Uncanny Valley: This Week in Tech & Politics (Episode Transcript)
This week on WIRED’s Uncanny Valley, our team breaks down the rapidly escalating feud between AI firm Anthropic and the U.S. Department of Defense— and what the standoff reveals about how the U.S. government navigates partnerships with private tech companies. Later, co-host Zoë Schiffer explains why sorting people into “agentic” or “mimetic” categories has become Silicon Valley’s newest favorite personality litmus test for AI hiring. We also break down the biggest takeaways from President Trump’s 2025 State of the Union address, and bid a fond farewell to TAT-8, the undersea internet cable that laid the groundwork for the modern web we use today.
Articles mentioned in this episode:
Are You ‘Agentic’ Enough for the AI Era?
Say Goodbye to the Undersea Cable That Made the Global Internet Possible
You can follow Brian Barrett on Bluesky at @brbarrett, Zoë Schiffer at @zoeschiffer, and Leah Feiger at @leahfeiger. Send us your questions and feedback at [email protected].
How to Listen
You can stream this week’s episode directly via the audio player on this page. If you’d like to subscribe for free to get every new episode delivered automatically, here’s how:
On iPhone or iPad, open the Apple Podcasts app, or tap this link to subscribe.
You can also search for Uncanny Valley in podcast apps like Overcast or Pocket Casts, and you can find us on Spotify too.
Transcript
Note: This transcript was generated automatically and may contain minor errors.
Brian Barrett: Hey everyone, it’s Brian. Zoë, Leah, and I have absolutely loved stepping into the host role for you all these past few weeks, and we want to hear from you. If you enjoy the show and have a minute to spare, please leave us a review on whatever podcast app you use. It really helps new listeners find us. And as always, you can send any questions or thoughts our way at [email protected]. Thanks for listening, let’s jump into the show.
Leah Feiger: Hey, how’s everyone doing today?
Zoë Schiffer: I’m feeling great! How about you, Brian?
Brian Barrett: I’m feeling fantastic, and I know Leah is too— because Survivor is back tonight. It’s another one of our favorite things that, apparently, you don’t care about.
Zoë Schiffer: How do you know I don’t care? I mean… I don’t. I really don’t. Though my childhood best friend tried to get on the show once and didn’t make the cut, so it’s not totally irrelevant to me.
Leah Feiger: For the record, one day I’m actually going to apply. Both Brian and our colleague Tim have already promised me I can take a month off to go hang out on a beach in Fiji and still have my job when I get back.
Zoë Schiffer: Most people would be like “Leah would never last out there,” but they don’t know about your deep-sea diving skills.
Leah Feiger: I actually think I’d do just fine. I really, really want to do it. One day, guys, mark my words.
Brian Barrett: But Leah, that would mean you might have to kill some fish to eat them, which isn’t exactly something you…
Leah Feiger: That’s fine.
Brian Barrett: Oh, okay then.
Leah Feiger: No, no, fishing for survival is totally fine. What I have a problem with is the big industrial overfishing that’s massacring our oceans, not catching one fish to feed yourself.
Zoë Schiffer: And on that very cheerful note, welcome to WIRED’s Uncanny Valley. I’m Zoë Schiffer, WIRED’s director of business and industry.
Brian Barrett: I’m Brian Barrett, executive editor.
Leah Feiger: And I’m Leah Feiger, senior politics editor.
Zoë Schiffer: Okay, I think we have to kick things off today with this feud that’s escalating by the hour between Anthropic and the Pentagon. Let me set the stage: Anthropic landed a contract worth up to $200 million with the Department of Defense last summer. Speaking of which… are we renaming it the Department of War these days? No?
Brian Barrett: We’re sticking with Department of Defense— it’s technically only a secondary title per executive order, so we’re good.
Leah Feiger: A name change would require an act of Congress, that’s a whole thing. We’re fine with DOD, I actually love calling it DOD.
Zoë Schiffer: Great, before we get totally derailed by nomenclature. After Anthropic won this contract, things got really tense really fast, because the two sides have wildly different views on how DOD can use Anthropic’s AI tech. Anthropic has pretty strict rules for how its tools can be deployed: no domestic surveillance, no use for fully autonomous weapons, for example.
