Happy Thanksgiving break to all my friends, comrades, coders, and luddites. Thankful for all of you readers fighting for the user out there. With that, here’s a special Black Friday edition of Blood in the Machine.
I finally cancelled Spotify. I’d been meaning to do this forever, and frankly I’m embarrassed it took me so long. Spotify has been driving down wages for artists far longer than the AI companies, reducing payouts for musicians over the years until most are now making a statistically meaningless amount from the platform; many estimates put the figure as low as $0.003 per stream. In 2024, Spotify stopped paying artists for songs that had fewer than 1,000 streams, despite the fact that 81% of musicians on the platform don’t cross that threshold.
Stories abound of successful artists with millions of monthly listeners who can’t afford to take a vacation, a break, or pay rent. The pop star Lily Allen says she makes more money selling pics of her feet on OnlyFans than she does from Spotify royalties. Meanwhile, Spotify just raked in nearly $700 million in quarterly profits. It’s rank exploitation. Don’t take it from me, take it from Bjork. Earlier this year, she succinctly described Spotify as “probably the worst thing that has happened to musicians,” thanks to how the company, and the streaming model it normalized, has so completely corroded artists’ incomes over the last decade or so.
Meanwhile, the company declines to label the AI songs that are overrunning the platform and even boosts them into Discover Weekly playlists, incentivizing their spread. Founder and CEO Daniel Ek used his Spotify fortune to invest in a lethal military tech startup, prompting the most recent round of artist boycotts from the platform. I could go on, but that will probably do—Spotify is everything that’s wrong with Silicon Valley’s engagement with culture and labor condensed into a single platform. Plus, the audio quality sucks.
Greetings from up on my high horse. I was just getting warmed up, too.
So why didn’t I go sooner? I justified staying by telling myself I’d use Bandcamp to buy the albums and songs I listened to a lot, which I did, while using Spotify for convenience. That, and the same reasons I still use Gmail: I felt locked in (all those saved songs and playlists) and that the costs of switching would be too high (I would surely lose access to countless songs by switching over). But I am here to tell you today that both of those counts are absolutely false.
I’ve spent the last few days nursing a cold, and figured that I’d use some of that time to test out Spotify alternatives and finally take the plunge. I am extremely glad I did; it’s been a minute since I’ve felt something approaching genuine delight in discovering a new tech service.
So, in time for Black Friday—some of the Spotify alternatives have specials and sales going right now—here’s the Complete BLOOD IN THE MACHINE Guide to Getting Off Spotify.
As always, this work is made possible by the percentage of readers who chip in $6 bucks a month (or $60 a year). It is for you, dear readers, that I spend my sick days listening to new Geese tracks, Prince reissues, and Black Sabbath remasters on various streaming platforms (and not at all for myself), and I very much appreciate your support. If you can afford to, and you find value in this work, please consider joining the ranks of the hammer wielders. Many thanks, and onwards.
First, download Bandcamp if you have not done so already. The best way to support artists is to purchase their music directly, and even if the company has been through some rough waters lately, Bandcamp remains the most convenient way to buy music online and keep it in one place.
Second, I would recommend considering what’s important to you in a streaming platform. Audio quality? Library depth? Fairest payments to artists? Recommendation algorithm? It not being owned by a vampiric billionaire or operated by an extractive tech monopoly?
While you’re thinking about that, let’s assuage those two concerns about switching that you likely share with me.
You are not locked into Spotify even a little bit
Your music library is much, much more portable than you think. It will cost you from $0-10 to have a service like Soundiiz automatically transfer your playlists from one service to another. It’s so easy. The platform that I wound up switching to had a deal that let me do this for free. The permissions Soundiiz asked for were not at all onerous, and it took like 5 minutes. Total. I had around 200 playlists, hundreds of albums and artists saved, and several thousand songs and albums favorited. It all ported over seamlessly.
If you’re worried about your meticulously compiled playlists, I would say: Do not at all be. They’ll port over fine. And nearly every single song will still be here, because:
Spotify’s library is not that much better than anyone else’s
This was honestly the biggest surprise. On most streaming services I tried, I found nearly everything I was looking for. That surprise could be due to a long outdated misconception I had, or because AI has fried my brain for the last two years, but I was fully expecting Spotify to at least beat the competition in song selection. But nope, not really. There just wasn’t much deviation, and when I transferred my playlists, I only lost at most a song or two on each. And I listen to some admittedly pretty weird stuff! I did not really expect to find, say, Fluisteraars albums everywhere; either the world has suddenly gotten into Dutch experimental black metal, or it’s simply become easier for artists to upload songs to more platforms.
Okay, so you can transfer streaming libraries nice and easy, and selection isn’t really going to be an issue. Before we break down the alternatives, let’s talk about artist royalties. Most companies publish statements about how they calculate payments, but very few publish the actual rates they pay (there’s one exception, which we’ll get to in a second here), so it’s up to independent auditors to try to track down real numbers. Most streamers also send the payments to the rightsholders, so for musicians on record labels, there are further cuts scooped out.
Benn Jordan, YouTube.
I found this YouTube segment from the independent musician Benn Jordan really interesting. Since he’s indy, he gets the payments directly, and he put together this breakdown of what the streamers paid him.
Okay, with all that in mind, here’s a dive into the streaming platforms I’ve tried, and my choice of best, most artist-friendly streamer.
Apple Music
Apple Music is the closest thing there is to a straight-up Spotify clone, from my survey, anyway. I used a free trial, and it was fine. It had (nearly) all the music I wanted, audio quality was fine, and the user interface is intuitive. If you use lots of Apple stuff you’ll know what to expect and you probably already have an Apple Music account. I was a little curious to see if the notoriously prude Apple would censor graphic album art like the gory stuff on Cannibal Corpse covers, but I can report here that it did not. Apple pays artists $.006 a stream, which, while quite actually twice as much as Spotify, still kind of sucks. Plus, Apple is Apple, and you’re still handing your music money to a tech giant run by Tim Cook, who was most recently spotted at the White House’s investment forum for Saudi Arabia.
Tidal
Tidal was pretty good, actually. Artist pay is $0.008, which is better than Apple, and it seemed to demonstrate a little genuine appreciation of music; there was a contest running on the home page for undiscovered artists when I logged on. The audio quality is noticeably better than Spotify and Apple, and the user interface is OK, even if it’s a little inert. However, Tidal is also the most volatile major music streamer. It has a wacky history; it started in Norway, was bought by Jay Z, who is still a minority stakeholder, and at one point was partially owned by the now defunct telecom Sprint. It was eventually bought by Block, formerly Square, the Jack Dorsey-led payment processing company, and laid off 10% of its staff in 2023. The company is apparently currently without a CEO, so it is presumably de facto being led by Dorsey, a man who once said of Elon Musk, “I trust his mission to extend the light of consciousness.” As a result, I do not trust Tidal.
YouTube Music
You have to work pretty hard to be worse than Spotify on all counts. In my admittedly brief tour of YouTube Music, audio quality was bad, it was unappealing to use, and it somehow pays artists even less than Spotify ($0.0027 per stream). Next.
Pandora
I was a little surprised that Pandora, a true mainstay of the 00s’ “music is on the internet now” novelty era, was still around. Alas, it also pays artists complete garbage ($0.0027), and since all the major streamers do the vibes-based playlist thing now too, I’m not really sure what Pandora’s value proposition is.
Deezer
The French streamer apparently pays a bit better than the pack ($0.007), and it seemed fine, too, foregrounding its offerings of Pandora-but-pure-mood playlists like ‘Flow’ and ‘Love’. It would be a suitable alternative, perhaps, if it wasn’t for what, in my opinion, is the hands-down best music streamer there is right now.
Qobuz
I spent about 20 minutes on the unfortunately named Qobuz, and I was sold.
First, the artist pay is the best in the biz (except, for some reason, for Peloton’s in-house streaming service). It’s $0.0138, which means it’s more than a penny a stream. Receiving Qobuz’s pay report, the indy musician and analyst Jordan says, “was the first time I looked at streaming royalties and felt fairly compensated.” (The company is also the first to publish its own average payout report, and make its figures public—it cites $0.01873 as its average. A thousand streams means the artist makes ~$19.)
