I’ve been writing the demand side of this story for a while. Cargo Cult Civilization is about what the feed does to human attention. This essay is the supply side: what’s about to be poured down the funnel.

AI Slop is inbound. You already knew this, so it’s no surprise. “Is that real?” is going to be the question of the next year, maybe a little more. Right now you can tell the difference between real and AI. Mostly. Soon, you won’t be able to tell. At all. And in either case, I’m not sure it matters, because these horses left the barn already, and exactly no one really cares. Why? The feed, amigos. The feed is you. It’s your brain.

You’re the product.

It’s 6am and you just got up. Show of hands here, how many of you reach for your phone literally before even going to the bathroom? I bet it’s more than half.

Just gonna stop right there. Because actually this little essay isn’t about AI Slop. It’s about an artificially low PE of 22.

I keep reading the analyst takes on Meta’s capex and they’re all running the wrong model. They evaluate Zuck’s $135B 2026 infrastructure bet by comparing it to the foundation labs. OpenAI burning money on reasoning. Anthropic burning money on alignment. xAI burning money on, I dunno, vibes. The bull case for those companies requires AGI, or at least an enterprise reasoning agent that can do a paralegal’s job without hallucinating its way into a malpractice suit. That’s a long, narrow path with a lot of “and then a miracle occurs” steps.

Meta’s path? They have to invent a slightly better cat video.

I’m not being glib. The product is a feed. The feed is optimized for dopamine. The dopamine response is well-characterized: variable reward, novelty, parasocial mimicry, mild outrage, outgroup dunking, animal cuteness, mild horniness. There’s thirty years of literature on what hijacks a primate attention system, sitting on top of an empirical layer that Meta has been writing since 2007. They know what works on you. They know what works on the version of you sitting on the couch at 11pm on a Tuesday after a hard day. They know it down to the millisecond.

The bar for synthetic content to beat user-generated content is hitting a known dopamine target for nine seconds at a time. That’s a much easier problem than the Turing test. Video gen models cleared that bar eighteen months ago. The 2026 models clear it cheap and at scale.

The asymmetry gets uncomfortable when you list what Meta actually has:

  • The largest dopamine telemetry dataset on Earth, fifteen-plus years deep
  • The world’s most sophisticated ad targeting system, already trained on that telemetry
  • Distribution to roughly 3.6B daily actives
  • Inference infrastructure being built out at $135B a year
  • No requirement to share revenue with creators
  • No requirement to moderate content they themselves generated
  • No morals

That moderation one is the part nobody’s pricing in. UGC carries a giant operational tax: moderation, copyright enforcement, creator relations, the whole trust-and-safety apparatus that exists because humans are messy and post shit you cannot lawfully host. Synthetic content can be constrained at generation time. Never produce X. Always produce Y. The brand safety nightmare that broke Twitter and made TikTok into a regulatory piñata? Actually solvable when you control the generation function.

TikTok proved the social graph is dead weight. Nobody scrolls Reels to see what their cousin posted. They scroll because the algorithm serves them whatever lights up the limbic system, and the limbic system doesn’t care if cousin Brad shot the video. Creators are feedstock, not the mission.

Synthetic content is the obvious next step. If you can generate the dopamine hit directly, you don’t need the creator. The creator was a workaround for not having a generator.

(Cory Doctorow has been writing about this for years under “enshittification.” He’s right about the dynamics. He has been characteristically generous in assuming there’s a floor where the platform breaks. There isn’t. The floor was the supply of replacement humans, and synthetic content removes the floor entirely.)

The Zuck capex bet, viewed through this lens, is the single most rational capex bet in the industry. OpenAI is spending to win a research race against unknown technical risk on an unknown timeline. Meta is spending to build a content factory whose product is already validated, whose distribution is already built, whose monetization is already proven, and whose competitors are still mistaking themselves for foundation labs.

I’m a big nobody in equity research, but I think the market is mispricing this by evaluating Meta on the wrong frame. The stock is down about a quarter from peak on “capex anxiety.” Q1 2026 came in at 33% revenue growth. The stock barely moved on the print. That gap is the mispricing.

The bear case lives in two places, neither of which is the financials.

The first is regulatory. The EU AI Act is going to require disclosure on synthetic content, probably more aggressively than the US, and in principle a disclosure regime could break the immersion that makes the feed work. I’m skeptical this matters. People know cigarettes cause cancer. The warning labels did not, in the limit, stop people from smoking. An “AI-generated” tag at the bottom of a Reel is a cigarette warning, and you’re going to scroll past it. The feed will still feed.

The second is cultural. A backlash where “real” becomes a premium signal, the way Kagi became a premium signal for people who actually want to control their information diet. Possible. Maybe even probable in some bounded slice of the population. The equivalent of the people who pay for Kagi today, which is a rounding error against Google’s user base. That carves out a niche. It does not break the main product.

Betting against dopamine optimization at scale has historically been a bad trade. Junk food won. Slot machines won. Pornography won. Reality TV won. Algorithmic news feeds won. The premium signal of “real” has always existed. It has always lost.

(I have kids. I see what they actually watch versus what they tell me they watch. Yes, vintage record collectors exist. So do a billion Spotify accounts. Do the math.)

So what’s the play? I don’t give investment advice and you shouldn’t take it from a guy writing blog posts about TTRPGs and bioelectric fields. But the structural argument is this:

Meta is not in the AGI race. Meta is in the synthetic content distribution race. The synthetic content distribution race has a clear technical path, a built-out distribution layer, an already-monetized customer base, and competitors who keep mistaking themselves for foundation labs. The capex spend looks crazy if you’re scoring it on “win the next intelligence breakthrough.” It looks completely fucking sane if you’re scoring it on “feed the supply curve that’s already eating the demand curve.”

Don’t sleep on Meta. The reason the stock is down is that the market is still trying to figure out which race they’re in.

It’s the cat video race. They’re going to win the cat video race.

Meow.