the world's longest email. This is the world's longest email. This is the world's longest email. This is the world's longest email. This is the world's longest email."

Then, once I had a paragraph, I would copy-and-paste that paragraph to make 10 or so paragraphs. Then I would keep snowballing: copy and paste the 10 paragraphs a few times, then copy and paste that larger chunk.

I know, I know, pretty advanced stuff.

After one of my attempts to send the world's longest email, my teacher called me over and pointed out that there were a few stray 'v's throughout the message. I had initially not confessed to using the copy-and-paste mechanism, but this was a smoking gun. A fingerprint, a signature of the tool I used to make the email. She caught me.

You may have realized by now that, unfortunately, this is another blog post about AI. Sorry! You don't have to read - the writing on this blog is mostly to help me think. I have only a slight illusion of an audience, and I am beginning to enjoy that more and more -oblivion is freedom.

The new copy paste

I found this recent soundbite from Hank Green quite fascinating:

"I need to come to terms with the fact that the level of dopamine I’ve been getting from interacting with LLMs... with doing more and more and more and more... is not healthy for me or good for the world" (via techcrunch).

I find the quote fascinating on two levels.

First, the dopamine thing. I am growing increasingly weary of this term, because I have noticed it being used as a scapegoat. Hank is doing the opposite here, and trying to take responsibility, but it's still a bit of a material outsourcing.

Like, I doubt that Hank or his doctor actually measured the amount of dopamine in his bloodstream over the last few years, and used this as the basis to determine the effects of his LLM interactions. The same is true for my students and friends who reference "dopamine addiction" - I am not aware of any way to gain a live view of one's dopamine levels, as if it were as measurable as insulin levels or heart rate or blood pressure.

I would imagine that psychologists feel about the word "dopamine" a similar way that computer scientists feel about the word "algorithm." It used to mean something very precise, and is now used haphazardly. We can generally recognize what people are referring to when they use the word, but it's not the same referent that the experts intended when the word first leaked into the general social lexicon.

I think what Hank is getting at, and what my students and friends are getting at when they mention dopamine, is less about a chemical and more about a hollow sense of satisfaction. A fickleness or a fakeness in the experience. Something feels fraudulent and deceitful. There is a shallow reward.

So that's what is interesting to me about the dopamine thing: people seem to use the word "dopamine" to reference an experience, not a chemical.

And then the second interesting aspect is that Hank is connecting this experience to LLM interactions.

I started considering this "addictive" phenomenon of the LLM experience earlier this year, when I decided to move away from hosting my website on ghost, and enlisted LLM-based coding tools to develop parts of the new site (I'm not good at writing javascript, and I don't necessarily want to or plan to ever be good at writing javascript, and LLMs can do it pretty well).

As I experimented with Claude Code, Hermes, Codex, Copilot, and a few other coding tools, I started to get kind of sucked in. The website project went well, and my website now runs on simple html that I can edit and modify quite easily without relying on ghost. The coding agents also helped me patch some security issues, and my website now gets "extra credit" in the Mozilla HTTP observatory.

After finishing up my website migration, I cranked out a few other tools I had wanted to make for a while. I have multiple degrees in computer science, but at some point you always end up reading documentation, searching error codes, browsing stack overflow, etc., and I did not have the time to do that for each of these projects.

For example, I have a consistent need for timers in class, where I can tell students "you have 10 minutes to discuss these question," and put something useful on the board to track the time. I've seen people embed YouTube videos of timers for this purpose, and it irks me. No, it infuriates me!

There are a few decent timers online, but nothing that had all the features I wanted. I didn't like the second-by-second (or even millisecond) tracking, the sounds, the ads, etc.

With the help of an LLM coding tool, I was able to put together a delightful little made-to-order timer widget for class, with all the features I needed, and no advertisements. It may seem like a small detail, but to me, it meaningfully improves the class experience. Plus, since I themed it after the CTA, it's much more fun to use/watch compared to the other options.

(Please use it, jackbandy.com/timer, if you are ever tempted to do something embarrassing like streaming a YouTube video just to keep track of time.)

So, that's the good/fun part. I was able to make a custom tool, one that I had wanted for a long time, but which I otherwise just did not have the time to make.

The bad part was, I think, what Hank called "not healthy for me or good for the world." There is something weirdly addictive about being able to crank out software (at least for people like me and Hank and this other guy I saw on TikTok who made a video about "saying bye to my wife before I go vibe-code for 4hrs").

The truth is, I don't really know how to describe how this interaction plays out. But I have done plenty of research about addiction and compulsion, as it was my initial critique of social media apps. At this point, the DSM has eleven criteria for addiction:

1. Using the thing more than intended
2. Trying to cut down or stop using the thing, but being unable to
3. Experiencing intense cravings or urges for the thing
4. Needing more of the thing to get the desired effect (tolerance)
5. Developing withdrawal symptoms when not using the thing
6. Spending more time getting the thing, using the thing, and recovering from using the thing
7. Neglecting responsibilities at home, work or school because of using the thing
8. Continuing to use the thing even when it causes relationship problems
9. Giving up important or desirable social and recreational activities due to using the thing
10. Using the thing in risky settings
11. Continuing to use the thing despite it causing problems to your physical and mental health

I've lightly edited this list from the source above, replacing "substance" with "the thing." For some reason, I still remember in one of my undergraduate courses, someone pointed out that fixating on substance addiction did not make sene, because people also get addicted to behaviors (like putting on chapstick or plucking hairs from their face). They were right!

I am not here to finger-wag about people getting addicted to LLMs. The point is that some of the addiction criteria do seem relevant to whatever it is about LLMs that are "not healthy for me or good for the world." One could reasonably argue that Hank's case, using the tool to help with his professional responsibilities, was a case of "using the thing in risky settings."

Personally, I have experienced criteria 1 at times, firing up LLM tools in the terminal more frequently than I anticipated. I never thought I would reach a "usage limit" - they literally cut me off!

I can honestly say that the other criteria do not resonate with me at this point. But they do worry me (again, things get pretty vulnerable here on the blog)! I do not have "intense cravings," for example, but sometimes I feel bored and think, "maybe I should fire up the old LLM and crank out another widget?"