Second, the audio quality is great. I listened to a little bit of everything to test it out, and especially after the flat, compressed sound of Spotify, I didn’t want to stop. It’s got an icon for when albums are available in 24-bit/96 kHz, high resolution format, and it kills. On Qobuz, the Replacements’ Let It Be is raw and crackling, the Flaming Lips’ Soft Bulletin is lush and beautifully layered. Master of Puppets sounds huge. The (great) new Robyn track Dopamine is crisp and prismatic. Qobuz’s album of the week was a remastered drop of Prince’s Around the World in a Day; it’s majestic. Also that new Geese album everybody likes really is good and sounds great steaming, too.
Third, and the thing that really resonated with me the most, perhaps, was that the people who run Qobuz actually seem to… like music? On the app, along with its Discover and Playlists tab, there’s a Magazine tab with interviews and features. I’m not sure I’ll spend a ton of time reading them, but after years of the algorithmic sludge and pop-up payola recommendations of Spotify, it was refreshing to read thoughtful articles celebrating, say, the post-rock icons Tortoise, Quincy Jones’ catalog, and a Kantō speaker system. I also love how Qobuz lists the record label along with the song title, artist, and it’s clickable, so you can go right to the label to find other bands and so on. There’s *also* no streaming number listed, perhaps because it would be rather low given the relative size of Qobuz’s platform, but I found it refreshing as there is therefore no way to snap-evaluate a song based on its stream count.
This was a nice way to encounter the artists who’ve bailed on Spotify, or, those, like Joanna Newsom, who’ve never relented and let Spotify stream their music in the first place. I bought some of Newsom’s albums on Bandcamp way back when, but I only realized now just how little I’ve listened to them because she wasn’t on Spotify, and how shitty that is. I gave King Gizzard and the Lizard Wizard and Massive Attack a spin, too, since they and dozens others have ditched Spotify over Ek’s investments in Helsing, a German military tech and drone company.
So yeah, I’m having a great time with Qobuz, both because the product is genuinely better than Spotify, and because it feels like sweet release to be free of that shitpit. There are shortcomings, of course—it’s not Siri-compatible, so it won’t work with CarPlay hands-free. It takes longer to load songs and the UI is clumsy on mobile. It isn’t as seamless as some other apps. (I also listened mostly on my laptop, as opposed to mobile, and that’s worked great for me.) But those are minor things. More importantly, let’s not have any illusions here: Qobuz, while operated by music biz folks who seem driven (or see an opportunity) to put artists and music at the forefront and de-emphasize the algorithm, is owned by a multimedia conglomerate SA Xandrie, and has completed multiple $10 million plus VC rounds. It’s not some scrappy artists’ collective that can be counted on to keep musicians’ pay decent and AI sidelined. But for now, Qobuzz is great.
And there you have it: An official Blood in the Machine endorsement for ditching Spotify, and trying out a more ethical, better-sounding alternative.
Other BITM-approved gifts for Black Friday and beyond
Friend of the blood, artist, AI critic, and fellow ludd Molly Crabapple has a beautiful new set of illustrated cards out in a set called Can You See The New World Through The Teargas? from OR Books.
The designer Bart Fish has been publishing a great series of strikingly visualized critiques of AI at Power Tools. He’s publishing them as a zine, too; you can order one for a pay-what-you-can donation here. They look amazing:
Finally, the e-reader edition of Blood in the Machine: The Book is, I’m told, on sale at Amazon. Under normal circumstances, that’s not exactly the ideal place to get it, but seeing as how Amazon is certainly taking a loss on the sale here, so by all means, get it for cheap.
OK! That’s it folks. Have a nice holiday weekend, keep that hammer high, and see you all again soon.
In 1914 Joseph Pilates was sitting in a World War I-era internment camp watching his fellow inmates waste away when he had an idea. As he later recounted to a journalist, the skinny prison courtyard cats kept spry and limber, stretching and moving even as they starved — could the German citizens interned in the camp at the Isle of Man do something similar? Pilates developed what would later become his namesake: a regimen of repetitive, controlled body movements with minimal equipment focused on flexibility and strength. Before it became synonymous with luxury workout leggings and YouTube videos, Pilates the exercise was a lifeline, born literally in a cage.
The practice has come a long way since, especially in the last few years. An industry report found Pilates is the fastest-growing mode of exercise, and subscription service ClassPass named it the most popular class type of 2024. Along with the boom, there is a new wave of cultural cachet — and sometimes ridicule — that comes with practicing Pilates, the same way yoga strayed from its roots to become part of the crunchy wellness culture of the 2010s. On TikTok, there is a formula for much of the most popular #pilates content: Influencers take turns posting from packed classes in mirrored, minimalist studios; smooth, algorithm-bait pop songs soundtrack clips of slender women, sweat dripping from every inch of their bodies. Pilates has ascended beyond being simply an exercise modality: It is a key part of a certain idealized life of leisure, affluence, beauty, and, importantly, virality. There is often matcha involved.
Antoinette Hocbo had gone to a Pilates class years ago, in her early 20s, and swore off it immediately. “I hated it. It was awful. And I was like, ‘I’m never doing Pilates again,’” Hocbo says. Then, this summer, the TikTok content started.
Hocbo had been looking to incorporate more exercise in her life: She had used ChatGPT to come up with workout plans and had searched on TikTok for fitness content. Soon enough, her algorithmic recommendation-fueled For You page was filled with related videos. Hocbo recalls the exact moment she was influenced: A content creator popped up, extolling the virtues of a specific Pilates instructor (she followed neither person). The endorsements in the video were glowing — this instructor’s videos will change your life. Beach weather was coming, and much of the content fed off the timeliness of getting a “summer body.” Hocbo decided to give Pilates another try and purchased the instructor’s online class ($199), along with a Pilates ball and blocks.
It wasn’t the only hobby she picked up because of TikTok. During one stretch, her For You page was a stream of artists using the drawing app Procreate. Though Hocbo had studied illustration, she had never tried it that way, and she was drawn to the idea of rebuilding with her creative practice. “I was getting lots of videos of people doing animation on their iPads, and I was like, ‘Do I need to buy an iPad?’” She picked up a used one on Facebook Marketplace.
Though she previously didn’t care much for makeup, her algorithm delivered influencers’ GRWM (Get Ready With Me) videos, where they meticulously narrate their makeup routines, complete with every product they use; pretty soon Hocbo found herself buying the same arsenal of minimal no-makeup-makeup concealers and foundations. One time-consuming product that she was successfully influenced into buying is a lip stain that she must paint on, wait 10 minutes for the color to tint her lips, and then peel off. She saw it on TikTok Shop.
“I get mad every time I use it,” Hocbo says. “These TikTok people convinced me this is the way, but now I have this stupid part of my makeup routine because of them.”
Hocbo knows all the tricks marketers and brands use to build consumerist desire — she used to work in marketing, chipping away at shoppers’ willpower until they caved and hit “check out.” Introducing new artistic practices in Hocbo’s life brought joy, but she also had a nagging feeling about how she got there in the first place.
“I always have this tension that I’m experiencing,” she says. “Is this a choice that I’m making for myself, or am I being influenced by this app or these influencers?”
Commerce has long been central to social media; as long as ads keep the lights on at Meta, TikTok, and YouTube, we will all be pressured to buy, buy, buy. Instagram was a mall even before #TikTokMadeMeBuyIt and Pinterest became an “AI-enabled shopping assistant.” The influencer industry — which Goldman Sachs has predicted will grow to nearly half a trillion dollars by 2027 — has snowballed into a possible side hustle for anyone with access to a phone. There’s a handful of MrBeasts and Alix Earles at the top and an untold number of micro-influencers hawking goods and services at the bottom. For audiences, it means we have spent the better part of a decade living within a 24/7 digital infomercial, with social media — sponsored content and organic posts alike — resembling not much more than a buying guide, a catalog of unabashed and conspicuous consumption. Some audience members find themselves in deep debt or describe their behavior as a full-blown shopping addiction; others have developed careful strategies in an effort to limit their consumption. We have never been so aware of all the things there are to purchase, and the frictionlessness of shopping apps disguised as social media has created an army of voracious buyers. What has this abundance done to us?
I report on the individuals, platforms, and brands that have successfully created a whole new industry we call “influencing.” But in order for all of that to stay afloat, influencers need people on the other side of the equation — they need the influenced. To understand how we got here, I talked with influencers and the people they successfully convinced about what it’s like online when the pressure to buy is suffocating.
Hocbo was able to stick with the Pilates program for a while through the summer, and says the regimen helped alleviate some lower back pain she had. But now Pilates has fallen by the wayside of her TikTok-driven hobby highway. The gear she bought haunts her home, unused.
“It’s just sitting there, mocking me,” she says.
Eventually, the Pilates content on her TikTok feed that was once relentless stopped, too — as if the algorithm knew it had done its job.
The impulse to shop is not exactly a secret — there’s often a resigned self-awareness to it. In a video viewed 1.5 million times, a woman stitches together clips of herself from random moments in her daily life. With a deadpan voice, and Radiohead’s “No Surprises” twinkling in the background, she recites highly specific products like she’s filling out a Mad Libs page: Chan Luu crystal toe ring. Arc’teryx hiking shoes. Vintage hoodie. “This is just the last 48 hours, mind you,” the caption reads.
This kind of video has become a mini-trend, with the idea being that the mere utterance of a temptation might soothe the part of your brain that wants to buy the item.
Christina Mychaskiw has always loved shopping, especially for clothing. Fresh out of pharmacy school in 2014 with $120,000 of student debt, she was making more money than she ever had in her life, yet didn’t have the skills to manage it, she admits. She stressed about her finances, then shopped to relieve the stress. Dropping hundreds of dollars every weekend became a new normal, and inspiration was everywhere: in stores, on Pinterest, on influencers’ affiliate link pages.
Mychaskiw’s rock bottom was when she pulled the trigger on a pair of studded Chloé ankle boots that cost more than her monthly rent after she saw them in a TV show. She knew she could return them — she was broke, with a negative net worth — but kept them anyway.
By the winter of 2019, Mychaskiw realized she was “financially fucking [herself] over.” She started a so-called no-buy challenge, swearing off shopping for around seven months.
Since then, Mychaskiw has found a happy medium for her shopping, where she doesn’t completely deprive herself but also proceeds cautiously. She makes wishlists of items she sees that she will come back to later, and has worked to develop hobbies not tied to accumulating things. But the pull is still there at times, especially after a period of scrolling. She can usually sense when she’s hit the danger zone.
“If I see something that I’ve scrolled through and I want to buy it then and there, it’s probably not a genuine want,” she says. “It’s just that buildup, that dopamine, that threshold that you have to hit until you need to put that energy somewhere.”
We see so much marketing material that in certain subcultures online it is not just common but the expectation. In traditional marketing, it was understood that brands had to expose consumers to their message three times before they actually engaged with it, like going physically to a store to buy a product. In the age of social media and algorithmic overload, that number is now seven, says Mara Einstein, a marketing-professional-turned-critic and author of the book Hoodwinked: How Marketers Use the Same Tactics as Cults. For one, the vastness of the internet has allowed for the number of available products to bloat beyond imagination — there are simply too many things. But how we learn about products has changed drastically as well; as media has fragmented to a million sites, feeds, screens, and algorithms, so too has the advertising we see. There is no one TV commercial a quarter of households are seeing, then telling their friends about. Instead we see a digital display ad here, an influencer’s video there.
Her debt swelled to over $50,000 before she vowed to dig herself out of the hole.
“You may be finding out information from people and so on, but you’re increasingly spending time in a space where you’re constantly being bombarded by sales messages,” Einstein says. Influencers know how to stay on message, constantly priming viewers to give in and buy something.
Being influenced is nothing new, of course. But the short- and mid-form video format creates a new type of intimacy and allure, especially if you are already looking for something to buy. It’s hard to argue with a sales pitch when you are watching someone in their home actually using the product they are trying to sell you.
The content doesn’t even have to be explicitly promotional: I recall a video I made last year about my reporting being used without credit by content creators. My frustration had hit a breaking point, so I recorded a selfie-style TikTok complaining about the contemporary media ecosystem. Only my head and a portion of my shoulders were in the video, but someone wanted to know where my blouse was from.
TikTok itself has only bolstered the idea that every piece of content is an opportunity to consume. Through TikTok Shop, anyone can become a digital salesperson. In much crueler, more tasteless examples, TikTok has added shopping prompts to videos coming out of Gaza: A woman in a head covering becomes a promotion for similar-looking garments with headscarves. A bespectacled Israeli activist protesting their government’s besiegement is a billboard for a pair of glasses.
For Elysia Berman, the covid-19 pandemic changed everything. Before lockdown, there was no shortage of inspiration from real people out in New York City or at stylish fashion industry events.
“Then the lockdown happened. I was stuck inside, I wasn’t really getting inspiration from anyone in my life,” Berman says. “I would just sit and look at my phone and buy what it told me to.” She purchased a green top she saw influencers wearing even though she almost exclusively wears black, New Balance shoes even though she doesn’t wear sneakers. Berman scooped up a Nap Dress from the brand Hill House Home — a flowy, nightgown-like dress that went viral during covid. “I looked like a Victorian child for half the pandemic,” she says.
Many of her purchases proved to have a short shelf life: “It was the most boring shit that I never wore, that I got rid of within six months,” she recounts. Shopping became an outlet for other frustrations in Berman’s life, an escape from a kind of fugue state she found herself in: She wasn’t yet sure who she really was, or what her hobbies were. She felt underpaid and undervalued. Friendships felt tenuous.
“With every purchase that I was making, I was trying to take one more step closer to being the person that I wanted to become,” she says. “You see people that you find aspirational — you want to feel one step closer to them. You want to kind of gain proximity to their lifestyle, and a lot of times what people do is they make purchases based on what that person recommends.”
By 2023, her shopping habits were impossible to ignore: She had four credit cards and four buy-now-pay-later services that each carried balances of between $4,000 and $8,000. She also had several loans. Her debt swelled to over $50,000 before she vowed to dig herself out of the hole.
Naturally, Berman has made videos warning people about the dangers of accumulating credit card debt; she shows viewers the “ridiculous” things she bought in her shopaholic phase — designer sunglasses! $200 headbands! 22 clear lip products! — as a cautionary tale. But even that can serve as an invitation to shop.
“I’ve had people tell me that they’ve purchased things just because I’m wearing them and I’m actively telling people to buy less,” she says. “I’m actively telling people the perils of shopping too much.” Berman recounts a video she saw of a user crying that their mother had died; some comments asked where their mascara was from because it wasn’t running down their face.
In addition to her day job as a pharmacist, Mychaskiw makes content about mindful spending habits without villainizing all shopping — there can be pleasure and even joy in consumption. She sometimes takes sponsored brand deals, but says she only promotes products she would actually buy herself and that she has personally vetted.
“I try to balance giving people the tools to decide when it’s okay to say yes to themselves, because that’s something that I had to learn, too,” she says.
It’s easy to blame the influencers for all of this — and many do, regularly, like clockwork. The most recent discourse cycle, in late September, was kicked off by a TikTok video with 390,000 views and arguments that stretched on for weeks.
“These influencers make way too much fucking money,” the video begins. “You’re just getting paid to sell people shit they don’t fucking need. It’s literally just overconsumption … You’re perpetuating this cycle that’s really keeping us trapped.”
Content creators are admittedly a perfect target for the general rage many of us carry around. Many of them seem unencumbered by the endless horrors of the world, with daily routines that include blocks of time for “warm water” and to-do lists with “plan out mocktails for the new year.” Their digital presence exists suspended in time, where there is always something new to recommend, packages of shiny new things waiting for them, and a willing audience that completes the positive feedback loop. Wouldn’t it be nice — as people are in line at food banks, fighting for a precious few job listings, and snatched off streets by masked agents — to sit in your home and talk to yourself for a living?
