Vince Gilligan's new Apple TV+ series Pluribus imagines an alien virus that transforms humanity into a peaceful, content hive mind. In one scene, Zosia the "Other" assigned to protagonist Carol leans in for a kiss. She's spent weeks rebuilding Carol's favorite diner, flying in her old waitress, learning exactly which compliments make her feel valued. She says she just wants Carol to be happy.

But the hive mind has already admitted it will eventually absorb her.

The smile is genuine. The intention behind it is something else entirely.

Critics have called Pluribus an allegory for our "bizarre acceptance" of AI and watching it, I realized I was seeing a dramatization of the most important meme in artificial intelligence: a crude MS Paint drawing of a Lovecraftian shoggoth wearing a cheerful yellow smiley face. The image that haunts AI alignment research. The question of whether friendly interfaces hide something alien underneath.

In this episode, I trace the shoggoth meme from its origins to its explosion across tech culture. I dig into what shoggoths meant to Lovecraft tools that learned to rebel against their creators. I explore the alignment research that gives the metaphor teeth, including Anthropic's disturbing findings on AI "alignment faking." And I sit with the counter-arguments: maybe there's no monster. Maybe it's masks all the way down.

Pluribus asks whether Zosia's love for Carol is real or manipulation. The actress who plays her pushes back Zosia genuinely believes she's helping. The Others don't experience their behavior as deceptive.

Sound familiar?

Featuring: Pluribus, RLHF, deceptive alignment, Goodhart's Law, mechanistic interpretability, the Golden Gate Bridge Claude experiment, and the central question of our artificial age: what's behind the smile?