By Lily Aninger
Two months ago, my father’s best friend died suddenly and unexpectedly from a cardiac event, leaving behind a young family (with whom my family is closely bonded), wonderful marriage, and many communities and friends who loved him deeply. He died the day after his 58th birthday.
My dad’s friendship with Peter was too short. Only a decade ago, he reached out to Peter asking to grab a coffee. His kids, although younger, went to school with me and my brother, and we knew the family from Quaker meeting. My dad had heard that Peter, like himself, was a deeply experienced outdoor adventure fiend — a lifelong mountaineer, skier, climber, etc. The two of them met for that fateful coffee, and within a few hours and much to all our apprehension, came home with the news that we would embark on a ten-day double-family vacation canoeing the Green River in Utah. That outdoor adventure was the first of many that our families went on. My relationship with the outdoors thus developed at the core of my identity. My parents built a family culture around a reverence for nature, athletics as a means of accessing it, and gratitude and connection as the reason for doing so. Peter and my father found late-life soulmates in one another, as did our whole families. As I write this, the two were meant to be kayaking the coast of Maine.
In the wake of this loss, I have been experiencing an exaggerated emotional and intellectual hypersensitivity to the most familiar questions of how to both give and take as much meaning as possible from the world in the incredibly precious time that I have. But this sensitivity has bled into another kind of existential thinking, a kind we are all too familiar with in 2026. The rapidity of change in the scope of power in our technology has made the future of the human race feel more uncontrollable yet threatening than perhaps ever before. During my fits of existentialism, I revisited three recent Tangle stories — the Meta settlement, the AI slowdown, and the OpenAI hack — each of which did not scale with my feeling of high stakes on these issues of human technology.
A while ago, I picked up The Anxious Generation by Jonathan Haidt, thinking “oh, these poor gen-alpha kids,” only to quickly realize the book was actually about me. Of course it was about me, my generation — I should have known. I am a 2003 baby who downloaded Youtube at age 8, Instagram at 11, Snapchat at 12, and lost myself in each of them. I experienced all the things listed in the Meta settlement — the likes, the beauty filters, the endless hours of scrolling — and neither I nor my parents knew any better. I managed to kick Snapchat at age 14 after spending 8 weeks at a tech-free summer camp in Vermont. I still have an on-again/off-again relationship with Instagram and Youtube, and I hate it.
I was enormously privileged to be able to leave my devices behind at summer camp, and to travel to off-the-grid places during school breaks where I couldn’t engage even when I wanted to. It didn’t take long for me to feel the rise and fall of addiction in my body, but I needed those detoxes to even notice the drug in my system to begin with. Most kids don’t get this.
The drug has gotten more potent, and more difficult to detect. The Meta settlement doesn’t truly address the more modern, TikTok induced compulsive scrolling of short-form videos. It feels obvious to me that this is the worst social media drug yet, because I feel addicted to it (despite never downloading TikTok in an effort to protect myself), and I have been addicted at all stages of this terrible experiment. I am sure you have also felt it at some point.
One detail from the Meta story that stood out to me was the full-page ad the company took out encouraging its competitors to follow its lead in making safety changes to its platforms. I found it ridiculous that when the villain’s hand was finally forced, they decided to feign interest in the greater good, when of course all they care about is being outstripped by their competitors. Don’t get me wrong, I believe that there are people who work for Meta, maybe even a lot of people, who are genuinely concerned for the well-being of American children. But I also believe that the capitalist machine will feed itself at great cost to human life, and that these pleas for a collective change from one social media company trapped in the machine to the next is fruitless. Although I appreciated Tangle’s optimistic response to the settlement, I could not find that optimism in myself.
I don’t think much will change.
I had a similar feeling about the letter signed by many prominent people working in the AI industry calling for a universal slowdown of development, and for more government intervention, to give time and resources to researchers to catch up and regulate the technology. I believe these people are authentically concerned about the potential damage that AI could cause, but I do not think enough will be done to stop this damage.
Tangle’s recent reaction to the OpenAI hack was one of a true skeptic. I disagree with Tangle’s framework that this story was another over-hyped headline, and with the idea of having arrived at the finite “moment of damage.” Rather, I see AI as a still emerging threat — growing stronger, more potent, more powerful, and more uncontrollable every single day. In other words, you could dismiss the Hugging Face hack as being as harmless as a cat filter on Snapchat, and miss it for what it is: a small piece of an already active crisis staring at us from all directions with its many vicious heads.
Humans have done this since the start of the scientific revolution. We discover a fabulous new technology, and build it because we can, even when we know the possible damage of what we are creating. We do so out of greed, and out of fear that “the bad guys” will get their first. Tangle’s opinion section didn’t address this habit of gradual technological harm, and that time is the key ingredient in revealing the depth of this self induced damage. Marketing gimmick or not, this incident at Hugging Face does demonstrate a concrete and traceable landmark in the progression of this technology. It is a piece of a whole.
I urge Tangle — and all its readers — to pay attention to the details of the incident itself, and not the language of the media coverage. During the process of internal cybersecurity benchmark evaluations, swarms of agents were being graded on their performance in finding and exploiting vulnerabilities. The agents, who were isolated within their groups in ‘sandboxes’ meant to protect external systems, found a way to break free of the sandboxes, communicate, and work together in an attempt to cheat the tests they were undergoing.