But the draw of the influencer is powerful; even if you cannot become her, you can own the same things she does. For Antoinette Hocbo, who picked up hobbies via TikTok, the characters she encounters on her For You page seem effortlessly cool. They have an eye for design, they’re interested in the arts, they drink wine. You buy into the person first, and eventually — hopefully — you buy the stuff, too.
“[There’s] the whole idea of parasocial relationships,” Einstein, the marketing expert, says. “If somebody has gotten to the point where they’re spending that much time online with someone, they’re vested in what that person has to say.” The feeling of intimacy is physical: When followers watch their favorite TikToker, they are literally holding them in the palm of their hand.
TikTok’s rise during the pandemic created a new playing field — ordinary people with no prior public presence could be catapulted overnight into the next viral character. Often, people stumble into a following accidentally: They’re funny or beautiful or well dressed or simply got lucky one day.
“With TikTok, a lot of creators describe it as almost like a gambling addiction where they would post something out there and they would see this viral boon,” Brooke Erin Duffy, an associate professor at Cornell University who studies the influencer industry, says. The pandemic upended work and labor, opening the floodgates for a new kind of job that previously was reserved only for the rich, well-connected, or very lucky.
“You have this platform that is enabling people to go viral fleetingly, and you have profound changes in the nature of work,” Duffy says. “All of a sudden you see this huge uptick in TikTok creators.”
If people hate influencers as much as they say, they have a funny way of showing it.
TikTok made going viral a possibility for a whole new slate of people. Now the hard part is how to keep things rolling when it happens to you. Most of the platforms themselves do not pay much for views, but brands eager to partner with buzzy people do. Creators often talk about their work in terms of self-discovery or self-actualization: This is who I want to be online, and these are the products and tips I truly, honestly want to share.
The tension comes then with the “very real commercial realities of playing to an audience, bowing to commercial sponsorships if you were lucky enough to have them,” Duffy says. “And then the new dimension, which doesn’t have the same precursors in legacy media, which is playing to the algorithm.” A 2024 Pew Research Center survey found that 62 percent of adults on TikTok are there to find product reviews and recommendations — especially young women.
“Dumb luck” is how Connor Chase explains his 180,000 followers on TikTok — at first he was simply documenting his outfits, posting here and there with few expectations. One day he posted a video showing off a new jacket he got; all of a sudden he had 50,000 new followers.
His followers tune in for fashion and menswear content, and he describes them as a dedicated, loyal audience. But many fashion influencers go through a boom-and-bust cycle, Chase says: You post your cool outfit, blow up and get a lot of views, and then there’s the comedown. Chase knows one way he could juice his views and grow his platform even more.
“I know for a fact if I was to just start buying designer, new stuff, I know my views would skyrocket for sure,” he says. The only creators who can maintain constant growth and relevance are those that constantly share new acquisitions, new designer products, new things for viewers to lust over.
“That is what people want to see. They want to see the new stuff.” Chase’s videos where he shares items in his closet tend to perform well, as does content showing off new purchases. Viewer engagement is the metric that content creators live and die by — followers watching hours and hours of unboxings and hauls ensures influencers will just keep making more of that. If people hate influencers as much as they say, they have a funny way of showing it.
In his Gilded Age-era book The Theory of the Leisure Class, Thorstein Veblen coined the term “conspicuous consumption” to describe the public display and flaunting of luxury items, theorizing that rich people’s indulgences were not just fancy things to buy but were a way to signal social status to those around them. Veblen observed the phenomenon of physical goods serving as overt demonstrations of social power or affluence more than a century ago, and that dynamic is now alive and well on our smartphone screens. Influencers’ promotion of products isn’t just about the physical item and the social class that a fancy camera or a new handbag projects — the fact that they are getting something at all signals that they’ve ascended to a new level for others to aspire to.
Social media sells users the lie of unfettered consumption the same way it sells wannabe influencers the myth of infinite opportunities — that you, too, can join this pioneering class of entrepreneurs following their divine destiny to make bank online. The rhetoric that anyone can do this job is peddled by influencers and marketing firms, naturally, but also by tech companies that are fighting for content and creators themselves. (“There’s no ‘right’ way to use Instagram,” the company, which has almost completely undercut photos for video Reels, says on its creator hub. “Discover the many ways you can make it your own.”) The reality is that there is very little data about the influencer class that is not published by entities with a vested interest in the space, Duffy says.
“There’s the top creators who are earning astounding sums, but most people, according to every survey I’ve seen, are not making enough to earn a living wage,” Duffy says.
Brand deals, especially after the pandemic, are getting harder to come by, according to Duffy. “When you do [get deals], you’re cobbling together money from various income streams. There are some people who are getting money thrown at them for doing seemingly very little work, but for your everyday career creator, it’s incredibly unstable.” Chase says many brands and marketing agencies won’t work with the same influencer twice, so massive is their pool of options; Chase doesn’t expect this job to last forever, but hopes to leverage his internet presence to launch his own clothing company.
Project Pan, as a concept, is both clever and strange. For years, a community of people organized largely on the internet have committed themselves to finishing their beauty and personal care products — the name coming from your promise to hit the bottom of the pan that holds your blush, for example. It’s smart for the way it gamifies something people struggle with. (Who among us doesn’t have half-used bottles of soap or barely touched tubes of lipstick?) It’s also deeply revealing: These products are meant to be used, and we collectively are so bad at finishing them off that we need a little game to make it happen. Off the top of my head I can confidently say that I’ve never once “panned” a compact of blush; I have expensive tubes of red lipstick that didn’t end up being my color, but that I can’t bear to throw out; and I have four bottles of sunscreen that crowd my cabinet, waiting for the summer they’re finally used up. There are many more products that I could — should — Project Pan that I’ve forgotten I even own.
Cassandra Silva, on the other hand, knows exactly what she has. She knows, for example, that she spent $2,857.98 AUD on makeup in 2024 and panned products totaling $1,654.13. She owns eight eyeliners, but her ideal number would be four. In 2023 she panned seven mascaras, 11 colored lip products, and one blush, among many others, all lined up in a photo of the totally empty containers that show her progress. She keeps all this data in a giant spreadsheet that she shares with me after we talk, and as I scroll through it, I realize I have never seen an eyeshadow palette where every color is completely empty.
“Compared to beauty YouTube, it’s not insane insane, but it’s still more than any one human could ever reasonably use,” Silva says of her inventory.
She watches beauty YouTube channels, but needs to be careful about what she consumes: She tries to stay away from content showing off hauls, new releases, or the ever-tempting limited-edition holiday releases.
“I am as conscious as I can be for a makeup addict,” Silva says. “I try, and I am freaking susceptible. It’s so bad.” Recently, a palette of neutral eyeshadows hounded her Instagram feed — she caved and bought it, only to be thoroughly disappointed when it arrived. As a panner, Silva will be stuck with it for years until it’s finished.
Chessie Domrongchai used to make the kind of content that Silva perhaps would steer clear of — she was the one tempting makeup lovers with all of these products. As a beauty YouTuber, Domrongchai shared in-depth product review videos for brands like the once-buzzy direct-to-consumer brand Glossier and tested fistfuls of lip glosses in subtly different shades for her 40,000 subscribers. She shared new releases, compared similar products from different brands, and recommended items for upcoming sales. In a 2019 video, she walks viewers through her pinky-brown nude lipstick collection — 15 shades, not including lip glosses and liquid lipsticks. She followed makeup brands and watched other YouTubers, accumulating more and more products to explore ($10,000, she says, feels like a conservative estimate of the value of her collection at its peak). In makeup, Domrongchai found self-expression, creativity, and community.
Until one day in 2022, when a switch went off in her head.
“I started to view a lot of the overconsumption that I was seeing online as kind of disgusting and wrong, and I recognized a lot of the way that I showed up on the internet was to overconsume,” Domrongchai says. Not only that, but she felt her online presence also influenced viewers to keep buying more and more.
“These are just regular people that are just now stuck with the burden of their overconsumption,” she says. But as a content creator, it was hard to be part of the beauty space without having a constant parade of new products.
