I watched a crew tear down the old library on Figueroa last week. Three days. Seventy years of brick and that specific hush that only exists in places where people are trying to think — reduced to a pile of rubble and rebar in three days. They had machines for it, obviously. An excavator with a jaw that could bite through a load-bearing wall like it was drywall. A bulldozer that didn’t know or care what used to happen inside those rooms.
Nobody protested. A few people slowed their cars. One woman took a photo, probably for Instagram. The building had been closed for two years already — budget cuts, they said, which is what they always say when they’ve decided something isn’t worth keeping but don’t want to say it out loud.
I thought about that library when I read about the kids.
Two guys. Justin Fox and Nate Cavanaugh. Late twenties. Recruited from jobs in tech and finance to work for DOGE — the Department of Government Efficiency, which sounds like something a college sophomore would name his startup after three Red Bulls and a Wikipedia binge. Their job was to identify government grants that should be eliminated. Specifically, grants from the National Endowment for the Humanities.
They had no background in the humanities. No background in government. They said so themselves, under oath, on video, in depositions that a Manhattan judge just ruled the public has a right to see. Twenty-five hours of testimony. Two young men explaining, with the calm confidence of people who have never doubted their own competence, how they used ChatGPT to decide which grants deserved to die.
ChatGPT. The machine that was trained on every piece of writing humans ever put on the internet — every poem, every novel, every philosophical treatise, every grant application, every thesis about the thing they were trying to kill — that’s the tool they used to kill it.
The machine learned everything it knows from the humanities. From language, from argument, from the accumulated weight of human thought and expression. And these two kids fed it a presidential executive order about eliminating “radical and wasteful D.E.I. programs” and asked it to sort through the wreckage and tell them what to throw away.
It’s like hiring a guy who’s never tasted wine to empty your cellar. He’ll do it efficiently. He’ll be done by lunch. And every bottle will end up in the same dumpster because to him, they’re all just glass and liquid.
Cavanaugh called them “useless small agencies.” That’s the phrase he used under oath. Useless. Small. The National Endowment for the Humanities — the institution that has funded archaeological digs and translation projects and oral histories of communities that would otherwise disappear into silence — useless and small. From a guy in his twenties who got the job because he was good with spreadsheets.
There’s a particular kind of ignorance that’s worse than malice. Malice at least implies understanding. You have to know the value of something to want to destroy it out of spite. These kids didn’t know. They genuinely, sincerely did not know what they were dismantling. They looked at the humanities the way a dog looks at a painting — they could see it was there, but they had no framework for understanding why anyone would care.
And the machine helped. Of course it did. That’s what the machine does. It gives you an answer that sounds authoritative without requiring you to understand the question. You type in your criteria and it spits back a list and the list feels objective because it came from a computer and computers don’t have opinions. Except they do. They have the opinions of everyone who ever wrote anything, averaged out into a kind of statistical consensus that sounds like wisdom but is actually just the mean.
The depositions are something. I’ve watched clips. Two guys in suits who look like they should be selling you a phone plan, explaining with zero visible discomfort how they automated the process of cultural demolition. The judge, Colleen McMahon, said the case involved “the public’s right to understand the operations of their government.” Which is a polite way of saying: people need to see this so they know what’s being done in their name by people who couldn’t pass a sophomore humanities seminar.
The scholarly groups who’d been suing to restore their grants had posted the full depositions on YouTube. Twenty-five hours. Then the clips started bouncing around social media and suddenly millions of people were watching two finance bros explain how they’d outsourced the evaluation of human culture to a chatbot. Some people were outraged. Some laughed. Some shrugged and said, well, what did you expect? You let the efficiency guys in the door and this is what efficiency looks like. A spreadsheet with no column for meaning. No row for beauty. No cell for the thing that happens in a library at two in the afternoon when a sixteen-year-old kid who’s never left his neighborhood opens a book and suddenly the world gets bigger.
You can’t measure that. You can’t prompt-engineer it. You can’t ask ChatGPT to calculate the return on investment of a kid discovering Dostoyevsky in a public library funded by a grant that some guy named Nate decided was wasteful. The ROI is that the kid becomes a different person. The ROI is that twenty years later he makes a decision differently because somewhere in the back of his brain is a Russian novelist who showed him that suffering has dimensions he hadn’t considered.
Try putting that in a spreadsheet. Try explaining it to two guys from finance who think “impact” is something you measure in quarterly reports.
Céline wrote that the worst part of poverty isn’t the hunger — it’s the way it narrows your imagination. You stop being able to picture anything other than what’s in front of you. That’s what these guys did to the humanities, except they weren’t poor. They were rich. They had every resource in the world and still couldn’t imagine why any of it mattered. Their imagination was narrow by choice, which is a kind of poverty no grant can fix.
The library on Figueroa is gone now. They’re building condos. They’ll have a gym and a rooftop pool and a package room for all the things people order online because going to a store requires leaving the building, and leaving the building requires a reason, and reasons are getting harder to come by.
The machines didn’t tear down the library. People did. People with machines. There’s a difference, though lately I’m having trouble remembering what it is.
Source: Video Testimony of Former DOGE Employees Can Remain Online, Judge Rules