The same goes for "risky settings." This summer I am preparing materials to teach Intro to Data Science, a class I have not taught before, and I use LLM to help with some of the materials. I am using quarto for slides, to avoid Google, and sometimes I prompt an LLM tool to draft a particular format for a slide.

Although I think my use cases are reasonable and defensible, this still seems a bit risky! Ideally, using an LLM tool for some tedious formatting will "free me up" to think more carefully about content, pedagogy, etc. But if LLMs, are in some ways, to some extent, a little bit, sort of like drugs, then I should be careful. We should all be careful!

If Hank Green can get hooked, any of us can get hooked.

My current approach is evolving and multi-faceted. I am trying to be transparent about when, where, and why I use LLM tools. I try to keep unvetted LLM-modified / LLM-generated materials in a "sandbox" and only move them out once I have reviewed and tested them. I keep everything in a public GitHub repository, so that anyone can see the various human decisions that went into the making the materials (and/or propose their own changes).

The Hank Green "situation" has reminded me to stay vigilant on this front. I have definitely had days where it feels like "I" am producing more and more and more... copy, paste paste paste paste. Copy. Paste paste paste paste paste paste. Copy. V paste paste...

I was fortunate enough to have a teacher in fifth grade who was paying attention to students and recognized my dishonesty. It was not even for a school assignment, of course, but maybe that makes my dishonesty all the more condemnable.

Did I violate all that's ever sacred between people? My teacher had been generous enough to exchange emails. A piece of "electronic mail" should have been an offering of peace - a message from one soul to another.

By automating part of my message (command-v, command-v, command-c), I made it fraudulent. I was not sincere. The whole purpose of the exchange was to have a real human sender and a real human recipient, and my message broke that purpose.

Email is long gone as a useful means of communication.

But what about code?

Personally, I never thought I would write important messages in code, in programming languages -- certainly not in javascript. I have had seasons of work where I felt proficient in various facets coding, or specific programming languages, and that can certainly be satisfying. But the point was always to make something, to create something - to express something?

I know people who deeply enjoy programming, who find beauty and joy in it. If I can be vulnerable... yes, I work in a computer science department, and no, I do not consistently find beauty and joy in computing. Maybe in data, but not in computing.

For me, the computing/automation was always a means to an end. Do you know what I mean? And that is why I hopefully will not feel the pressure to simply create a larger volume of work. I actually want to keep my volume rather low, to stay focused, to move slowly and fix things.

Now, I will save this markdown file, and run it through a script to convert it to html, and then commit the changes in git, then push those changes to GitHub, where the html will be copied around their servers, and perhaps - perhaps! - one day be accessed by you, fellow human.

Although I am trying not to be obnoxious about it, I did go on a 10-day silent meditation retreat earlier this month, and it was indeed a profound experience.

When I tell people about the experience, and the fact that yes, I willingly chose to go, many express surprise and confusion. No phone for 10 days? No internet for 10 days? Not talking for 10 days? Some people close to me were even somewhat resistant. Would I be brainwashed? Indoctrinated? Dragged through the cruel tutelage of S.N. Goenka?

On the winding path that led me to a 10-day silent meditation retreat in Rockford, Illinois, one of the key junctures was taking AP Literature ("AP Lit") with Hallie Karen Brinkerhoff, also known as "Brink," at West Jessamine High School -- just outside of Wilmore, Kentucky.

Brink was strict about graded participation, and I was shy about speaking in class. A month or so into the semester, my football coach pulled me aside during practice and said, "Did you know you're failing AP Lit? I was surprised to see you on 'the list.' Better get that figured out..."

I bandaged the situation and temporarily fixed my grade by doing copious amounts of extra research about class readings, and planning elaborate monologues that I could go through in class when Brink would ask open-ended questions: "Any other thoughts? Comments? Ideas? If you haven't shared yet, we would love to hear from you!"

My monologues did not really add to the class conversation, but it was enough to change my grade. And at the time, the grade was my main concern (although I also wanted Brinkerhoff to like me).

Brinkerhoff also taught writing courses at the University of Kentucky, and on her "rate my professors" page, one comment caught my eye: "Very hard grader, tough teacher. RUDE. but learn a lot."

But Learn a Lot

On the quest to improve my grade in Brinkerhoff's class, I discovered that one of my friends had earned hundreds of bonus points for his annotations of a book we read for class. Students had to turn in physical copies of our books, which Brinkerhoff would look through and grade based on our engagement with the text. (It's a good exercise, and one I might borrow for my classes to help mitigate AI usage, at least a bit...)

When I looked at my friend's book that had earned him extra points, every page had two or three written notes in the margins. I was desperate to get my grade up, so I resolved to do the same. It was daunting to consider how much time and effort it would take, but it was easier for me than speaking up in class.

One of the first books I annotated extensively was Siddhartha, by Herman Hesse.

Siddhartha came to mind a lot while I was sitting in silence for 10 days. And it's been interesting to think about this book/story at the same time The Odyssey is hitting theaters. The Odyssey is the classics hero's journey of "the west." I even saw an ad for a drink called "the hero's journey":

"Embark on an epic voyage with a tart, honey-spiced cocktail with a legendary finish. Order the Hero’s Journey mixed with Maker’s Mark when you see THE ODYSSEY, now playing. Available at MacGuffins Bar."

Man, I love when a drink description includes three advertisements! And you're telling me it's only $18?

I digress, but... this comical product description actually kind of captures the whole phenomenon. The "Hellenistic" tradition of Homer's Odyssey (aka "the west") is desperately attached to materials: the honey-spice, the b0urb0n, the tart lemon.

The heroic aspects of Odysseus' journey are material victories: conquest, survival, return. And while he does face some challenges and temptations through inward clarity ("tie me to the mast"), his victories are decidedly material. He strings the bow, slaughters the suitors, reunites with his dog, etc. You know the story.

But "the east" and "the west" have very different kinds of heroics. This is not a simple binary, of course, and ordinal directions are relative. So instead of eastern/western, let's compare "Hellenistic" heroics with "Dharmic" heroics. There are probably better terms. If you are one of the three people (or bots) who read this blog, and you know the better terms, please share!

Dharmic Heroics

The journey of Siddhartha, and the journey described by S.N. Goenka at 10-day meditation retreats, is an inward journey. The "wheel of Dhamma" turns toward no destination in particular. There is no material arrival, no literal pilgrimage, no final stop.