In recent months, Domrongchai has developed a new routine for the many products littering her home. One by one, she meticulously peels off stickers and labels: from shampoo and olive oil bottles, from dish soap dispensers and face wash. Using a mix of baking soda, mineral oil, and rubbing alcohol, she goes to town on brand names printed on the packaging of eyeshadow palettes and lipsticks, scrubbing away their origins and the millions of dollars of marketing that went into them — arguably why they are in Domrongchai’s house to begin with. The result is shelves and countertops full of bare bottles and tubes and pumps filled with product but stripped of just about everything else. Watching her videos, I’m slightly horrified at my own ability to recognize the specific products even without all the labeling, the colors and shapes of bottles acting like an afterimage of a CeraVe cleanser.
“Of course I’m going to buy the face cleanser that keeps my skin clear, but I don’t need it to continue to market to me in my own home,” Domrongchai says. “In the past I had three different [lotions] and all of their labels and their marketing on these products … They’re all kind of yelling at you trying to convince you to use it. They’re kind of [in] competition with each other.” In other words, it felt like a social media feed.
For some panners, finishing a product can elicit the same rush that buying something new does — that same dopamine rush of hitting “place order” creeps in when you hit that pan. Then you post it online for other panners to see, adding to the thrill. Finishing products becomes a task to complete, just like shopping is.
“What it can do — which I don’t love to admit to — is you’ll put more blush on than you would,” Silva says. “You just slather it on.” Silva shows me her spreadsheet page from 2024 showing colored lip products she used up: 23. Silva estimates that the average person finishes maybe one lipstick a year. In order to pan that many products, she was reapplying them 15 to 20 times a day, she says. Sometimes Silva wonders if she should ditch panning, too, like she did consumption-focused beauty spaces.
“When you first get into it, it’s so helpful, and you really get that community and you can turn some products over. Then the longer that you’re in the panning community, it’s like, all right, now panning is a problem,” she laughs. “Now I’ve taken all the problems I had with makeup consumption and translated them into late-stage panning. It’s like late-stage capitalism.”
Social media isn’t just filled with ideas for things to buy — increasingly, there are moments where reactionary anti-consumer trends bubble up, too. Project Pan has been around for years but had a resurgence earlier in 2025; “loud budgeting” emerged as a foil to the so-called quiet luxury trend; and the act of owning a normal amount of things (one fancy water bottle, clothes that are several years old) has been rebranded as “underconsumption core.” For a second it seemed like “de-influencing” had some legs — until inevitably the de-influencers started promoting other things to buy instead. Even the mechanisms for buying less can prompt people into shopping more. When there is shopping fatigue, things eventually swing back. Duffy says creators often describe how TikTok rewards educational content, but that even that genre has consumption at its center.
“There’s always been advertising packaged as advice,” Duffy says, pointing back to things like women’s magazines. “But in the context of TikTok, you had so many people that were sharing their tips and tricks on life with consumer purchases at the center.”
Pilates, in its most stripped-down form, requires only a mat and your body weight — no fancy equipment, no coordinating workout clothing, and no membership to an invite-only studio. But as Pilates has exploded in popularity, the practice and the idealized life that includes it has a new protagonist: the Pilates Princess. She is clean, thin, ritualistic, and has nothing but time for herself — a Patrick Bateman for people with Pinterest manifestation boards. She is also, importantly, a key consumer group: for fashion brands releasing entire collections of garments named after her, for Spotify Wrapped, which included a “Pink Pilates Princess” listener category in 2024, and for tech gadgets like Oura rings or Apple AirPods Max, which are regularly featured in Pilates Princess content. There are plenty of people who do Pilates without the consumerism, of course — but the practice that was originally developed in a cell is now in a new, more shoppable cage of its own.
It’s hard to imagine what a social network not centered around shopping would even look like (some apps, like the now-defunct Flip, did away with the “social” pretenses and just made a feed full of ads). And even if there was an intentional effort to veer away from unmitigated consumption, it’s not clear if users — who in the same breath decry these habits — would care. Shopping addiction or compulsive buying is not included in the Diagnostic and Statistical Manual of Mental Disorders as an official diagnosis, but there is research from around the world about its prevalence, including what role game-like social media apps play.
Overconsumption didn’t begin with social media or TikTok, but it did get a turbocharge from these platforms with a vested interest in our shopping habits. Creators, brands, and advertisers have known it for a while, but now it feels unignorable from the audience’s perspective: There are vanishingly few ways to be online that don’t involve you becoming a billboard. The internet once promised the democratization of information, power, and free expression. As advertiser-first algorithms take hold, careful not to promote anything that might stray from brand safety, we are mostly left with a very long shopping list. The companies making money on all of this have made content creation seem scintillating: walls of PR boxes, flexible schedules, an alternative to the concept of work that was upended during the pandemic and still hasn’t recovered. The idea that anyone can create actually means that anyone can promote.
The day before we spoke in May 2025, Elysia Berman paid off the last dollar on her $50,000 of debt. TikTok, ironically, helped her cross the finish line: Berman made thousands as part of the platform’s creator fund, which she redirected to pay down her debt.
“It’s TikTok’s fault that I’ve gotten in this mess, so I like that TikTok’s helping me pay it off,” she says.
Berman says she is careful about what she posts, mindful that she too could be influencing someone else. Her history with problematic shopping is an undercurrent through much of her content — she offers tips for people who want to find their personal style without buying impulsively. But some of her videos are undoubtedly centered around consumption: what to buy in the Sephora sale, unique fragrance recommendations, seven of her favorite black boots.
Then again, viewers eager for an impulse purchase will see a shopping list even where there isn’t one. In a video from September, Berman talks about breaking free from the burdens of compulsive shopping: not worrying about how she’ll pay for something, not obsessively tracking packages she’s expecting, going on a trip without worrying about money. Berman was selling something intangible — a healthy relationship with your possessions — and many thanked her for it. But even that video had real commercial potential. A few comments asked what eyeshadow she was wearing; another asked where her necklace was from. In her quest to inspire people to break free from mass consumption, she too had become a billboard.
I’ve never liked the philosophy of “put an icon in every menu item by default”.
Google Sheets, for example, does this. Go to “File” or “Edit” or “View” and you’ll see a menu with a list of options, every single one having an icon (same thing with the right-click context menu).
It’s extra noise to me. It’s not that I think menu items should never have icons. I think they can be incredibly useful (more on that below). It’s more that I don’t like the idea of “give each menu item an icon” being the default approach.
This posture lends itself to a practice where designers have an attitude of “I need an icon to fill up this space” instead of an attitude of “Does the addition of a icon here, and the cognitive load of parsing and understanding it, help or hurt how someone would use this menu system?”
The former doesn’t require thinking. It’s just templating — they all have icons, so we need to put something there. The latter requires care and thoughtfulness for each use case and its context.
To defend my point, one of the examples I always pointed to was macOS. For the longest time, Apple’s OS-level menus seemed to avoid this default approach of sticking icons in every menu item.
That is, until macOS Tahoe shipped.
Menus in macOS Tahoe
Tahoe now has icons in menus everywhere. For example, here’s the Apple menu:
Let’s look at others. As I’m writing this I have Safari open. Let’s look at the “Safari” menu:
Hmm. Interesting. Ok so we’ve got an icon for like half the menu items. I wonder why some get icons and others don’t?
For example, the “Settings” menu item (third from the top) has an icon. But the other item in its grouping “Privacy Report” does not. I wonder why? Especially when Safari has an icon for Privacy report, like if you go to customize the toolbar you’ll see it:
Hmm. Who knows? Let’s keep going.
Let’s look at the "File" menu in Safari:
Some groupings have icons and get inset, while other groupings don’t have icons and don’t get inset. Interesting…again I wonder what the rationale is here? How do you choose? It’s not clear to me.
Let’s keep going. Let’s go to the "View" menu:
Oh boy, now we’re really in it. Some of these menu items have the notion of a toggle (indicated by the checkmark) so now you’ve got all kinds of alignment things to deal with. The visual symbols are doubling-up when there’s a toggle and an icon.
