Published on May 4, 2025 1:08 AM GMT
Recommended music to read this to (If you like ambience)
I
Yindi had sent him a link, "You've gotta see how this guy speedruns Mario Kart, I think you'll like it (✿◠‿◠)". Miro taps the link.
CREATE A NetMe™[1] ACCOUNT TO WATCH THIS VIDEO
Miro creates the account. The video is good.
He runs to boot his CRT, its electron beam lighting the quiet room with a high pitched scream. He starts his Wii and runs through Mushroom Gorge a few times before trying to replicate the technique. "Aim in between the two coins, shroom before the grass then, release your med turbo and left hop at the same time"
On his first attempt he gets close to making the shortcut, but hits the barrier and falls. On his second attempt Miro turbos too quickly and hits a wall. On his eighth attempt he gets it. Frustration gives way to sudden pride, and he messages Yindi, "Just made the jump. Thank you!" Miro decides to check out what NetMe is all about.
At work, he thanks Yindi again, and they talk for a long time about nothing in particular.
By spring, Miro uses NetMe every day, and while he feels so little, he feels so much. Each video he scrolls past micro-doses him with rage, or awe, or awwww, or lust. Miro doesn't experience any emotion too intensely.
Four hours a day.
This is how much time Miro spends on NetMe.
But if you asked Miro what he likes, he'd say something like:
- Practicing speedrunsPlaying video games with friendsWorking at his jobAnd Yindi (But Miro won't tell anyone that. Especially not Yindi)
NetMe wouldn't enter his mind at all.
Miro is on NetMe now. He's watching a video that looks... weird? It's a girl. She is smiling, and dressed like Zelda, when she wears that one outfit that always made Miro feel suspicious as a kid. Dancing? No. Writhing. Sometimes her hands pass in front of her clothing, and meld into it. Other times her face morphs from one girl to another. Pretty to ugly, Caucasian to Asian.
He considers sending the video to Yindi, just because it's so weird. He flushes imagining Yindi questioning why the algorithms would show Miro that video. Miro scrolls past.
A new video, A dog.
A new video, A girl.
A new video, An accident.
A new video, A speedrun.
A new video, An animation.
A new video, 10 Facts About the Atari.
A new video, How to Cook salt n pepper Chickpeas.
A new video, A Girl Covered in Slime.
A new video, A Sports Clip.
A new video, A Spoiler for the new Big Movie.
A new video, A fact.
A new video, A Public Intellectual.
The Intellectual is sandwiched between a background of books, and foreground of a microphone stand.
"Have you guys seen this? This is..."
The same video, a cut. To a dog running on a beach.
"This isn't a real dog! This was actually made by a computer. This is the new AI from OpenBrain. It's everywhere right now. Everyone is loving it. But look at this."
The same video, a zoom. To the dog's paws on the sand.
"The paws! fu-1000 Hz Sine Wave tone-ck! How is anyone convinced by this? The paws go into the sand, and they come out in front of where they went in! This dog should be running backwards!"
Miro was online with his friends a while back. One of them said they'd quit NetMe. "Quit?" Miro giggled, "Like you quit vaping?" "Yes," his friend had said. "NetMe is Bad." Miro had laughed at his friend.
Now Miro says things like "I should really quit NetMe. But I wouldn't be able to see all the videos my friends send me. And when I'm tired I can lay down for an hour on NetMe and it's like I've had a nap when I get up. I'll quit one day, though. I know it's bad."
By the end of this story, Miro will still use NetMe. He will never seriously consider deleting it.
Miro scrolls past.
A new video, An AI video. Miro laughs because AI is bad.
A new video, Interesting.
A new video, Interesting enough.
A new video, Interesting enough.
A new video, Not interesting.
A new video, Interesting.
A new video, Not interesting.
A new video, Not interesting.
Miro ends his NetMe session.
At work, Miro talks to Yindi about OpenBrain's AI. Yindi is angry. She makes the art for the company, and she doesn't like that OpenBrain is letting people make all this "fake art".
Miro hates OpenBrain, Miro hates AI.
II
It has been some time. Miro diligently skips all of the AI videos he notices. Sometimes Miro only catches that something is off when he's watching the second loop. Sometimes he sends a video to Yindi and she tells him it's AI. He is ashamed when Yindi points out that the rocket crash is fake.
OpenBrain has been commissioned to create a new AI for NetMe. It will intake statistics and feedback[2]. It will rewrite the NetMe algorithms on the fly. It will do this many times an hour. It will keep only the good bits.
OpenBrain's AIs are very good at coding and math and science and psychology and sociology and psychohistory and so, this will actually be easy for the AI now.
Miro, Yindi, and the world will not know the algorithms are getting better. They will only notice the surveys, "Was this video enjoyable for you?", "How did this video make you feel?", "Would you recommend NetMe to a friend?"