Siddhartha did observe that "he had to go into the world" (page 53 in my copy) to learn about the insufficiency of that world - to learn that "worldly pleasures and wealth are not good things." But that discovery could be made almost anywhere.

It's the classic aphorism that another one of my great high school teachers, Wayne Stevens, used to talk about a lot: "you can lead a horse to water, but you can't make it drink." (Stevens used to say something like, "well, you can also pull the horse's head really close to the water, or even into the water...").

You can give a student a failing grade because he doesn't participate in class, but you can't make him participate in class, or grasp the importance of participation. You can reward students lots of points for annotation, but you can't make them appreciate the significance of really engaging with a text.

And you can require students to read Siddhartha, but you cannot force them to go on their own inward journey.

Reading Siddhartha 13 years ago softened my mind. As I read, re-read, and annotated the book in high school, I found myself learning a lot. At the time, I thought other traditions had a monopoly on truth -- I had been told of only one hero who renounced the material world and taught of a kingdom within. And yet here was Siddhartha, echoing many of the parables I knew, sometimes with greater clarity.

I still have my copy of Siddhartha from high school, with all my annotations therein (e.g. "not going to lie, this is the most cheesy part of the book.") I used a black pen for most of my annotations, and a red pen to highlight parts about the Ferryman. We performed parts of the story as a play in class, and Brinkerhoff assigned the role of the Ferryman to yours truly.

One of my favorite lines (page 27 in my copy):

"You can learn a lot from a river"

Something I am often re-learning is that I essentially write for my own benefit. Admittedly, I thought this might change when my book came out. "Look! I'm a published author! I wrote about big tech and social media and enshittification!" A small, egoistic part of me imagined I might be busy with speaking gigs and interviews for the next few months.

And, admittedly, that part of me felt slightly disappointed when the book release date came and went with little more than a slight engagement boost on the very services that my book tries to critique.

Hmmmmm...

But, as I have realized and re-realized and re-re-realized before, that is not the purpose of writing. For all intents and purposes, nobody reads anymore. I mean yes, some people read. I even see quite a few books on trains and busses these days, at least in Chicago.

But if you read more than one book per month, you are pretty weird. And if you read books for pleasure, you are super weird. If you are reading this blog post, yikes, sorry, it doesn't get much weirder...

Or, to reframe it, if you are reading this blog post, you are super special! A real-life snowflake! I have even seen some creators try to leverage this argument recently: "if your algorithm is showing you this video, you are part of a rare group of people who..."

Anyway.

Over the yeas I have gone back and forth as to how much audience actually matters. At times, I really want an audience. I spend months or even years on written work, particularly for "peer-reviewed" pieces. Please, somebody, take a look! I spent two years of my one wild and precious life trying to make this text useful and meaningful - will you please check it out? No worries if not!

(Also this is not the case for blog posts, sorry, I only spent a couple hours on this).

I also know that part of me still cares about audience because, in some cases, I have submitted article pitches or applied to speak at various events. When those proposals were rejected, I felt disappointed. I fear the inescapable conclusion of this disappointment: my egoistic desire for an audience remains.

And yes, the acceptances still feel good. But the drug wears off. I used to feel quite a high when Google Scholar emailed me to add in new articles. When they emailed me that I had crossed 1,000 citations, that was also a bit of a rush, although I have slowly learned that most work seems to cite other work without more than reading an abstract or looking at a couple important figures.

To get back to the point: we write for our own benefit. When I write something, I change my brain. I can almost feel the synapses firing and the new neural pathways expanding and connecting throughout my mind. I know that after I write this blog post, because I wrote it with my own mind, I have a much stronger grasp on the latent ideas that were trying to come to the surface.

For me, this is enough. Although my ego may still interfere at some points, in principle, I believe that the writing process itself is reward in and of itself. If zero or one or two people read this post and connect with it in any way, that is phenomenal. Miraculous, even, given the staggering cosmic odds otherwise.

AND (not but - and), if I am the only soul who ever comes in contact with these words, that too is phenomenal. I wrote them myself and they helped me makes sense of the world. And, more than likely, at some point in the coming days or weeks or month, I will talk to another soul who will speak a word or phrase that activates some of the same synapses and neural pathways that I have activated here. And we will forge new pathways as we make sense of the world together.

Perhaps that is the point of writing, regardless of who is in the audience. Perhaps the purpose is "just to create something beautiful and worthy of contemplation," something that helps us make sense of the universe and find meaning in it. And perhaps that was the point all along.

Why does TikTok feel so different in 2026 than it did in 2020? I have been thinking about this for a few years, honestly, and I have yet to actually share any thoughts because, well, maybe I'll answer that later...

This is a slightly longer (by longer, like ~2600 words, 5-minute) explanation of the theory, just for the nerds, so if you are not a nerd, feel free to move on.

The reason I am sharing this now is because I sense a growing recognition of just how terrible our media environment has become. "We" (the working-class nerds?) are developing shared, common knowledge showing that these systems not working well for any of us. Many of us have come to agree that they are not serving us in our personal lives, they are not supporting our communities, and they are definitely not serving our larger democratic project. More and more people are seeing this, and are ready to make some change.

And as we pursue change, I think it is helpful (at least for us nerds) to discuss the specific reasons why these systems are failing us. That is the reason why I am sharing this.

Also, a few weeks ago I saw a 10-minute video by Kevin James Thornton, discussing the long, sad story of TikTok's crashout and the decline of social media more generally. And that basically gave me some kind of subconscious permission to share my perspective on it.

Thornton shares some great insights. In particular, he describes his experience of "amassing an audience of people who are interested in a niche topic" and then slowly losing the ability to actually communicate with that audience through the very channel which provided that audience. I have nowhere near the audience size of Thornton and other creators, but I experienced some similar dynamics on a small scale.

Anyway, here are some thoughts.

What has provably changed

I will do my best to separate some factual, provable elements of changed from my own hypotheses. I do not have secret access to the TikTok control room (I did request access to the Academic API a few years ago, but they never got back to me), I am just trying to make sense of things.

So, first, some facts.

Fact 1: TikTok sold out

TikTok got bought. The company literally sold out. More precisely, after years of political pressure, TikTok's U.S. operations were moved to a majority-U.S.-owned venture called "TikTok USDS Joint Venture LLC." That venture seems to be a few billionaires in a trench coat (including Larry Ellison), and ByteDance keeps a 19.9% stake.

