The Screenwriter's Fallacy

David Shapiro argues that the AI story has entered its "Messy Middle" — Act Two, in screenwriting terms — where complications replace the comfortable hypotheticals of Act One. The Anthropic-Pentagon confrontation is his inciting incident. The hero has been thrust into the new world.

He's right about the observation and wrong about the frame.

What's actually happened is straightforward: AI stopped being a technology story and became a power story. The Trump administration labelling Anthropic a national security threat, the "woke AI" rhetoric, the deregulation sprint — none of this is about what the models can do. It's about who gets to decide what they're used for, and who profits from that decision. That shift is real, and Shapiro spotted it earlier than most.

His concept of "identity stacking" is the most useful thing in the piece. The idea that having any opinion about AI now comes bundled with an entire political identity — pro-AI means MAGA, pro-safety means woke — is corrosive. It makes it nearly impossible for a normal professional to say "I think this tool is useful for my finance team but the vendor's claims are overblown" without someone slotting them into a tribe. That's the actual problem most working people face right now, and almost nobody in the AI discourse is talking about it clearly.

But the three-act structure is doing real damage to the analysis. Stories have protagonists and resolutions. What Shapiro is describing — overlapping institutional power struggles with no central character and no obvious endpoint — is the opposite of a story. It's a condition. Calling it "Act Two" implies an Act Three where things click into place. There is no evidence that's coming. This might just be what it looks like from here on out: permanently contested, structurally unresolved, no catharsis.

And then the ending undercuts everything. After a thousand words warning about narrative polarisation and tribal thinking, he closes with "imagine how much worse it's going to get as AI surpasses humans across all cognitive dimensions." That's not analysis. That's the hype cycle in a trench coat. The honest version — capabilities are advancing unevenly, specific jobs are already affected, nobody knows the ceiling — is less cinematic but more useful.

The piece is worth reading for the middle. Skip the metaphor and the finale.