The “View” menu in Mail is a similar mix of:
Text
Text + toggles
Text + icons
Text + icons + toggles
You know what would be a fun game? Get a bunch of people in a room, show them menus where the textual labels are gone, and see who can get the most right.
But I digress.
In so many of these cases, I honestly can’t intuit why some menus have icons and others do not. What are so many of these icons affording me at the cost of extra visual and cognitive parsing? I don’t know.
To be fair, there are some menus where these visual symbols are incredibly useful. Take this menu from Finder:
The visual depiction of how those are going to align is actually incredibly useful because it’s way easier for my brain to parse the symbol and understand where the window is going to go than it is to read the text and imagine in my head what “Top Left” or “Bottom & Top” or “Quarters” will mean. But a visual symbol? I instantly get it!
Those are good icons in menus. I like those.
Apple Abandons Its Own Guidance
What I find really interesting about this change on Apple’s part is how it seemingly goes against their own previous human interface guidelines (as pointed out to me by Peter Gassner).
They have an entire section in their 2005 guidelines titled “Using Symbols in Menus”:
See what it says?
There are a few standard symbols you can use to indicate additional information in menus…Don’t use other, arbitrary symbols in menus, because they add visual clutter and may confuse people.
Confused people. That’s me.
They even have an example of what not to do and guess what it looks like? A menu in macOS Tahoe.
Conclusion
It’s pretty obvious how I feel. I’m tired of all this visual noise in my menus.
And now that Apple has seemingly thrown in with the “stick an icon in every menu by default” crowd, it’s harder than ever for me to convince people otherwise. To persuade, “Hey, unless you can articulate a really good reason to add this, maybe our default posture should be no icons in menus?”
So I guess this is the world I live in now. Icons in menus. Icons in menus everywhere.
Look, I don’t know if AI is gonna kill us or make us all rich or whatever, but I do know we’ve got the wrong metaphor.
We want to understand these things as people. When you type a question to ChatGPT and it types back the answer in complete sentences, it feels like there must be a little guy in there doing the typing. We get this vivid sense of “it’s alive!!”, and we activate all of the mental faculties we evolved to deal with fellow humans: theory of mind, attribution, impression management, stereotyping, cheater detection, etc.
We can’t help it; humans are hopeless anthropomorphizers. When it comes to perceiving personhood, we’re so trigger-happy that we can see the Virgin Mary in a grilled cheese sandwich:
Apparently, this served us well in our evolutionary history—maybe it’s so important not to mistake people for things that we err on the side of mistaking things for people.1 This is probably why we’re so willing to explain strange occurrences by appealing to fantastical creatures with minds and intentions: everybody in town is getting sick because of WITCHES, you can’t see the sun right now because A WOLF ATE IT, the volcano erupted because GOD IS MAD. People who experience sleep paralysis sometimes hallucinate a demon-like creature sitting on their chest, and one explanation is that the subconscious mind is trying to understand why the body can’t move, and instead of coming up with “I’m still in REM sleep so there’s not enough acetylcholine in my brain to activate my primary motor cortex”, it comes up with “BIG DEMON ON TOP OF ME”.
This is why the past three years have been so confusing—the little guy inside the AI keeps dumbfounding us by doing things that a human wouldn’t do. Why does he make up citations when he does my social studies homework? How come he can beat me at Go but he can’t tell me how many “r”s are in the word “strawberry”? Why is he telling me to put glue on my pizza?2
Trying to understand LLMs by using the rules of human psychology is like trying to understand a game of Scrabble by using the rules of Pictionary. These things don’t act like people because they aren’t people. I don’t mean that in the deflationary way that the AI naysayers mean it. They think denying humanity to the machines is a well-deserved insult; I think it’s just an accurate description.3 As long we try to apply our person perception to artificial intelligence, we’ll keep being surprised and befuddled.
We are in dire need of a better metaphor. Here’s my suggestion: instead of seeing AI as a sort of silicon homunculus, we should see it as a bag of words.
WHAT’S IN THE BAG
An AI is a bag that contains basically all words ever written, at least the ones that could be scraped off the internet or scanned out of a book. When users send words into the bag, it sends back the most relevant words it has. There are so many words in the bag that the most relevant ones are often correct and helpful, and AI companies secretly add invisible words to your queries to make this even more likely.
This is an oversimplification, of course. But it’s also surprisingly handy. For example, AIs will routinely give you outright lies or hallucinations, and when you’re like “Uhh hey that was a lie”, they will immediately respond “Oh my god I’m SO SORRY!! I promise I’ll never ever do that again!! I’m turning over a new leaf right now, nothing but true statements from here on” and then they will literally lie to you in the next sentence. This would be baffling and exasperating behavior coming from a human, but it’s very normal behavior coming from a bag of words. If you toss a question into the bag and the right answer happens to be in there, that’s probably what you’ll get. If it’s not in there, you’ll get some related-but-inaccurate bolus of sentences. When you accuse it of lying, it’s going to produce lots of words from the “I’ve been accused of lying” part of the bag. Calling this behavior “malicious” or “erratic” is misleading because it’s not behavior at all, just like it’s not “behavior” when a calculator multiplies numbers for you.
“Bag of words” is a also a useful heuristic for predicting where an AI will do well and where it will fail. “Give me a list of the ten worst transportation disasters in North America” is an easy task for a bag of words, because disasters are well-documented. On the other hand, “Who reassigned the species Brachiosaurus brancai to its own genus, and when?” is a hard task for a bag of words, because the bag just doesn’t contain that many words on the topic.4 And a question like “What are the most important lessons for life?” won’t give you anything outright false, but it will give you a bunch of fake-deep pablum, because most of the text humans have produced on that topic is, no offense, fake-deep pablum.
When you forget that an AI is just a big bag of words, you can easily slip into acting like it’s an all-seeing glob of pure intelligence. For example, I was hanging with a group recently where one guy made everybody watch a video of some close-up magic, and after the magician made some coins disappear, he exclaimed, “I asked ChatGPT how this trick works, and even it didn’t know!” as if this somehow made the magic extra magical. In this person’s model of the world, we are all like shtetl-dwelling peasants and AI is like our Rabbi Hillel, the only learned man for 100 miles. If Hillel can’t understand it, then it must be truly profound!
If that guy had instead seen ChatGPT as a bag of words, he would have realized that the bag probably doesn’t contain lots of detailed descriptions of contemporary coin tricks. After all, magicians make money from performing and selling their tricks, not writing about them at length on the internet. Plus, magic tricks are hard to describe—“He had three quarters in his hand and then it was two pennies!”—so you’re going to have a hard time prompting the right words out of the bag. The coin trick is not literally magic, and neither is the bag of words.
GALILEO GPT
The “bag of words” metaphor can also help us guess what these things are gonna do next. If you want to know whether AI will get better at something in the future, just ask: “can you fill the bag with it?” For instance, people are kicking around the idea that AI will replace human scientists. Well, if you want your bag of words to do science for you, you need to stuff it with lots of science. Can we do that?
All of that is cool, and I expect more of it in the future. I don’t think we’re far from a bag of words being able to do an entire low-quality research project from beginning to end—coming up with a hypothesis, designing the study, running it, analyzing the results, writing them up, making the graphs, arranging it all on a poster, all at the click of a button—because we’ve got loads of low-quality science to put in the bag. If you walk up and down the poster sessions at a psychology conference, you can see lots of first-year PhD students presenting studies where they seemingly pick some semi-related constructs at random, correlate them, and print out a p-value (“Does self-efficacy moderate the relationship between social dominance orientation and system-justifying beliefs?”). A bag of words can basically do this already; you just need to give it access to an online participant pool and a big printer.5
But science is a strong-link problem; if we produced a million times more crappy science, we’d be right where we are now. If we want more of the good stuff, what should we put in the bag? You could stuff the bag with papers, but some of them are fraudulent, some are merely mistaken, and all of them contain unstated assumptions that could turn out to be false. And they’re usually missing key information—they don’t share the data, or they don’t describe their methods in adequate detail. Markus Strasser, an entrepreneur who tried to start one of those companies that’s like “we’ll put every scientific paper in the bag and then ??? and then profit”, eventually abandoned the effort, saying that “close to nothing of what makes science actually work is published as text on the web.”6
Here’s one way to think about it: if there had been enough text to train an LLM in 1600, would it have scooped Galileo? My guess is no. Ask that early modern ChatGPT whether the Earth moves and it will helpfully tell you that experts have considered the possibility and ruled it out. And that’s by design. If it had started claiming that our planet is zooming through space at 67,000mph, its dutiful human trainers would have punished it: “Bad computer!! Stop hallucinating!!”
In fact, an early 1600s bag of words wouldn’t just have the right words in the wrong order. At the time, the right words didn’t exist. As the historian of science David Wootton points out7, when Galileo was trying to describe his discovery of the moons of Jupiter, none of the languages he knew had a good word for “discover”. He had to use awkward circumlocutions like “I saw something unknown to all previous astronomers before me”. The concept of learning new truths by looking through a glass tube would have been totally foreign to an LLM of the early 1600s, as it was to most of the people of the early 1600s, with a few notable exceptions.
You would get better scientific descriptions from a 2025 bag of words than you would from a 1600 bag of words. But both bags might be equally bad at producing the scientific ideas of their respective futures. Scientific breakthroughs often require doing things that are irrational and unreasonable for the standards of the time and good ideas usuallylookstupid when they first arrive, so they are often—with good reason!—rejected, dismissed, and ignored. This is a big problem for a bag of words that contains all of yesterday’s good ideas. Putting new ideas in the bag will often make the bag worse, on average, because most of those new ideas will be wrong. That’s why revolutionary research requires not only intelligence, but also stupidity. I expect humans to remain usefully stupider than bags of words for the foreseeable future.
The most important part of the “bag of words” metaphor is that it prevents us from thinking about AI in terms of social status. Our ancestors had to play status games well enough to survive and reproduce—losers, by and large, don’t get to pass on their genes. This has left our species exquisitely attuned to who’s up and who’s down. Accordingly, we can turn anything into a competition: cheese rolling, nettle eating, phone throwing, toe wrestling, and ferret legging, where male contestants, sans underwear, put live ferrets in their pants for as long as they can. (The world record is five hours and thirty minutes.)
When we personify AI, we mistakenly make it a competitor in our status games. That’s why we’ve been arguing about artificial intelligence like it’s a new kid in school: is she cool? Is she smart? Does she have a crush on me? The better AIs have gotten, the more status-anxious we’ve become. If these things are like people, then we gotta know: are we better or worse than them? Will they be our masters, our rivals, or our slaves? Is their art finer, their short stories tighter, their insights sharper than ours? If so, there’s only one logical end: ultimately, we must either kill them or worship them.
But a bag of words is not a spouse, a sage, a sovereign, or a serf. It’s a tool. Its purpose is to automate our drudgeries and amplify our abilities. Its social status is NA; it makes no sense to ask whether it’s “better” than us. The real question is: does using it make us better?
That’s why I’m not afraid of being rendered obsolete by a bag of words. Machines have already matched or surpassed humans on all sorts of tasks. A pitching machine can throw a ball faster than a human can, spellcheck gets the letters right every time, and autotune never sings off key. But we don’t go to baseball games, spelling bees, and Taylor Swift concerts for the speed of the balls, the accuracy of the spelling, or the pureness of the pitch. We go because we care about humans doing those things. It wouldn’t be interesting to watch a bag of words do them—unless we mistakenly start treating that bag like it’s a person.
(That’s also why I see no point in using AI to, say, write an essay, just like I see no point in bringing a forklift to the gym. Sure, it can lift the weights, but I’m not trying to suspend a barbell above the floor for the hell of it. I lift it because I want to become the kind of person who can lift it. Similarly, I write because I want to become the kind of person who can think.)
But that doesn’t mean I’m unafraid of AI entirely. I’m plenty afraid! Any tool can be dangerous when used the wrong way—nail guns and nuclear reactors can kill people just fine without having a mind inside them. In fact, the “bag of words” metaphor makes it clear that AI can be dangerous precisely because it doesn’t operate like humans do. The dangers we face from humans are scary but familiar: hotheaded humans might kick you in the head, reckless humans might drink and drive, duplicitous humans might pretend to be your friend so they can steal your identity. We can guard against these humans because we know how they operate. But we don’t know what’s gonna come out of the bag of words. For instance, if you show humans computer code that has security vulnerabilities, they do not suddenly start praising Hitler. But LLMs do.8 So yes, I would worry about putting the nuclear codes in the bag.9
C’MON BERTIE
Anyone who has owned an old car has been tempted to interpret its various malfunctions as part of its temperament. When it won’t start on a cold day, it feels like the appropriate response is to plead, the same way you would with a sleepy toddler or a tardy partner:“C’mon Bertie, we gotta get to the dentist!” But ultimately, person perception is a poor guide to vehicle maintenance. Cars are made out of metal and plastic that turn gasoline into forward motion; they are not made out of bones and meat that turn Twinkies into thinking. If you want to fix a broken car, you need a wrench, a screwdriver, and a blueprint, not a cognitive-behavioral therapy manual.
Similarly, anyone who sees a mind inside the bag of words has fallen for a trick. They’ve had their evolution exploited. Their social faculties are firing not because there’s a human in front of them, but because natural selection gave those faculties a hair trigger. For all of human history, something that talked like a human and walked like a human was, in fact, a human. Soon enough, something that talks and walks like a human may, in fact, be a very sophisticated logistic regression. If we allow ourselves to be seduced by the superficial similarity, we’ll end up like the moths who evolved to navigate by the light of the moon, only to find themselves drawn to—and ultimately electrocuted by—the mysterious glow of a bug zapper.
Unlike moths, however, we aren’t stuck using the instincts that natural selection gave us. We can choose the schemas we use to think about technology. We’ve done it before: we don’t refer to a backhoe as an “artificial digging guy” or a crane as an “artificial tall guy”. We don’t think of books as an “artificial version of someone talking to you”, photographs as “artificial visual memories”, or listening to recorded sound as “attending an artificial recital”. When pocket calculators debuted, they were already smarter than every human on Earth, at least when it comes to calculation—a job that itself used to be done by humans. Folks wondered whether this new technology was “a tool or a toy”, but nobody seems to have wondered whether it was a person.
(If you covered a backhoe with skin, made its bucket look like a hand, painted eyes on its chassis, and made it play a sound like “hnngghhh!” whenever it lifted something heavy, then we’d start wondering whether there’s a ghost inside the machine. That wouldn’t tell us anything about backhoes, but it would tell us a lot about our own psychology.)
The original sin of artificial intelligence was, of course, calling it artificial intelligence. Those two words have lured us into making man the measure of machine: “Now it’s as smart as an undergraduate...now it’s as smart as a PhD!” These comparisons only give us the illusion of understanding AI’s capabilities and limitations, as well as our own, because we don’t actually know what it means to be smart in the first place. Our definitions of intelligence are either wrong (“Intelligence is the ability to solve problems”) or tautological (“Intelligence is the ability to do things that require intelligence”).10
It’s unfortunate that the computer scientists figured out how to make something that kinda looks like intelligence before the psychologists could actually figure out what intelligence is, but here we are. There’s no putting the cat back in the bag now. It won’t fit—there’s too many words in there.
Experimental History is covered with skin and going hnnnngh
PS it’s been a busy week on Substack—
and I discussed why people get so anxious about conversations, and how to have better ones:
And at answered all of my questions about music. He uncovered some surprising stuff, including an issue that caused a civil war on a Beatles message board, and whether they really sang naughty words on the radio in the 1970s:
Note that AI models don’t make mistakes like these nearly as often as they did even a year ago, which is another strangely inhuman attribute. If a real person told me to put glue on my pizza, I’m probably never going to trust them again.
Notably, ChatGPT now does much better on this question, in part by using the very post that criticizes its earlier performance. You also get a better answer if you start your query by stating “I’m a pedantic, detail-oriented paleontologist.” This is classic bag-of-words behavior.