The Wikipedia page for selling out includes a musical/artistic definition, "associated with attempts to tailor material to a mainstream or commercial audience." In the context of TikTok, selling out relates more to compromising integrity, morality, authenticity, or principles "in exchange for personal gain, such as money or power."

For TikTok, this was really just a matter of survival. I am not sure what other options they had after years of political pressure: the first threat to ban TikTok came July 31, 2020. A few days later the president signed an executive order which would ban TikTok if it was not sold within 45 days. After a few years of ambiguity, in 2024, Congress passed a divest-or-ban law, ("Protecting Americans from Foreign Adversary Controlled Applications Act"). And In January 2025, the Supreme Court upheld the law. The U.S. venture was finalized in January 2026.

(Side note on this point - I'm intentionally using "the president" and "the administration" because this is one case, among many, where the disagreement is much more top-verus-bottom, elite-versus-plebians, owning-class-versus-working-class, rather than left-versus-right or dnc-versus-gop or republican-versus-democrat. The elite's efforts to buy out TikTok was remarkably bipartisan.)

Fact 2: TikTok changed their algorithms

In a sense, this goes without saying. The algorithms are always changing.

When I was trying my hand at BeInG a CreAtoR in 2020, I developed the following mental model of the TikTok algorithm. It largely aligns with the "batch theory", which is a shared conceptual model of how TikTok's algorithms work at a high level. It goes something like this:

• phase 1: after you post the video, TikTok shows it to a batch of about 100 random people in their "for you" feeds.
• phase 2: TikTok analyses engagement from the first batch and compares it to other videos
• phase 3a: "instant hit" - if engagement is looking really good, TikTok will push it to a larger batch (e.g. 1,000 people)
• phase 3b: "on deck" - if engagement is looking okay, TikTok might keep testing it
• phase 3c: graveyard - some videos go to the graveyard, for whatever reason, though they might get to compete in the loser's bracket and resurface later.

This is a generalization, but the basic idea is that there are gradual, recursive phases of the "instant hit." That is, if your video did well enough in the first 100 random feeds, it will put it in another random pool of, say, 1,000 feeds. There is another competition there among videos that "made it out" of the first round. Then the next round might show it to 10,000 feeds...

It's like a bracket, where each round gives the video more views.

TikTok has shared some details about its systems, which mostly align with the batch theory.
The company blog posts mention specific signals such as user interactions, video metadata, and account settings, which are used to place the videos and evaluate performance.

But the knobs are always in flux -- what Cory Doctorow calls "twiddling." How much weight does watch time get? How much is a comment worth? Does shopping content make a difference, and if so, how much? These are the kinds of features that are always fluctuating as TikTok continues experimenting on us.

In September 2025, the BBC reported that Oracle was "auditing and inspecting the source code and recommendation system underpinning the app, and rebuilding it for US users using only US user data."

So, in addition to the ongoing changes, there may also be some distinct step changes post-sale. It's tricker to determine whether those changes to the algorithm really explain why the app feels worse, or if there is something else. One more fact before I offer my hypotheses...

Fact 3: surrounding culture shifted

TikTok was associated with a specific moment in U.S. culture: early pandemic, lockdowns, indoor boredom, the 2020 elections, and the sense that everyone was online basically all the time.

There was something special about TikTok at that time. Whatever made it "special" is somewhat ineffable, but I do think Eugene Wei does a good job breaking down the details. In particular, TikTok and the Sorting Hat describes TikTok as an automatically sorted network based on people's interests, which placed us into invisible subcultures without requiring us to follow anyone. It just analyzed what we watched and liked, and put us in a bucket with similar people.

But our lives and our participation on the app evolved. I enjoyed The Noisy Room website (for the most part), and how it frames this evolution: a few voices became amplified in a way that really distorts our sense of what is happening and what people are actually saying.

Again, early-pandemic TikTok felt like a fun little secret passageway to connect with other people forced online. Now that passageway goes to a loud casino-like room where it seems most people are yelling at us and selling stuff (right now TikTok really thinks I'm going to buy an Apple CarPlay add-on device).

Some hypotheses

My general hypotheses is that TikTok got bullied into reducing its more subversive features. TikTok created an amazingly effective algorithmic system for creating and discovering social content. That's how it took off: the app provided a fun, functional, collaborative communication system.

Yet after the the political threats and bullying, TikTok was forced to prove it would/could obey the powers that be. It had to show that it could be controlled, that its mysterious recommendation system would not show videos of Momo asking children to attack people (a total conspiracy, by the way).

I propose three prongs of this argument - three changes which I hypothesize TikTok made to their systems: (1) more "manual overrides," (2) decreased randomness/quirkiness, and (3) increased exploitation. Again, these are just hypotheses with supporting evidence, and I can't yet prove them. At the same time, perhaps more importantly, I have not been able to disprove them (but maybe you can help, dear reader?).

Hypothesis 1: TikTok Increased Manual Overrides

There was a sense in the early days of TikTok that anyone could go viral. It was not necessarily unique to 2020 TikTok, but it was palpable. There was a sense of excitement and possibility.

But the record now clearly shows that TikTok is not a merit-based popularity machine. For example, The Markup reported in 2021 about how Tinuade Oyelowo and other creators constantly struggle to figure out TikTok’s changing algorithm. Then in 2023, Forbes reported on TikTok's internal heating button, which they could use to manually boost selected creators/videos.

TikTok has made some noteworthy efforts in response to these critiques. I was especially excited about the transparency center, which was announced in late 2020 and promised to share more information about how the recommendation process works. An important opportunity here was for TikTok to be able to address various folk theories about their algorithm and say bluntly, "no that's not how it works."

But at this point the transparency center, as well as the academic API and related efforts, seems to have been mostly theater. That is, TikTok was just taking performative measures in attempt to prove their app was safe, predictable, and controllable.

Hypothesis 2: TikTok Dialed Down the Quirk

One of the knobs that can be "twiddled" in an app like TikTok is the randomness of the recommendations. In technical terminology, this is part of the explore/exploit tradeoff.

Early TikTok seemed to achieve an impressive balance of exploration and exploitation. TikTok's app description, as quoted by Vice, described its ability to "offer the most relevant, interesting, fun, quirky, head-turning videos that you'll never want to stop watching."

