Picture the moment. It’s late on a Saturday night in September 2026, the new season of Saturday Night Live is barely an hour old, and Michael Che is doing the thing Che does — a slow setup, a flat delivery, a turn toward the guest seat. What sits down next to him is Jane Wickline in an elaborate wig, playing Anthropic CEO Dario Amodei. Not a generic hoodie-and-slides tech founder. Him, specifically. And rather than play him as one person, the bit splits him into two versions of the same man, both answering Che’s questions.
I review AI tools for a living. I spend my weeks poking at agents that promise to book my travel and refactor my repo, then writing up how they actually failed. So I want to be clear about why a sketch comedy segment is worth 800 words on this site: it’s a measurement. Not of Anthropic’s products, which a wig can’t tell you anything about. Of something else.
Getting spoofed is a data point
SNL doesn’t write impressions of people the audience has to look up. The whole mechanism of a Weekend Update character depends on a critical mass of viewers going “oh, that guy” within about four seconds. When the writers decide a specific AI executive is a recognizable type — recognizable enough to justify the wig budget — that tells you the general public has moved past “AI companies” as an abstract category and started sorting the individual people inside them.
That’s new. For most of the past few years, the public read on this industry was one undifferentiated blob of men promising the future. Now one of them has a bit. That’s the kind of cultural absorption you can’t buy with a keynote.
The essay that earned it
The material writes itself because Amodei handed it over. He published a 3,800-word essay on a Saturday calling for a global slowdown of AI development on safety grounds. Thirty-eight hundred words. On a weekend. About slowing down the thing his company builds.
You do not have to agree with a single line of that essay to see why a comedy room circled it. The setup is inherently funny in the structural sense: a man runs a frontier AI lab, and he is also the person most publicly worried about frontier AI labs. Comedy runs on contradiction held in one body, which is exactly why the two-versions-of-the-same-man staging works better than a straight impression would have. The joke isn’t “this guy is scared.” The joke is that both answers are his.
What I’d tell you if I were reviewing the person instead of the tool
Here’s where I’ll be less charitable than the average tech writer covering this. Being spoofed for your safety warnings is a strange kind of prize. It converts a serious position into a character trait. Once “worried about AI” is a recognizable costume with a wig attached, the position gets cheaper to dismiss. Not refuted — dismissed, which is worse, because there’s no argument to answer.
That’s the risk for anyone in this industry who talks publicly about risk. Do it long enough and clearly enough and you stop being a participant in a debate and start being a bit. I don’t think that’s fair. I do think it’s what happens.
There’s a second thing, and it’s the part that actually matters for people who use these tools. Warnings are cheap to publish and expensive to act on. I can read an essay about slowing down; Those are the questions I get paid to answer, and they’re settled by testing, not by prose length.
So my read on the SNL segment is split too, appropriately enough. On one hand: good. An executive said something specific and unpopular about his own field, loudly enough that a comedy institution noticed, and that’s more than most of his peers have risked. On the other: a sketch is not accountability. It’s recognition. Those feel similar from the inside and are not remotely the same thing.
What this actually changes
For you, as someone deciding which AI tools to put in your workflow, nothing. Zero. The wig doesn’t ship features. What it does change is the conversation around those tools, and conversations tend to drift toward whatever framing is easiest to repeat. If the easiest available framing of AI safety becomes “the guy SNL made fun of,” that’s a loss for everyone who wants the technical arguments taken seriously — including the reviewers who have to explain why an agent quietly deleted a folder.
Watch the sketch. It’s funny, the two-versions structure is a genuinely sharp piece of writing, and Wickline commits. Then go back to reading the documentation, because that’s still where the useful information lives.
🕒 Published: