Dumb and Dumber
Dumb and Dumber
It’s 6:23 AM, and my thirteen-year-old, after a shower lasting approximately one thousand years, is dressed and is now prowling the kitchen for breakfast. He’s past the age where I make this happen on his behalf. He has a cardboard cylinder of Quaker Oats instant oatmeal in his hand. I am one room away, on my second cup of coffee, prepping for my work schedule. I can see him out of the corner of my eye. He’s holding the oatmeal for maybe five seconds.
“Dad…”
I do not answer. I will not answer.
Then his voice, in that vomitous cadence every kid his age has etched into their brains, you know the one, the one that YouTube “content creators” pour forth. It sounds the way a day-glo cereal box looks. He’s using the YouTube voice to himself at six-fucking-thirty in the morning. Not talking to me now. Talking to… who? Himself? This is unclear.
Nope, I was wrong. As it turns out, he was talking to… (wait for it)
ChatGPT.
Asking for directions. On how to make the oatmeal.
He’s holding the goddamn canister in his hand. The instructions are right there. On the ever-loving canister. Half cup oats. One cup water. Microwave. Fourteen-point type. In his hand.
“How do I make… (pause, reading front of canister) ‘instant oats’?”
Instead of reading the instructions, in his hand, my teenage child asked a datacenter instead, and the datacenter answered. Correctly, it seems. Thoroughly. Bulleted list, serving suggestion (add blueberries). And the oatmeal got made, and got eaten. At no point was there anything specifically wrong here, except maybe that he didn’t rinse out the bowl. I’m still waiting for that to latch in his brain.
But. This isn’t just my kid. It’s your kid, too. All of them, the whole Croc-wearing, sweatpants-clad herd of them. They all meme-speak, they all streamer-talk. They seem to think that life should imitate the content they’ve been served for the entirety of their lives. Someone right in their face, throwing stream-of-consciousness scenery-chewing overacting at them. Undeserved and incommensurate surprise. Shock. Anger. A small palette of base emotions, with the affective volume turned up to 11.
(If you don’t know what I mean by streamer prosody, ask any middle school teacher. Watch the eye twitch.) I once heard one of my teenage sons address the dinner table conversation as “this chat.” Like it was Discord.
Anyhow my thirteen-year-old boy cannot make fucking oatmeal without supervision, and the supervision he chose was a server farm.
Yay.
I already wrote the feed essay. Cargo Cult Civilization is the long(ish) version. You won’t go read it so I’ll summarize. Engagement-optimized feeds are the most successful behavioral engineering project in human history. Billions (many billions) of dollars of tuning aimed straight at the oldest exploits, the OG zero-day vulnerabilities: the human nervous system.
Thinking (actual cognition!) is metabolically expensive. It’s hard to train. It takes decades to educate a homo sapiens to use their own brain for more than 3.3 seconds at a time, as it turns out. So, yeah, literal energy cost to think. The Feed (enagagement feed, algorithmic feed, whatever you prefer) is a gas station stocked with twinkies and Dr. Pepper. It’s cheap. It’s so very easy. Nothing about it is real, or uplifting, or nourishing to the human spirit. But our brains are lazy, so we know what wins. Big Gulps of TikTok.
Don’t take my word for it. It works against you. When researchers audited Twitter’s engagement ranking against a plain chronological baseline, they found it amplified emotionally charged, out-group-hostile content. But wait! There’s more! Users themselves reported that this material made them feel worse, content those same users did not prefer even by their own stated preferences.
Sit with that. The machine ostensibly built to serve your revealed preferences feeds you things you say you don’t want, because you linger on them anyway, for longer. Shit you do not want, fed to you despite your preference. The slot machine doesn’t need you to like it. It needs your hand on the lever. Good little rat. Pull the lever again.
Yes, Virginia, we are getting dumber. (I’m not making this up.) Fifteen years of systematically atrophying cognition. Cognition is a muscle. It takes years to develop, and you have to use it to keep it strong. We ain’t using it. The cause is environmental: the supercomputer in your pocket. In a billion pockets. Did I mention I’m not making this up?
I’m not going to spend words on this. It’s already baked in.
I’m going to rant about what happened next - and is still happening.
Until roughly 2023, it seemed that our cognitive nosedive might pull back up and fly straight.
Work still required thinking (mostly). School demanded it (mostly). Making oatmeal demanded. A little, anyhow. However weak your cognition muscle was by the end of the day, every new day forced that muscle through some reps. You had no choice, the way manual labor kept farm workers reasonably fit, like it or not. The Feed weakened the muscle, but the world wouldn’t let you crash because the world forced you to keep flapping your damn arms.
Then, we got AI.
Now, at any hour, day or night, there floats a ubiquitous, ever-present, seemingly omniscient, encyclopedic thing that will answer any question, solve (or attempt to solve) any problem, talk about whatever you want to talk about, and generally think on your behalf for more-or-less “free”. (New flash: not free.)
The difficult thinking-work that the Feed trained you to avoid, on the rare occasions you need this work done, is now 100% outsourceable. Instantly. Fluently. Better than your Dad or your boss or your spouse, your expert cousin Eddie, or anyone else you actually fucking know. Especially, and palpably: better than YOU, most of the time.