"Quirky" is key. These unexpected, off-beat, unusual, eccentric videos laced our feeds in precise, effective doses.

Even TikTok acknowledges the importance of the balance. In Casey Newton's piece about the transparency center, TikTok executives discussed the risk of "repetitive experiences" and the need to interspersed diverse types of videos. They even talked about avoiding filter bubbles! But the point was: random discovery and quirkiness in the feed was really crucial.

This likely requires a bit more of a hands-off approach to content moderation. It entrusts the people scrolling to engage in ways that elevate "quality" content and leave behind inappropriate / undesirable content.

To put it another way, it's easier for a company to review and moderate videos when it's just ten videos that are all getting a million views. But when those ten million views are dispersed across a thousand videos, the company either needs to pay for a lot more content moderation, or, trust their users are ready to separate the chaff from the wheat with their winnowing forks.

My hypothesis -- again, it's just a hypothesis -- is that part of the "retraining" and "rebuilding" of the TikTok algorithm has been a lower dose of exploratory recommendations in the feed. They made TikTok less quirky. The feed still shows plenty of "good," safe content, but the fun surprises are mostly gone.

Hypothesis 3: TikTok Spammed the "Exploit" buttons

The flipside of lowering the quirkiness and randomness is a higher dose of "hits" in the feed. Sort of like a radio station playing safe songs over and over again. They are good enough to keep the radio on, but not good enough to make you feel alive.

Again, it's known as the explore/exploit balance. It's a classic dilemma: are you going to go somewhere new, or go back to your same spot? In the TikTok library, the algorithms are balancing similar questions: do we recommend our greatest hits (exploit), or recommend new stuff that we're not sure about (explore)?

Exploration helps a system learn and adapt, while exploitation allows it to "cash in" on what has already been learned. My hypothesis here is that TikTok has leaned into the exploitation side of things.

This "cashing in" often means copying what has worked in the past, whether that be in specific formats, topics, aesthetics, sounds, user interests, etc. The videos and feeds all start to feel more homogenous -- more like slop.

And don't forget about the literal "cashing in" via TikTok shop! Every video you watch, and really, every tap in the app (and elsewhere) can eventually be used to make you buy something. Just keep watching and tapping -- let TikTok handle the rest (don't worry, it will tell you what you want and where you where to click to buy it).

The cycle is somewhat complicated, and it's why this question of "why does TikTok feel so weird now" has no single simple answer. Everything interacts with everything else: recommendation algorithms, creators, audiences, TikTok executives, politicians... It's kind of a mess.

How do we make the internet fun?

TikTok grew as a fun, quirky, high-volatility video sharing app system that put a spotlight on remix culture and spontaneous creativity during a unique cultural moment in 2020. (Sorry for the run-on sentence). As of 2026, its enshittification cycle appears to have come to a close. After selling out, changing its algorithms, and meeting a changing cultural environment, it lost its magic. The unicorn is alive, but not well.

Notably, it also followed the standard three-step recipe for enshittification. At the beginning (step 1), it was essentially fun and enjoyable for the larger mass audience. Then (step 2), it used that accumulated mass audience to entice advertisers and creators, making their experiences a little better and our experiences a little worse (e.g. with ads and shopping links). Finally (step 3), now in the hands of U.S. owners, the TikTok app is a miserable experience for everyone except TikTok owners.

An important clarification: I am not saying that "TikTok sucks now because the algorithm changed." TikTok does suck now, and its algorithm did change. But it is a larger ecosystem that fostered algorithm changes, ownership changes, and other operation changes that eventually enshittified the app.

I don't know if a fun internet is possible anymore. And maybe fun is not the right word. Perhaps we want more value, or more meaning, or more connection, or more quirk. Personally, I'm not even sure what I want from the internet, if anything, but it's definitely not Instagram or TikTok.

And that's basically the end. I don't even have anything to sell you. I did write a book that is somewhat related, but I don't get commission if you buy it (I get a one-time payment from the publisher that will almost cover all the coffee I bought while writing it). So this post is truly just one nerd trying to share some of what I learned with other nerds out there. I don't even have an alternative app to recommend.

Anyway, I hope you learned something from all this and I would be curious to hear your ideas and questions through an uncorrupted channel.

Step 1: Combine ingredients via blender or whisk or shaking
Step 2c: For a sweeter version, add simple syrup or agave or similar
Step 2b: For a more citrusy version, add lime juice or similar

Sources consulted:

• Stress free Mickey
• Disney Food Blog
• Attempts at Domestication)

Yeah, online recipes are a pet peeve of mine. More than that, really: online recipes are a significant source of pain and struggle in my life. I hate them. These days I print them. And sometimes when I invest time wading through seventeen advertisements and clicking through five popup videos and such, I write up my result in hopes that other humans may some day benefit from my labor.

Every few years I get on a Star Wars kick and invite some people over to watch, which has now twice resulted in attempts to make blue milk. The version that Disney makes and sells for $19 a cup is quite good! This base recipe is not quite as tasty, but it's a solid base for customization.

I prefer the full-liquid version that you can chill or put over ice, but I think Disney's is more like a slushy. Homeade slushies tend to separate too much for my liking, but if you poured this into one of those icee machines it would probably turn out pretty well.

This version is super simple and dairy-free, and in a small sample of taste-testers, most people say "wow this is actually pretty good!" It's also a very safe and versatile base, so you could start with this ratio in the blender, and then add ingredients and taste test until it's just how you like it.

The only catch is that if you drink this before watching Andor, there will be part of you that dies a little more with each sip. So it's probably best to watch every episode of Andor before making this at home.

Cheers!

Most weekdays, before I start making numbers and letters appear on a screen, I ride the train. Earlier this week on the Blue Line, I sat on train #3412. Each 3200-series car holds about three dozen people.

Similar to the rituals at McDonald's, for these three dozen or so people sitting on the train, we agree to a "temporary subordination of individual differences in a social and cultural collectivity."

That's the kind of sentence I could think about for a whole month. I read it in a book chapter called "Power, Technology and the Phenomenology of Conventions: On Being Allergic to Onions" by Susan Leigh Star, and it made me want to keep reading Star's work. So I followed it up with her "Reflections on the Origin of a Concept" and came across yet another quote I could think about for a month. She describes her original scholarly desire as to understand "the nature of cooperative work in the absence of consensus."

