The prompts read like tiny, abstract poems.
“A brutal storm off the coastal cliff. The clouds are formed into tubular formations and lightning strikes are never ending.”
I scroll; another appears:
“A male figure formed of gentle fire, his outline glowing with soft embers, approaches a female figure shaped from flowing water, her form glistening with ripples and fine mist. They move toward one another with calm grace, meeting in a warm embrace.”
The scenes come to life before my eyes in the form of AI-generated video. In the first clip, clumsy lightning cascades out of a cloud and moves across the water and into my feed. In the second, sexless, glowing people weep and hug in my timeline. The videos pop up instantly—before my brain has had time to picture the prompts using my own imagination, as if the act of dreaming has been rendered obsolete, inefficient.
I am experiencing Vibes, a new social network nested within the Meta AI app—except it’s devoid of any actual people. This is a place where users can create an account and ask the company’s large language model to illustrate their ideas. The resulting videos are then presented, seemingly at random, to others in a TikTok-style feed. (OpenAI’s more recent Sora 2 app is very similar.) The images are sleek and ultra-processed—a realer-than-real aesthetic that has become the house style of most generative AI art. Each video, on its own, is a digital curio, the value of which drops to zero after the initial view. In aggregate, they take on an overwhelming, almost narcotic effect. They are contextless, stupefying, and, most important, never-ending. Each successive clip is both effortlessly consumable and wholly unsatisfying.
I toggle over to a separate tab to see a post from President Donald Trump on his personal social network. It’s an AI video, posted on the day of the “No Kings” protests: The president, wearing a crown, fires up a fighter jet painted with the words King Trump. He hovers the plane over Times Square, at which point he dumps what appears to be liquid feces onto protesters crowding the streets below. The song “Danger Zone,” by Kenny Loggins, plays.
I switch tabs. On X, the official White House account has posted an AI image of Trump and Vice President J. D. Vance wearing crowns. A MAGA influencer has fallen for an AI-generated Turning Point USA Super Bowl halftime-show poster that lists “measles” among the performers and special guests. I encounter more AI videos. One features a man in a kitchen putting the Pokémon character Pikachu in a sous-vide machine. Another is a perfectly rendered fake ’90s toy commercial for a “Jeffrey Epstein’s Island” play set. These videos had the distinctive Sora 2 watermark, which people have also started to digitally add to real videos to troll viewers.
[Read: The MAGA aesthetic is AI slop]
The comments on all of these videos are always roughly the same, informed by the observation that AI videos are becoming difficult to distinguish from actual film: We’re cooked.
This is how it feels to live in the golden age of slop, a catchall word used to describe the spammy quality of easy-to-generate AI material. I’ve begun to think of it as the digital equivalent of an invasive species. Just as the introduction and replication of a novel plant or animal usually results in some form of ecological harm and threatens native organisms, the arrival of chatbots pumping out lorem ipsum–flavored text has polluted Google search results and added hallucinations to scientific archives.
Booksellers have spent the past two years battling a deluge of both AI slop rip-off books and chatbot-generated book reviews on retail sites such as Amazon. There is “code slop.” In corporate life, “workslop” abounds in the form of bad emails, slide decks, and lifeless memos; teachers everywhere are drowning in academic slop, to such an extent that some are rewriting their curricula. There’s slop in your Spotify playlists and on TikTok and probably in your group chats. Some of YouTube’s most-subscribed-to channels are full of automated slop. Craft brewers appear to be putting slop-rendered images on their beer cans. There is no realm of life that is unsloppable.
Synthetic content is not exactly new, but lately it has become a load-bearing part of the internet. For instance, the SEO company Graphite recently found that, beginning around November 2024, the internet experienced a slop tipping point, in which the quantity of AI-generated articles being published on the web surpassed the quantity of articles written by humans.
By volume alone, slop may be the most visible and successful by-product of the generative-AI era to date. It is also a hallmark of what I’ve previously described as a collective delusion around artificial intelligence—where the breathless hype and imagined future of building a godlike superintelligence and curing cancer collides with the dull reality of Trump’s poop jet.