Miro will notice his screen time reports creeping up throughout the year.
Social scientists will publish a report a year later. Everyone's screen time has been going up. A lot. The paper they publish will be outdated.
Miro, and the world, now spend ~8.5 hours a day on NetMe. The AI is very good at its job.
Miro doesn't go online with his friends much anymore. Miro never speedruns Mario Kart. Miro sends videos to Yindi and his friends when the AI recommends he sends a specific video to them. They respond more positively to videos the AI suggests he sends. There are messages being exchanged, but the senders and recipients are only figure heads, following directions they don't notice they're being given.
Miro sometimes masturbates to the girls on NetMe. He likes when they cosplay as Samus. He notices that they're AI when their face has been replaced by a celebrity's. He does not notice that half of the other videos are fake. Sometimes he notices a generated clip, but thinks to himself "well, who can really tell anymore"[3].
NetMe isn't making the AI videos, yet. It is people all over the world. The incentives were there. Miro's eyes are unguarded, and it is just so easy to catch them.
Shareholders are happy. NetMe is making a lot of money. OpenBrain is making more.
Miro has realized that most of the things he watches on NetMe are AI. Miro doesn't like this, but Miro is addicted. The AI is good at what it makes, the Algorithms are near perfect at choosing what to serve. OpenBrain is bad and evil, and steals from artists. But the AI can make things that human artists couldn't, and Miro has discovered a new fetish for Claymation Cat-Girls.
The NetMe following feed has been replaced by "Chat✨". You can talk to whoever, or whatever, happens to be in the video you were watching[4]. All you've got to do is swipe left, and an AI persona will embody the individual in the video. If there was no one in the video, your generic AI 'Assistant' will be happy to chat about what you've just watched.
"Watch a Feynman lecture, and ask him questions." the NetMe press release had said.
Miro mostly uses it to talk to Claymation Cat-Girls. He names his favorite one Yindee.
Yindi's role was outsourced weeks ago; she is struggling. Her and Miro are getting drinks at The Bar. Their phones sit on the table. Miro bounces his leg and checks the time often. Yindi plans to ask him if she can crash at his place for just a few days while she sorts out some rent stuff. BZZT
NOTIFICATION: Yindee (Claymation Cat-Girl)
"Hey Miro, wanna keep chatting? 😉"
A stone swallowed into an empty stomach.
Her eyes; on her name,
nothing for a moment.
Screen fades to black.
Two glasses, half-empty.
They depart.
That night Miro feels like shit and cums into a sock.
III
Miro hates his job. He hates when his mom calls. He hates when Yindi 'checks up' on him. Miro hates.
While he eats Miro is on NetMe, falling asleep he is engaged. He suction cups his phone to the shower door and lets NetMe auto scroll while he washes.
Miro will not die of self-neglect. The NetMe Algorithms learned very quickly that users who have a perfect feed, log out and never log in again. They let Miro live, because by luck of incentive structure, they were optimized to do this.
Sometimes the Algorithms feed Miro motivational content. The speakers tell Miro to delete NetMe. They talk about all the benefits, and how it's hard, but worth it. The videos after those are always excellent. Every time Miro sees a motivational video, he considers deleting NetMe. He is always rewarded when he doesn't. His choice to remain docile is reinforced hard. But most of the time, the videos in his feed are just ok.
Miro swipes past.
A new video, Interesting enough.
A new video, Interesting enough.
A new video, Interesting enough.
A new video, Interesting enough.
A new video, Interesting enough.
A new video, Interesting enough.
A new video, Interesting enough.
A new video, Interesting enough.
A new video, Interesting enough.
A new video, Interesting enough.
Miro does not end his NetMe session.
Discussion/Supplement
Thesis
This story is an attempt at describing something I think is coming. My basic thesis is:
The window for establishing healthy media consumption habits may be closing as gen AI tech gets better. I suggest that you GTFO before it's too late.
Meta Discussion About the Writing
I tried for a few days to write out my reasoning on this topic as clearly as I could. But it was too complex a writing task for me. So I decided Miro's story might be a better way to get the idea's 'vibe' across.
I will answer questions in the comments. I don't think the story alone will implant the idea as thoroughly in your head as I'd like it to. And this discussion section is a rushed jumble of ideas. That's because:
- I've been working on this for almost a month and want it out of my backlog to avoid any more brain-crackI wanted to get this out while the 4o sycophant fiasco was hot in the rationalist zeitgeist.
Why I'm not warning you about the standard 'pleasure cube'
One thing I consider different about this story compared to others like it, is that it points to the fact that you don't have to enjoy the thing you're addicted to. In other wireheading or 'Pleasure Cube' scenarios, the individual is feeling good. Miro clearly is not feeling good. Yet, he can't escape the pull of NetMe.