Which brings me back to train car #3412 on the Blue Line this week. About three dozen other people sat on the same train car, all of whom are, to some extent, temporarily subordinating their individual differences. We travel in the same direction, but we do not share a final destination. "All passengers must leave the train."

The social norm is to avoid looking at other people on the train, but I always look around. This little metal room weighs about 55,000 pounds and travels around 60 miles per hour -- forgive me for breaking social conventions, but I need a glimpse of my fellow passengers.

We go above ground for three stops -- California, Western, Damen -- before going underground for four miles. Approximately 72,000 of us make this journey every weekday, entrusting the CTA to move us underground for miles on end.

When we step off the train, we all walk to different places. Even on the train, though we move in the same direction, we sit in different seats. Although I did see a happy couple recently try to sit on the same little backward-facing seat on a 7000-series train the other day.

Do you prefer center-facing seats, moving left to right or right to left? Or do you prefer facing forward, or facing backward? I have avoided sitting backward for months, and when I finally tried it earlier this week, it... wasn't that bad at all.

The direction of motion, though sometimes shared ("social and cultural collectivity"), is always relative and arbitrary and temporary. I am only "moving forward" from my own point of view. If someone was watching from outside the train, who knows what is the relative direction of movement? Much less if the person is watching from earth orbit or moon orbit... ("You can’t imagine what consternation your tuna fish question has raised...")

Even from my own point of view, I am not moving perfectly forward, although the seat faces that direction. There are bumps that move me up and down, and wobbles that move me left and right, to and fro. There is no such thing as a perfectly straight line.

As of April 2026, I have taken the train to work for about two semesters. Sixteen weeks each semester. Going in to the office four days per week, I get to ride the train about eight times any given week.

Eight rides per week, 32 weeks, and about 30 minutes on the train each ride makes for about 7680 minutes, or 128 hours. These are 128 hours of my life which has been ordered by the CTA. The number grows when accounting for the waiting times, walking times, and various other rides I take on CTA buses and trains.

Like McDonald's, the CTA "shares sovereignty with other enterprises which seek to order lives, and of coexisting principles of order which in fact stratify human life." Some passengers sit while others stand. Some face forward, some face backward, some face "the center." Some sit on 7000-series cars, with carpet-like seat backs, video screens, and improved air circulation.

In the Fall, I saw lots of people reading books on the train. These days, it seems like most people are back to earbuds and touch screens.

Again, forgive me for breaking social conventions. I know I am supposed to sit (or stand) and stare vaguely at the floor or look at an advertisement for a new weight loss drug or gambling website.

But my mind is occupied with other matters: How does the CTA move 70,000 people back and forth on this underground track every day using giant metal boxes? How have we subordinated so many individual differences and created a collectivity that moves us in the same direction for miles on end, even though we all have a different final destination?

I must look around to see which of my fellow passengers realize the profundity of our journey.

Kramer's piece points out that even Discord "is moving closer to offering stock and becoming a publicly-traded company." I was also disappointed to learn recently that BlueSky is headed in a similar direction, following a $100 million funding injection from "Bain Capital Crypto."

Ugh!

Perhaps I was overly optimistic about BlueSky. I mean, it is basically a rewrite ActivityPub, with a few superficial changes. Perhaps if BlueSky really believed in the spirit of protocols, they would work to improve the existing protocol. I am still on the app, but, my hope is dwindling.

In fact, I am still on pretty much all the apps. My defensible justification is mostly related to identity protection. That is, if I deleted my Twitter/X profile, for example, it would open the door for impersonators.

But that security justification often becomes more of a rationalization. When I lose an hour or two in one of these apps, I wonder, "why do I even have this app again?" "How is this helping me reach my goals in any way?" And then I think through deletion, and remember that I want to maintain an "active" account.

As I argue in my book, the "social" components of these "social media" apps have become so dilluted as to make "social media" a misnomer. The apps do not help me connect with people. And if I made a more thorough attempt to answer "why do I even have this app?", I think it would end in someone else's intentions (not my own values or goals).

So my question has become, how do I "quit" without deleting all my profiles and such? I have done plenty of unfollowing, scrubbing, and data exports across varous services. I have what I need from them. If I deleted these profiles, the main thing I would lose is the ability to represent myself on these platforms.

In short, "digital sovereignty" must mean more than just getting off these apps and doing something different. To me, the identity control, and the option to post and represent myself in some (shallow) way, is worth it. So I will not be deleting my accounts any time soon.

That being said, I am trying to build some new habits - I hear it's one of the best ways to kill bad habits! Namely, I am publishing the "text" portion of this website as an RSS feed.

Yeehaw!

Nothing groundbreaking here. Plenty of people have advocated for returning to RSS feeds, and the format dates back to 1999.

What makes sense to me is that I can "post" to an audience that actually wants to read what I have written, however small that audience may be. And I don't have to rely on a massive, for-profit multibillion dollar corporation to make it happen. I mean, my current setup relies on GitHub pages, but the code is all mine and I can move it somewhere else if I want.

Anyway, all this to say: welcome to my RSS feed.

Here is one quote that has made me think quite a bit: "In order to have a public voice, you have to care about something. So figure out what you care about."

Oof!

And he doesn't stop there: "It's not just being against something, and it's not just wanting to have a community. It means having values that make the world make sense."

Double oof! By the way, this is from his piece, "Find Your Voice."

Agre describes a public voice as in between a private voice and a commercial voice. A private voice only prioritizes honesty -- it's just for one's self. The posts here on my website lean towards a private voice, although I've found that the simple act of making a piece discoverable online helps me write (and think) more clearly.

Then there's the commercial voice, which has a "predetermined effect on the audience." A commercial voice tells the audience to do something, but in a way that the audience wants to hear. "Follow this account!" "Buy this book!" "Delete this app!" "Sign this petition!"

One alternative to the private voice and the commercial voice is the public voice. I've been thinking a lot about this concept of a public voice, particularly after a weird experience with posting recently.

I do not make any commission money on my book. I still feel some obligation to promote it, so I put together a series of posts and started dolling them out on Instagram and TikTok. I have over 8,000 followers on TikTok (🤪) from a brief stint of videos I posted during the 2020 pandemic lock-down. Something about being home all day every day allowed me to engage quite differently. I digress...