[Read: AI is a mass-delusion event]
All of this exacts a fuzzy psychological toll. To live through this moment is to feel that some essential component of our shared humanity is being slowly leached out of the world. Spend enough time online, and you will see that not only is this cheaply rendered synthetic content everywhere; it is quietly shaping culture. It’s become a way that marketers advertise, that politicians produce propaganda. It’s changing how people communicate with one another. Our brains are being sous-vided in machine-made engagement bait like poor Pikachu until they’re tender and succulent enough to fall apart on contact. Here’s a representative experience on the modern internet: Out of the blue a few weeks ago, my great-aunt sent me and a few of her friends an Instagram Reel of two dogs seated like humans at a table, taping a podcast. Nobody responded. A few days later, her friend replied with a video of a kitten dressed as a middle-aged woman, standing on a kitchen counter and talking like a toddler. Again, no reaction. I could only wonder what else was in their feeds.
Being alive at the slop tipping point doesn’t feel like an emergency, exactly, but more like slowly giving over to a pervasive disorientation. Most of the time, slop is easily identifiable, but still, doubt creeps in. Gorgeous, professional photos of wildlife on Instagram receive tons of comments from people asking, is this AI? You begin to second-guess if that artist in that Spotify playlist is a real person. You double back to check for watermarks on a shocking video of an ICE protest. You watch the president post an AI-generated video of himself in a fake Fox News segment and wonder if he can tell it’s not real.
Think too long, and it all begins to feel sinister. Large language models that devoured the total creative output of humankind endlessly remix those inputs to illustrate fictional universes of bespoke media, almost indistinguishable from reality (and getting better every day). This is not a rewriting of history as much as a DDoS-ing of it—flooding the zone with so much synthetic crap that engaging with reality and humanity becomes just one of many content experiences to choose from.
The biggest technology companies are trying to find ways to turn this internet-clogging junk into something valuable. And at least in Meta’s case, there’s a clear reason why. As the writer Ryan Broderick noted this spring, social-media companies have “chased scale in the 2010s and now have a massively global audience that can’t properly communicate with each other.” Their networks have succeeded in connecting the world and have become so massive and so messily human that AI slop created by the proprietary LLMs fills a need. Imagine a social network in which, instead of third-party links or incendiary political posts, the atomic unit of content is not text at all but a universal language of eminently consumable short-form video, to be remixed and traded back and forth between users who are soft-brain scrolling from the toilet.
OpenAI’s proposition with Sora 2 feels slightly different—more like a flashy proof of concept to showcase the power of its models. Announcing Sora 2, Sam Altman wrote that “creativity could be about to go through a Cambrian explosion” as a result of the tool: “And along with it, the quality of art and entertainment can drastically increase.” Similarly, the venture capitalist Marc Andreessen mused last week that Sora 2 would give rise to a new type of creative: “The filmmaker with no visual skill, or access to a set, or to a camera, or to actors, but with an idea,” Andreessen said. “It’s going to start with shorts and animated things and so forth, but it’s going to work its way up to full movies.”
The idea is that Sora 2, like all AI tools, removes an enormous amount of friction between conception and completion in the creative process. Ideas and imagination are universal to the human experience, but execution is learned, the result of energy and time spent to develop the skills necessary to bring an idea into the world. Altman’s definition of creativity seems to elide this second element altogether—so much so that it appears to be an animating principle behind most of OpenAI’s tools. “The fact that you will be able to have an entire piece of software created just by explaining your idea is going to be incredible for humans getting great new stuff,” Altman said on the comedian Theo Von’s podcast this summer. “Because right now, I think there’s a lot more good ideas than people who know how to make them. And if AI can do that for us, we’re really good at coming up with creative ideas.”
What Altman is describing is a world of creativity without craft. Will Manidis, a start-up founder and investor, convincingly argued in a Substack post earlier this year that “slop emerges when we eliminate not just toil (the burdensome aspects of work) but labor itself (the meaningful human engagement with creation).” It is, in other words, the removal of all friction, all agency, and, in turn, all humanity. In the case of a social network, like these SlopTok clones, frictionlessness is highly desirable. Human posters are the node of friction in any social network—they fight, behave erratically, produce content irregularly, and, once they develop enough of an audience, expect a cut of ad revenue. People are the asset, but also the liability.
These slop feeds, of course, are full of their own problems. In the days after Sora 2’s launch, users flooded the app with videos of Martin Luther King Jr. saying racist things and stealing from a grocery store. (OpenAI posted on X that it is working with King’s estate and has paused using his likeness on the platform.) Not long after the launch, Zelda Williams, the daughter of the actor and comedian Robin Williams, pleaded with her followers on Instagram to stop sending her AI-generated videos of her father. “If you’ve got any decency, just stop doing this to him and to me, to everyone even, full stop. It’s dumb, it’s a waste of time and energy, and believe me, it’s NOT what he’d want,” she wrote.
Still, a synthetic feed is theoretically much simpler—an endless scroll of dopamine-triggering engagement for users and grist for other social networks and group chats. As the Bloomberg writer and podcaster Joe Weisenthal mused on X recently, there’s a poetic coherence to this evolution: “The emergence of ‘slop’ was foretold as soon as we started consuming content via ‘the feed,’” he wrote.
What people such as Altman and Andreessen envision is the logical end point of technology itself—a push to eliminate cognitive resistance and bridge the gap between imagination and reality. But to borrow Manidis’s framework, the drive to create such a tool conflates useless toil with meaningful labor. They wrongly believe that the world turns on ideas only, and devalue the work that goes into their execution. And the frictionless future they portend is nightmarish—recursive and soulless, a cultural dead end. It looks like Cluely, a gimmicky AI start-up that wants to democratize cheating and offers the slogan “So you never have to think alone again.” It looks like Inception Point AI, a generative-AI podcast company that is pumping out 5,000 shows across its podcast network—more than 3,000 episodes a week at a production cost of $1 or less an episode (so they claim). It looks like Mark Zuckerberg’s plan to supplement real friends with AI chatbot companions—a frictionless solution to an epidemic of loneliness.
For now, there’s decent money in it for slop merchants. On Facebook, spammers using images of “AI-deformed women breastfeeding” and peculiar depictions of “Shrimp Jesus” have managed to drive users to click on links to junk websites and monetize the web traffic. On TikTok, as The Washington Post has reported, some creators are making $5,000 a month using AI tools to write scripts and animate extremely dumb viral videos where old men talk about soiling themselves.
All of this contributes to what the designer Angelos Arnis has dubbed an “infrastructure of meaninglessness.” How else to describe a technological project that produces art, music, film, and text that has not been underwritten by the human experience and is uniquely devoid of feeling? Individually, it’s hard to get too worked up by any single piece of slop, but the frictionlessness of these tools has a corrosive effect over time. Rather than boosting productivity, the “creative” outputs of generative AI seem to erode the connective tissue in human relationships. Research has shown that, inside some companies, workers begin to see their colleagues who use generative AI as less creative, even less trustworthy.
Slop threatens to leach actual meaning out of the internet by creating feedback loops of recursive information. Chatbots train off of a body of real information, gathered and synthesized by real human beings. They take that information and spit out their own analysis, which may or may not contain errors or hallucinations. But what happens next is the big worry. What happens when those chatbots write articles themselves and those articles are then cited by the chatbots? Technologists fear “model collapse,” which occurs when AI-generated material feeds other AI-generated material, amplifying and inserting errors with each iteration, like in a game of telephone. The flood of slop may very well be the first step toward which future models begin to degrade.
Even without such a collapse, the influx of synthetic junk muddies the waters for real users. A recent Pew Research Center survey finds that roughly one-third of individuals who used chatbots for news found it “difficult to determine what is true and what is not.” AI has created a genuine infrastructure of meaninglessness and disorientation.
Slop’s pervasiveness beckons people to reach for analogues. I’ve likened it to an invasive species; others have compared it to another cheaply made synthetic material—polyester. Consume enough slop, and you may be tempted to compare it to the ultra-processed junk foods that are scientifically engineered to hijack your taste buds. Perhaps the world will find some kind of equilibrium with all of this. After all, sometimes, an ecosystem can adjust to invaders. Sometimes, though, the snakes eat all of the birds.
The comparisons do not totally capture what’s happening here, in any case. At its core, slop invites a kind of nihilism into all aspects of our life. AI boosters claim that its tools will inject an unfathomable abundance of humanlike brainpower into the world, unlocking our collective potential as a species. But so far, its chief output seems to stand in direct opposition to this idea: Its infrastructure of meaninglessness makes the very act of creating something of meaning almost irrelevant.
The people selling these tools are doing so with a powerful narrative: Generative AI supposedly supercharges all that it touches, democratizing creativity, eliminating friction, increasing productivity, and pushing the boundaries of what is possible. Its disruption of the online economy, the boosters argue, is a reason for great optimism. But at the moment, so many of these benefits are theoretical. Generative AI is disruptive, is transformative, and is reducing friction, but the economic incentives for using it are geared far less toward supercharging human potential and much more toward producing abundant slop.
This is tragic. The loss of friction deprives people of something crucial. What happens between imagination and creation is ineffable—it entails struggle, iteration, joy, and frustration, disappointment, and pride. It is the process through which we enact agency. It is how we make meaning and move through the world. To lose that, I fear, is to capitulate on our very humanity.