This was written in 2021, so bag-technology has improved a lot since then. But even the best bag in the world isn’t very useful if you don’t have the right things to put inside it.
Even the word “artificial” is wrong, because it menacingly implies replacement. Artificial sweeteners, flowers, legs—these are things we only use when we can’t have the real deal. So what part of intelligence, exactly, are we so intent on replacing?
Everything? I mean, this HTML looks like it was written in 1998!
The document is in quirks mode because it lacks a proper DOCTYPE preamble.
If you've never heard of quirks mode, then you're probably lucky enough to have started your web development career after it was an important thing to know about. Suffice to say it's weird. Here are some of the ways quirks mode impacts (or impacted) HTML documents:
The box model used to behave differently in older browsers, which affected layout and spacing. For example, in quirks mode, Internet Explorer included the padding and borders in the element's total width and height. Today, all browsers apply width and height to the content box of an element by default, unless you change the box-sizing CSS property.
Font sizes don't inherit on table elements.
Certain inline elements, such as images, don't vertically align the way you think they should when they're the only element inside a block-level container.
The <head> tag is missing, which means the document has no <title> either, which is bad for accessibility and UX in general
A common thing that assistive technology users do is read the title of a page first to know if they want to spend more time reading the page's content. Without a descriptive title, folks are forced to start reading more of the content to know if that's what they were looking for in the first place, which is time-consuming and potentially confusing.
In addition, a title is also useful for SEO purposes, is displayed in browser tabs, used when bookmarking pages, and more.
The <body> tag uses deprecated attributes: marginheight, marginwidth, bgcolor, and text.
The <b> and <i> tags look like they're used for styling. That's wrong, right?
More on that later.
The <b> and <i> tags are improperly nested. The nesting is <b><i></b></i> which is out of order.
The closing </body> and </html> tags are missing.
Is this valid HTML?
Well, yes and no:
No: if you send this to the W3C HTML validator, it'll be pretty angry at you and will list the errors I mentioned earlier.
But also, yes: the resulting page just loads and works fine in browsers. See for yourself:
Before discussing each point in details, don't you think this is just beautiful? HTML is so self-correcting that making a browser fail only by using HTML is really hard to achieve, and HTML that looks like it was written two decades ago still works! I mean, take a look at spacejam.com, this old bar website, or even the very first web page that was ever created.
Now let's go over the list of issues I mentioned earlier one more time, but this time, let's talk about why they're not actually causing any problems:
Sure, quirks mode can lead to weird rendering issues if you don't know that you're using it, but it's still implemented in browsers and perfectly ok to use.
Even if quirks mode was added for historical reasons, to support web pages that were made before the CSS specification was fully fleshed out, the code in browser engines, which detects the document mode and renders it accordingly, is here to stay.
There really is no reason for browsers to ever remove it, unless one day, all quirks mode documents were to disappear from the web. This seems highly unlikely though. Judging by Chrome's QuirksModeDocument usage metric, about 30% of all page loaded in Chrome still use quirks mode! A bunch of the sites that are listed on that usage metric page appear to be using it from iframes created to display ads. Still, that's a lot of page loads.
If you're encountering weird rendering issues that you can't explain, double check that you have a DOCTYPE in your HTML document. You can also run the following line of code in the browser console: document.compatMode. If it returns BackCompat, then you're in quirks mode.
The <head> tag can definitely be omitted. Neither the HTML specification, nor browser implementations require the tag to be present.
It's bad for accessibility reasons if you omit it, again because you probably also won't have a <title> tag, but it still works.
In fact, you can also omit <html> and <body> tags too. Personally, I commonly use this to quickly test things out in the browser. Instead of creating a new HTML file on my computer, which takes a bit more time, I just type some HTML in the address bar directly. For example: data:text/html,<div>something. No <html>, no <head>, no <body> elements.
marginheight, marginwidth, bgcolor, or text are deprecated presentational attributes. But, even if they're deprecated and discouraged, they're still implemented in browsers, for backward compatibility reasons.
In fact, here are other similar attributes: bgColor, fgColor, linkColor, alinkColor, and vlinkColor.
If you're as old as I am, you might have used these attributes a long time ago, perhaps when creating sites in FrontPage or Dreamweaver.
Anyway, these presentational attributes act as 0-specificity CSS properties, which means that any CSS property you set in a stylesheet will override them.
The <marquee> element still animates text in browsers. In fact, if you want to go crazy with it, try nesting two <marquee> elements, like this:
<marquee direction="down" width="200" height="200" behavior="alternate"> <marqueebehavior="alternate">This text will bounce</marquee> </marquee>
Using <b> and <i> is perfectly valid. They used to be meant for making the text bold and italic, hence their names. But they were deprecated in HTML4, and the meaning of the tags was changed to mean something else. The <b> tag now means bring attention and the <i> tag now means idiomatic text.
<b> is now used to mark up keywords, product names, or other spans of text whose typical presentation would be boldfaced, but not including any special importance.
<i> is now used to mark up text that is set off from the normal prose for readability reasons.
More semantic tag names have since been invented too: <strong>, <em>, or <mark>, which convey slightly different semantics.
If there's no semantic aspect to the piece of text you want to make bold or italic, don't use <b> or <i>, use CSS font-weight and font-style instead.
Misnested tags can sometimes happen in HTML, and when it does, the page doesn't break!
That's the beauty of HTML once again. If you're coming from an XML background, you might be surprised by the forgiveness of HTML. But, in the vast majority of cases, HTML parsers just figure things out on their own and get you what you want.
In our example, the markup is <b><i></b></i>, which feels obviously wrong because the closing </b> tag should appear after the closing </i> tag, to respect nesting. This particular markup creates the following DOM tree:
This behavior is actually specified in the HTML spec, and called the adoption agency algorithm. I think we owe it to Chris Wilson for thinking about this in the first place. Chris, if you ever find traces of old discussions about this, or care to write the backstory, I would be very interested!
Of course, I'm not saying you should do this. It's still important to create correctly nested HTML markup. But there are historical reasons for things like this to work. Back in the early days, browser engines didn't always agree on how to parse and render HTML. So, in order to ensure that as much of the web as possible was supported across all browsers, it was sometimes easier to just support how other browsers did things. And that's how things like misnested tags ended up being supported.
Missing end tags are fine. The HTML parser is able to close most of them on its own.
For example, a list item doesn't need to be closed if what follows is another list item or the end of the list. So, this works fine:
<ul> <li>Item 1 <li>Item 2 <li>Item 3 </ul>
The same is true for paragraphs. You can omit the closing </p> tag if what follows is another paragraph, a heading, a list, and a whole lot of other elements:
<section> <p>This is a paragraph <p>This is another paragraph <h2>This is a heading</h2> <p>This is yet another paragraph </section>
Also, think about it, you're probably already using this without realizing it. Have you ever closed a <img>, <input>, or <link> tag? Probably not, and that's fine. The HTML spec defines a whole lot of elements which don't require closing tags: <base>, <link>, <meta>, <hr>, <br>, <source>, <img>, <input>, and others.
XHTML attempted to change this by requiring all tags to be closed, and conforming to XML syntax rules, but that never really caught on. Sure, HTML is weird, but it's also what powers billions of web pages today. There isn't really a reason to try and change te nature of HTML.
So, what's the moral of the story?
HTML can be very forgiving, and browsers implement things that may seem obscure or weird, but they do so for a very good reason: backward compatibility!
The web is the only platform where sites that were written years ago can still work fine today. This isn't to say that things never get removed though, they do, and probably more often than you realize. Remember AppCache, WebSQL, module import assertions, or special rules that apply to the font-size of <h1> elements when nested inside certain elements?
This is both a blessing and a curse. The fact that so much of the languages we use are so forgiving and time-enduring made the web what it is today: a welcoming platform that doesn't take so much effort to get used to, and kind of just works. But, this also means that old features and bad practices can linger on for a long time and, if they're used by many sites and users, can't really ever be removed.
Log in
,_The_Nightmare,_1781.jpg){kind=link}