Anyway, despite those 8,000 followers, my book promo posts have about 150 views each on TikTok. LinkedIn claims my post has over 3,300 views. My Instagram posts also seem to be averaging about 150 views, although my initial announcement did a little better.

I have to remind myself, this is not what I actually care about!

For transparency, audience metrics are alluring. Very alluring. Seductive, even. Maybe it's just the thrill of seeing numbers go up (at SIGCSE this year, someone presented results by saying "the number goes up, so, yay" and I had to contain my laughter). Or maybe our brains have been tricked to receive those numbers as a social signal of some kind.

This creates a kind of dilemma: "You must be extremely online to build an audience, but you must be extremely offline to actually have a thought worth sharing." It is also a dilemma "between visibility and depth... work must be formed in private conviction first, otherwise the demand for attention reshapes it before it knows what it is."

Side note - I came across this post via Taylor Lorenz on Instagram, who shared a screenshot of a tweet, and the tweet had an attached screenshot of the substack post. Truly it has been said unto you, the internet has degraded into "five giant websites, each filled with screenshots of the other four."

So, back to what I care about. Well, for the book promo posts I have been working on, I am curating vector graphics of 50 or so different mobile devices -- iPhones, iPods, iPads, and some Android devices as well. I always found it fascinating that a few engineers could add a button or change a screen resolution, and then millions of people would deal with that decision every day. I vividly remember following "antennagate" when I was in high school.

When placed in the Instagram grid, the graphics for the iPhone 4, the first-generation iPod touch, and the iPhone 5 all look pretty similar. Instagram shrinks them. It's just another post.

Although I am not an Instagram guru, I know enough about the app to realize that this series of posts will not "do well." They look similar and they contain a similar message ("this app sucks").

What I care about, here and elsewhere, is that the posts are actually quite different. The iPhone 4 was quite different from the iPhone 5! Apple stretched the screen from 960 rows of pixels to 1,136! Apple switched from the 30-pin iPod-style connector to the reversible "lightning" connector! The details...

From what I can tell, apps like Instagram and TikTok do not really care about details, which may be due to the fragmentation in these systems. By designing the feed around individual posts, each post is forced to be its own self-contained story. There is no way to know what kind of post will show up next. There is no way to know what kind of audience will see what you post. It all collapses into the vortex of the feed. If you want to know more about context collapse, time collapse, and relevance collapse, well, my book goes into more detail.

Perhaps this does not answer the titular question, and perhaps it does. It is tempting to make a list of "what I care about," however this does not seem like the kind of inquiry to answer in a list. If I was really forced to choose a single word, I think it would be democracy. If I were forced to choose a single phrase, it would be "the details of democracy," because I mean democracy in a deep and detailed sense: the day-to-day challenges of attempting self-governance and self-determination within a social system designed hundreds of years ago. Our democracy, if we are still willing to call it that, has yet to contend with the realities of the 21st-century. I am especially concerned about a handful of big tech companies managing what we see for several hours every day. That is one attempt at stating what I care about.

It's been said that absolute power corrupts absolutely, and it seems that the same goes for monopoly power. We saw it happen with oil and tobacco. It happened with steel and railroads. It happened with newspapers, television channels, and phone companies. It happened with computer operating systems. It seems to be happening with restaurants and airlines.

Heck, it happened with tea and spices as early as the 1600s -- they taught me about the Dutch East India Company in public high school!

Perhaps the title of Sally Hubbard's book says it most succinctly: Monopolies Suck.

And yet, here we are in 2026, with a small handful of companies that own and operate almost the entirety of the digital communication industry. Sure, it is technically an oligopoly, not a monopoly. There are a few of them, so the car manufacturing industry may offer a more accurate analogy than Standard Oil.

My book, Algorithmic Media, explores the unchecked power of big tech companies, particularly in the context of digital communication. The chapters discuss communication challenges that arise for three groups of communicators in the "digital town square:" (1) journalists and news organizations, (2) advertisers, and (3) common social connections like friends and family. Here are a few more details.

Marshall McLuhan famously said the medium is the message, and Neil Postman offered a revised version -- the medium is the metaphor. I prefer to say that the medium is a message or a metaphor -- a very important one, just not the only one.

(I am also attached to one other revision of McLuhan, related to the terminology of the "Global village." His idea was that the spread of new media technologies would connect everyone in the world, allowing us to communicate as if we all lived in the same village. Sounds nice. What the big tech oligopoly has created instead seems to resemble a global casino, where elite owners adjust dials to control the flow of different resources. The enshittified town square has a lot more in common with a casino than a real town square -- TikTok, SnapChat, and Instagram offer addictive entertainment, through features that encourage compulsive use.)

Back to "the medium is the message." In today's digital communication ecosystem, the medium/metaphor/message in question is often algorithmic. That is to say, the media through which we exchange messages will often use algorithmic processes to determine who receives what messages, as well as when and where they are received. To rephrase McLuhan and Postman, "the algorithm" itself has thus become a sort of message -- a central metaphor in our day-to-day thoughts and decisions.

There seems to be broad (and growing) awareness around "the algorithm" (or "my algorithm") as an influential mediator in our communication. But what exactly does this entail? What are the effects and implications of algorithmic media? And why does it matter?

One important implication of our current algorithmic media ecology is that big tech companies have come to dominate. They own us. Hundreds of millions of people spend hours a day on their screens, only to use a handful of services from the big tech oligopoly: YouTube, Instagram, Facebook, TikTok, and Snapchat are among the most dominant.

Whether the messenger is a news organization, an advertiser, or a friend / family member, there are fewer and fewer guarantees as to how their messages will be delivered. Many people know this intuitively -- your post may or may not reach your followers, regardless of its quality. Cory Doctorow describes this as a breakdown of the "end-to-end" principle. In a service that follows the end-to-end principle, messengers can reliably send messages to people ("receivers") who have willingly chosen to receive such messages.

Some services used to follow the end-to-end principle, for example, when your feed showed posts from people you willingly chose to follow. It was nice! But it turned out to be a trap -- a setup for future enshittification. The typical three-step enshittification, according to Doctorow, goes like this:

1. Be good to users
2. Be good to business customers, in part by abusing users
3. Abuse business customers and claw back all possible value

Here is how that plays out in the context of algorithmic media services.