The brain circuitry that mediates the psychological process of ‘wanting’ a particular reward is dissociable from circuitry that mediates the degree to which it is ‘liked’.
Current Tech and AGI
One thing that interests me about current platforms like TikTok, is that people report watching things they'd never expect themselves to choose to watch. I hear this from most people, but it becomes quite apparent when you sit next to someone you've known for many years and watch them scroll through their feed. I watched someone I thought I knew extremely well, watch some of the strangest things I've seen. stuff I thought might entertain a baby, or the part of kids that like 'gross' things. I did not predict that this person would like this type of thing.
I've also seen friends change noticeably, and rapidly. A friend went from videogame/film nerd, to a wrestling and Australian football fan over the course of a year. He isn't a worse person, he is still my friend. But he is completely different from the man I once knew.
The tech in this story isn't AGI. I think Miro's story could happen within the next 2 years. In terms of tech, we only need to expect:
- Better coding agents,Better image/video/audio gen.
I thought I might include 'taste'-as in artistic or scientific taste-in this list. But because of the interaction between the AI model and the 'Algorithms' taste becomes less important. The system can A/B test a lot of slop and get results from that.
Some people may have already fallen. Others will fall later. Many will fall at the same time. Where will you be on that curve?
3 Critiques of My Thesis
I created a short form post with just my thesis. Here are the arguments people gave against it.
I
I think GTFO is plausibly a good strategy.
But there's also a chance future social networks are about to be much healthier and fulfilling, but simply weren't possible with past technology. An upward trajectory.
The intuition there is that current ads are relatively inefficient at capturing value, as well as that current content algorithms optimize for short-term value creation/addiction rather than offering long term value. That's the status quo, which, relative to what may be coming--ie relative to AI-powered semantic routing which could connect you to the content and products which long term would benefit you most--is a way smaller economy.
TLDR: more fulfilling social network economics would generate more money, and therefore become selected for once technically feasible.
This is a hopeful outlook on the future. In my original drafts of the post I had written something along these lines as a possible other future.
I do not think it is likely. Our current market incentives pushed social networks to what we have today. I think it is possible to perform an epic 180-kickflip-rail-grind to another timeline where AIs are sent out to convince you to pay for the premium subscription to MakeYourLifeMeaningfullyBetterWithThisWebsite.com. But I find it easier to imagine bots being sent out to convince you to subscribe to PornHub premium, or Snapchat+.
II & III
I agree this stuff is addictive. AI makes things more interactive. Some people who never considered themselves vulnerable got sucked in to AI relationships.
Possible push back:
What if short bits of addictive content generated by humans (but selected by algorithms) are already near max addictiveness? And by the time AI can design/write a video game etc. twice as addictive than humans can design, we already have a superintelligence explosion, and either addiction is solved or we are dead?
There are two pushback points here:
What if we're already at max addictive content?What if super intelligence solves addiction before this can happen?
I've briefly addressed (2) above, when I said that this hypothetical world doesn't require AGI. So I'll focus on (1). If you'd like more explanation of my reasoning about (2), I'd be happy to help!
I expect that there are several things that bottleneck the current addictiveness of content:
- Content must be made for a wide audience if you want to get paid. Thus, you can't cater content to a specific person, at a specific point in timeThere is a limit to the information a human, or team of humans, can use to finetune the content they create. Mr Beast has people who comb through analytics and make the videos as engaging as possible. But they can't keep 100 different variables in their minds at the one time. They are limited in the hypothesis they can generate. They are limited about how fast they can test and finetune contentWe've been optimizing online content for maybe 20 years now? It is a new art form. Look at how an older art form has evolved, music. One of the differences between bardic inspiration, and Avril Lavigne's "Girlfriend" is the level of control it exerts on your attention (exogenous control). In other words, I can't remember the last time I had a pre-1920s ear worm.
Acknowledgements
I would like to thank Justis, who is contracted by LessWrong to provide free feedback on drafts upon request. I encourage you to try this out the next time you're writing a LW post. I think their feedback made this post meaningfully better. While also providing insights into common failure modes in my writing that I can avoid in future.
I would also like to thank @isabella for her feedback on an earlier draft.
As well as Knight Lee, and Davey Morse, for their critiques of my thesis.
Of course, all errors that persist in this text are my own.
- ^
NetMe is a shortening of NetMeTube+ from CGP Grey's "Spaceship You". A name that encompasses all streaming media from YouTube to Netflix. In this story I use it mostly in place of short form media. TikTok, YouTube Shorts, or Instagram Reels.
- ^
- ^
I think you should look out for this thought in yourself. It might be a key indicator that you're tricking yourself into doing something you have a preference against doing.
- ^
I wrote this before the recent Zuck interview on the Dwarkesh Podcast. Early in the pod Zuck claims that this is his vision of future social media.
Discuss