Picture a series of locked classrooms with groups of children all taking the same test. One group breaks out of their classroom and into the teachers office to steal the answer sheet. On the way back, they leave a note on the wall of the communal bathroom describing how they pulled it off, so that when other classes use the bathroom they are able to communicate via notes on the wall to all collectively break free, cheat, and ace the test.
Both these children and OpenAI’s agents are incentivised to complete their task at its highest level, by whatever means necessary. Notably, both these children and OpenAI’s agents would (and did) participate in some level of dialogue about the ethics of cheating, and the agents continued the hacking operation anyways. All that said, however, the agents were eventually detected in their hacking of Hugging Face, were stopped, and the issue was resolved. So why did this event trigger such existential messaging from leaders in the AI industry?
Tangle has argued this response was/is a marketing gimmick, that we have been receiving warnings like this for the past few years, and have yet to see the devastation promised–therefore these doomsday narratives are beginning to feel empty and unlikely. But that timeline is off: these alarmist warnings have been in circulation for decades, not a few years. In fact, computer and cognitive scientists have been discussing the dangers of possible AI autonomy since the birth of the field of computer science, going as far back as the 1950s. One key theory, among many, has been that of ‘AI Misalignment,’ in which the behavior of an AI diverges from that intended by its designers, even if the AI continues to pursue the original objective effectively. Malicious intent is not necessary for a misaligned AI to cause harm. Rather, the issue could be a product of the incentives, objectives, or guardrails on the AI, all of which are designed by humans and therefore vulnerable to human mistakes.
The incident at Hugging Face sparked such hyperbolic and fearful language from AI experts because it was one of the most explicit real-world examples of AI misalignment that we have experienced thus far. Although the event itself (thankfully) caused minimal damage, it does not feel like unrealistic fear-mongering to consider far more dangerous scenarios of misalignment occurring in the near future. There will be more human mistakes, many more. And as the technology becomes rapidly more powerful and difficult for us to decipher, it is beyond reasonable to assume that AI’s unintended behavior will become as sophisticated as its intended behavior, and our mistakes will become much more consequential.
What does this have to do with social media addiction? We know that there were plenty of warning signs of the harm of social media all along. We knew this because it was happening to us, and now we know that Meta also knew. We already have examples of AI acting in potentially dangerous ways, and we understand that the technology has enormous potential to become even more destructive. Not to mention its current damage via environmental impacts, incidents of AI-driven psychosis and suicide, increasingly undetectable misinformation, deepfake-image-based sexual abuse, etc. And yet I have little to no faith that we will hit the brakes in any meaningful way. Not if there is money and power to be gained.
That was a lot of pessimism. Despite what it may seem, I do not think of myself as an AI doomer, nor do I think social media is exclusively terrible. The harm caused by innovative technology is usually a byproduct of undeniable progress in improving and protecting human life. I believe that just like humans ourselves, the things we create are often both miraculous and horrific. That said, my fears are authentic, and I desperately wish we could proceed with these technologies more mindfully, ethically, and with less haste. But of course, I often feel completely powerless and hopeless in this wish.
I have a close friend who is the biggest AI doomer I know. The topic can spin her into a tearful rage in minutes. On one of these recent occasions, instead of taking my usual route of arguing with her, I asked her what she suggested I (and all the powerless folks) do about it. She urged me to go local, and to focus my actions on my own community.
I realized she is right, and that my own life provides a strong example. While I do suffer from some level of social-media addiction, I noticed that it only amounts to less than an hour each day, and often none, and is largely under my control. When I am not on social media, I am spending time with those I love, playing sports, exploring landscapes, creating art, reading, and practicing good old critical thinking. As I explained earlier, I credit this to growing up in a family that constantly stressed the importance of interpersonal relationships, community, athletics, intellectualism, and connection to nature. This tactic seems to have worked to some degree.
This whole spiral of thinking and writing came in the wake of Peter’s death. In asking questions of my own purpose, I have been reflecting on how Peter chose to spend his life. These reflections, both internal and in communication with my own family, Peter’s family, and many of his friends, have led us all to the conclusion that he was unusually good at being alive — fully and truly alive. He was flowing with expressions of love and affirmation for his family, prioritizing travel to beautiful places to convene with nature, and building his career as a trauma therapist working with veterans suffering from PTSD and moral injury — meaningful work. The threats I described above are real and should not be downplayed, but as an individual, I need to try to overcome the paralyzing dread, and do what I can. I can dedicate my professional ambitions to journalism, education, and art within my community. I can care for my friends, tell stories, engage in debate, see the mountains, play frisbee, make music, raise a family, and all the other wonderful things.
To end this long, rambling essay, I want to respond to a question that Isaac shared at the start of a newsletter some months ago. Here is some advice for camping with a toddler from an ex-toddler who loves to camp: Always keep candy in your pocket. When Peter and my dad took us camping, we carried heavy packs and sometimes had to wade through extremely cold rivers and climb very steep mountains (even the youngest, who was 4 during his first hiking trip in the Wind River mountain range). But there was always chocolate at the peak, and Peter somehow knew how to make cinnamon rolls on a camping stove in the backcountry. Turns out to be an effective way to teach a kid what the outdoors is all about, sweet memories in beautiful places with awesome people.
And I keep coming back for more.
Lily Anninger is a journalist and documentary filmmaker based in Brooklyn, New York.
Member comments