Phase 1

In phase 1, the communication service works well. It follows the end-to-end principle, such that you can expect your posts will reach your followers, and your feed includes posts from people you willingly followed.

Nostalgia can be dangerous, but sometimes it's worth remembering the good old days. Remember when you opened Instagram, and you got to see posts from friends and family? And you could scroll through the new posts until you saw a picture that you recognized, at which point, it really was a good time to close the app. It was nice, but it was short-lived.

Before November 2013, Instagram did not even have advertisements. In the beginning, before their enshittification, services like Instagram actually worked: they showed your posts to anyone who asked to see them. To use Cory Doctorow's terminology again, the service reliably delivered messages "from willing senders to willing receivers." It was an effective communication tool, attracting millions of active users.

Phase 2

Then came advertising. This is phase 2, in which the feed remains somewhat usable, showing posts from people you followed. But in phase 2 of enshittification, any advertiser willing to pay the price is able to put a post in your feed.

Phase 2 is the key growth period for companies to build up their business clientele. The advertisers have a heyday. As long as the feed is usable (not necessarily good, just usable), people keep using it. In this phase, the owners sit back and relax. All they have to do is get the advertisers hooked. And if your company is bringing in millions of dollars in advertising revenue, it's worth the investment to hire full-time employees whose only job is to make sure advertisers keep up their spending.

You can think of phase 2 like adding a toll to a road, but only after people have become used to driving on that road every day to get to work. Oh, and the alternative roads are no longer maintained. With advertising, the company starts turning communication into a challenge. If you want to reach your followers, you cannot simply publish a post and expect it to appear in their feeds. Now, to communicate with your friends, you need a strategy: multiple posts per day, carefully-selected hashtags, engaging captions, etc. etc. etc. Even then, your message might not reach your intended audience. Welcome to the great audience competition of algorithmic media, may the odds be ever in your favor.

Phase 3

Phase 3 is the current state of affairs. We all know some things are wrong. I can't speak for everyone, but at the very least, it seems like more and more people recognize that these apps suck.

The platform becomes so saturated with advertisements that users develop genuine fatigue. By the early 2020s, Instagram users were seeing about 4 ads for every 10 posts. The only one who benefits in phase three is the technology company itself -- users and advertisers are frustrated, but they have no meaningful alternatives (thanks to phase one and phase two locking everyone in). Instagram is no longer fun, unless you own it.

This third phase of enshittification makes for a structurally chaotic and unpredictable communication ecosystem. The primary goal is no longer to facilitate interpersonal communication, but to keep advertisers paying and keep users coming back -- to build a better mousetrap. Feedback loops emerge that reward a small portion of users while undermining the more balanced, two-way communication that made these services usable in the first place. Pew Research found in 2024 that about half of adult U.S. TikTok users had never posted a video, and the top 25% of users produced 98% of public videos. A similar study of Twitter users found that the top 10% created 80% of tweets.

The monopoly / oligopoly patterns thus show up in multiple ways. We have a few companies that own these apps. And within these apps, there is a select group of users who dominate the attention field. Again, welcome to the audience competition of algorithmic media!

What next?

As I have talked to more people about the topics in this book, I have become more and more encouraged to realize that more and more people are simply fed up with big tech companies and their services. We hate it. We know these apps suck. Maybe that means this book is late to the party. And, maybe, that means we are ready to organize for better media futures.

How do we do that? Perhaps by some combination of public media, public education, real democratic governance, and probably some other changes that will require us to do more than click. I am convinced that we need to keep making a big fuss about how much these apps suck, and why big tech companies are to blame. Technology can still be cool!

But the purpose of this blog post is just to give you a taste of my book, not to solve all these problems. So if you enjoyed reading this post, you might enjoy the full book. If not, well, thanks for your time.

So, when I published a new website that was substantially shaped by Claude Code, I felt the need to share some of my reasoning.

To get a little meta, let me start by saying that drafting (and/or editing) my writing is one of the tasks for which I refuse to use LLMs. I believe writing is thinking -- that is, (I think) one of the best ways to understand what is happening in my brain is to translate my thoughts into language (and/or visualizations). Within this framework, every word outsourced to a generative model is a missed opportunity to strengthen my own thinking.

The analogy of "bringing a forklift to the weight room" has been an effective way to communicate this principle to students. If you don't want to strengthen your thinking, then by all means, go ahead and have LLMs do your writing. And I have to ask, if you don't want to strengthen your thinking, what made you decide to enroll in school?

There does seem to be a difference between outsourcing a thought process and outsourcing a building process. In the case of my website, for example, I have gone back and forth between managing it on my own versus outsourcing it (Ghost, Wordpress, etc.). Writing html is time-consuming. But when I let Ghost handle all the html, I lose some control. It also costs more money, and creates a dependency on their hosting services.

I wanted the best of multiple worlds:

• Quick editing
• File hosting
• Customizable/controllable html
• Fast and lightweight

It's only a little personal website, but ultimately, all technologies are political. So I sketched out my idea for a new website and fired up Claude code. I checked all the outputted html, performed some tests, and kept track of my prompts. For the most part, it felt like working one layer of abstraction above existing high-level programming languages -- for example, I still used html vocabulary (e.g. div, span) to guide the output.

Why all the fuss? Who cares if a website is vibe-coded? Well I do, and if you are reading this, you probably do as well. We are the working-class nerds! We know it's not just about code and control, it's about freedom! And as Eliza Ludd wrote, "curs’d his the man that even lifts a straw against the sacred cause of Liberty." Dare I say, we are the Neo-luddites?

As you may have heard, Claude code is pretty capable. It made the website renovations pretty quick, and eliminated a lot of time that I would have spent searching online for different html and css snippets. Since it is my website, I do feel responsible for the code that is published -- understanding it well enough to maintain it and adjust it without an LLM.

My main hesitation at this point is that Anthropic kind of sucks. They're running on billions of dollars from Amazon and ruthlessly stealing intellectual property wherever they can. This is not to mention their violent partnerships with Palantir and the U.S. military. And don't forget about all the unnecessary resource-hungry data centers!

So, as of writing, I am exploring ollama launch in order to hopefully "get off" Anthropic as well. The struggle continues.

Thanks for reading, fellow nerd!