The Feed and the AI are sequenced (not through intention, but synchronicity, thanks civilizational-Jung) to cause the most possible wreckage across our fragile and complex society. Either one of these, by themselves would be a lot for us to absorb as a global civilization, but together? Oy vey.
If you just handed AI to a cognitively healthy population, you’d get leverage, force multiplicative effects. Like power tools are to woodworking, so AI to thinking. But what we’re getting is outsourced cognition handed to a seething mass of humanity who would rather eat Cheetos and doomscroll more than anything else. AI arriving at the moment when the world is like, “Sure, something that will replace my own thinking, for those (what, really annoying) times when I still have to think? I’m in, where do I sign up?”
Dumb and Dumber. Harry you could deal with. Lloyd, sure you can deal. Together? Not so much.
The Feed subsidized cognitive avoidance (or, maybe just counterfeit cognition? time will tell) paying you in dopamine to click “like” and keep scrolling for more. Now, AI drops the marginal cost of outsourced cognition to zero. (Used to be, you had to hire an executive assistant or a Chief of Staff to do real work for you.) People do what? Pick the lowest cost thing. That’s Moloch, right on schedule. Not sinister conspiracy. Optimization functions.
Cruel irony: the thing the Feed atrophies is the thing that we might use to escape the optimization basin. Working AI as a fulcrum (sparring partner, red teaming) rather than a replacement requires sustained, effortful, self-directed cognition. The exact muscle. So AI doesn’t shift the distribution down uniformly on humanity. It bifurcates it. A wicked small cohort with intact attention gets superpowers, but plebes get a very fluent prosthetic and they better fucking be content with that. My kids’ cohort isn’t losing the ability to think unaided. You can’t lose what never finished growing. Which is really, deeply sad.
We all already despise half of this, the AI half. The internet rage at AI slop is righteous and still climbing. Shrimp Jesus, the dead-eyed video slurry coating every surface. This hate is (mostly) good. But observe: the hate attaches only to the part that’s obvious, the slop, because the algorithmic wood chipper that selects and spews out the Feed and delivers the slop is… invisible. The phantom hand on the Feed, driving it. Ya can’t see it. But it’s there.
Rage needs something tangible to hate. Slop is such an object: you can screenshot it, forward it, make memes out of it. Easy to hate on. The Feed? By comparison it actually quite hard to point at What It Is or even What It Ain’t No More. Before it was, what? Better? And now it’s Worse? This is harder to hate.
Anyhow the rage has it backwards. The AI slop is a symptom, not the disease. Slop is a consequence of optimizing for engagement. The Feed is the disease. Slop is the natural basin, the evolutionary pinnacle, when engagement is the fitness function and marginal cost of production drops to zero. Nobody would mass-produces shrimp Jesus for a chronological feed of people who chose to follow them. It wouldn’t propagate. The algorithm is the selection environment, the petri dish. Slop is the phenotype, a fungus. AI just made the phenotype cheap enough to saturate the pond.
Weirdly, AI slop isn’t new - it’s just an evolved form of the content the Feed was already selecting for. For years. Engagement optimization selected for this exact content for years, and what did subhuman “content creators” do? Contorted themselves to produce it at human speed. AI steps in and pumps out enough of it, fast enough, that the vile endpoint is finally visible to the naked eye. Slop is the Feed’s algorithmic portrait, ugly enough to notice at last.
The worst part? This “AI Slop!” won’t last. Soon you will literally not be able to tell the difference. Mark me. This fervor will not last, because just like an MP3 didn’t sound as good as the hiss and pop of analog vinyl, but goddamn it’s so nice to have the entire catalog of all human music at your fingertips, the anti-AI thing will simmer down into a permanent, vocal minority of purists. No one else will give a shit.
Vinyl collectors and unassisted thinkers: one demographic, two hobbies.
Meanwhile, offstage, some actual adults have noticed a few things.
Australia first: a full under-16 social media ban starting December 2025, world’s first, no parental override, sold by the communications minister as protection from “predatory algorithms.” Eight months later, the regulator’s own study found more than eight in ten under-16s still on the platforms, most of them scrolling just as often as before, using VPNs and faked face scans, or decamped to whatever app the ban forgot.
This is somewhat akin to naming a new epidemic virus, and then announcing that the mitigation was that humans should leave. Kids, being rational optimizers just like my son at the pantry, routed around the law like it was packet loss. Meanwhile the actual mechanism, the predatory Feed algorithm the minister stood at a podium and named, remains perfectly goddamn legal for every human on earth aged sixteen and up, and de facto available to everyone under.
Le Sigh™
Back in the US, we’re anemically circling around something like the 1965 tobacco playbook. The Surgeon General called for a warning label, explicitly modeled on the cigarette one. Senator Schatz wants algorithmic feeds banned for kids. He got one thing right, anyway: open skepticism (even contempt) towards “just let users toggle the algorithm off”. Lulz. We all know exactly no one toggles anything, bro. It’s 2026, c’mon. But notice the contours of every proposal: protect the children, presume the adults are fine, leave the objective function untouched. “BUT THE KIDS!”
Apparently that’s what passes for political effectiveness these days.
Maybe the kids are OK? But no, the adults are not fine. Adults are an atrophied cohort holding the voting power and the checkbook. Aim at the kids and you concede the mechanism; concede the mechanism and 2026’s AI layer closes the second jaw on everybody, sixteenth birthday or no.
Ok, yeah, it’s bleak. Don’t shoot the messenger. So here’s my fix.
Ban the monetization of engagement.
All of it, both directions. No advertising served on any feed ranked by engagement in any way, and no creator payout keyed to engagement metrics. Nobody buys the attention the ranker farms, and nobody gets paid for farming it. The feed itself stays legal. Rank however you want; the money just can’t touch it.
This will never happen. But it should.
I’m not saying throw out AI, to clarify. AI is dual-use, a lever for the strong and a crutch for the weakened, and you regulate dual-use tech rather than banning it; nobody outlaws forklifts because some people stopped lifting. (I’ve spent a whole essay series arguing these models don’t think. I’m not here to defend the crutch. I’m here to tell you the crutch is only lethal to people the feed already broke.) And not “social media,” not phones. Those are containers.
Engagement - as an optimization endpoint - is different in kind. It has zero legitimate use to a globalized civilization. There is no useful version, and not even any banal version of “maximize time-on-site by exploiting limbic response” that serves the user - or humanity - and the audit literature shows users don’t even want its output by their own account. It’s a pathogen - a parasite - wearing a product. You don’t put a warning label on a pathogen. You cut off its food.
“But… you have to rank stuff!” Sure. But my friends, discovery was solved. And can be re-solved. Humanity solved ranking-at-scale in 1998, and the company that solved it had “don’t be evil” in the handbook at the time. (The irony…) PageRank ranked on authority: the link graph. Who vouches for whom. Turtles all the way down. Every input was a property of the corpus - the thing itself. None of them consulted your heart rate, your outraged nervous system, your anger-driven blood pressure. Old Google ranked the books. The library. The Feed ranks your human frailties. That’s the bright line, and it’s draftable as law: platforms may rank on attributes of the content (relevance, authority, provenance, recency) and may not rank on predicted or observed user behavior. Rank the book, not the reader. A regulator can audit that sentence.
We have seen this movie. We know how it ends when the engagement objective eats the authority objective, because it went down in public, to The Google, with the emails and entrails sitting right in federal discovery. In 2019 Google’s ads side called a “code yellow” on search because people weren’t asking Google enough questions; the head of search wrote that he was “deeply, deeply uncomfortable” with the growth plans. Shockingly (not really) and soon after, the former Head of Ads had his job. Zitron read the emails so you don’t have to. We already know the ending, because we lived through it: same corpus, same product, objective function swapped, and a billion people watched search rot in real time.
And now it’s all but gone. Zero-click was always the endpoint. It’s the bottom of the basin.
Blah blah, yes, SEO was gameable. That’s a bullshit objection and you know it.
No platform will fix this voluntarily. Even the ones whose employees hate the damage they are intentionally causing. This is Moloch’s home turf. (Longtime readers know coordination failure is my jam.) Any platform that unilaterally drops engagement ranking hands its users’ eyeballs to the competitor that keeps it, and every executive knows it, which is why the ones with consciences quit and write memoirs. Multipolar traps are blind to any virtue of the trapped; they yield only when the payoff matrix changes for every player at once. In this case, that would require force of law, which is the entire reason laws exist. We should use them.
I think a ban could actually work. I don’t think it will ever happen, but I do believe it would work. Smoking got canceled, even though cigarettes are still legal today. It took a generation or three, sure. (Surgeons used to smoke in operating rooms. Ask anyone over seventy.) The indoor ban was the fulcrum: plenty of restaurant owners privately wanted smoke-free rooms, and none could disarm alone without handing the smokers to the bar next door. The law disarmed the whole street simultaneously. And the unlock for all of it was secondhand smoke: the moment your habit became my lung problem, personal choice stopped being a defense.
The feed has secondhand smoke too. Your uncle’s engagement diet votes in your elections. My son’s feed writes his prosody. Human cognition is the lung.
I could go on. There would be corner cases, sure. Maybe one could voluntarily pay for a Feed. Tax the shit out of it. Etc. Designated smoking areas become designated waivers for Feeds. Something like this, I dunno, I’m not a policy guy.
We don’t even need to manufacture any new rage-energy to get this done, we already have the AI slop hate. We got a surplus of hate, really. It just needs to be aimed properly.
Dumb and Dumber.
So you’re telling me there’s a chance.
My kiddo finished the oatmeal, did not rinse the bowl, and wandered off narrating something to nobody in that bright, fake, borrowed voice. Good kid. Great kid, honestly. They’re all great kids, and that’s the point. The oatmeal canister went back in the pantry, instructions facing out.
Nobody will read them.
Colin Steele deliberately leaves typos, grammatical errors, cusswords and mangled metaphors lying around so that you know that this was written by an actual human bean.