Leah Feiger: The fully autonomous weapons rule is not some crazy left-wing “woke” take, by the way. All Anthropic is saying is machines shouldn’t be allowed to push the kill button entirely on their own, right?
Zoë Schiffer: Exactly.
Leah Feiger: They can’t kill people without a human involved. That feels like a totally reasonable line to draw.
Zoë Schiffer: Right. But that stance didn’t fly with certain DOD leaders, specifically Defense Secretary Pete Hegseth. He met with Anthropic CEO Dario Amodei earlier this week and essentially gave him an ultimatum: the company has until Friday to cave to the Pentagon’s demands, which would let DOD use the AI however it wants with no restrictions, or DOD could cancel the entire contract.
Brian Barrett: Zoë, one of the most fascinating parts of this to me— beyond the DOD threatening to use state power to force a private company to let them use its AI for whatever they want— is that DOD already has other options. They have a contract with xAI, they work with all the biggest AI players, other companies don’t have these kinds of restrictions. Why is it so important to them that Anthropic specifically drops its rules? Can’t they just use Grok for whatever they need?
Zoë Schiffer: They already have a deal with xAI, yeah. My read on this— and I’m curious what you two think— is that this is mostly performative. They’re trying to make an example: any company that wants big government or military contracts can’t embed their own corporate values into those agreements. Basically, if you want our money, you do whatever we ask to get the job done.
Brian Barrett: Let me circle back to something I touched on earlier: to flex their power here, they’ve even floated potentially invoking the Defense Production Act. That’s a power normally used during wartime, right? Like, if we need more tires for Humvees, you force Goodyear to ramp up production. We saw it used during COVID to force manufacturers to make more masks. It’s almost always tied to physical goods and a formal state of emergency. This is just “we want to play with your fancy new AI and you won’t let us,” which is such an outrageous overreach of that power, if you ask me.
Zoë Schiffer: Yeah. I will say, this isn’t necessarily the worst position for Anthropic to be in, purely from a branding perspective. The company has always differentiated itself from competitors not just by its tech, but by its values-focused brand— it’s positioned itself as the more ethical, “holier-than-thou” AI option next to its rivals. So while hundreds of millions of dollars are on the line, and Anthropic does want to be a government contractor, this standoff is already cementing their brand as the company that’s willing to take a principled stand, which is something Sam Altman and OpenAI probably wouldn’t do right now.
Leah Feiger: What you’re saying lines up exactly with what Hegseth said when he announced the xAI partnership, actually. You guys have to hear this clip.
[Archival Audio: Pete Hegseth]: We will judge AI models on this standard alone— factually accurate, mission relevant, without ideological constraints that limit lawful military applications. Department of War AI will not be woke. It will work for us.
Zoë Schiffer: Right, that’s exactly the point. This reminds me of the “anti-woke AI” executive order the administration rolled out last year. I talked to dozens of people in the industry asking if they’d heard from the government about this, if they were being told to change their models, and almost every source told me no. That told me it was all for show, just to make a political point, not actually to change the products these companies build.
Brian Barrett: I was trying to think of a real-world analogy for this, if it wasn’t code. It feels like going to iRobot and saying “we need your Roombas to be bombs. We need exploding Roombas to deploy in the Kremlin.” That’s the level of overreach we’re talking about here, to me.
Leah Feiger: This is wild overreach, full stop. The news is breaking by the hour on this, new statements coming out from both sides every day. It feels like we’re heading straight for a breaking point. And so far— correct me if I’m wrong Zoë, this is your beat— we haven’t actually seen any big tech company stand up to the second Trump administration in any meaningful way yet.
Zoë Schiffer: Yeah, it’s way more normal for companies to just fall in line. But if Anthropic caves here, it would be a total disaster for their brand. They’ve positioned themselves as the “good AI company,” so the second they backtrack, everyone will call them a hypocrite immediately— and that’s not even counting the earlier drama over them considering taking investment from Gulf states. What’s more, I’ve heard from sources at OpenAI that they’re already poised to pounce on this, because they’re sick of Anthropic’s executives running around saying they’re better than OpenAI because they have principles and OpenAI doesn’t. I’m really curious how this plays out, but in my view, Anthropic really only has one good move here.
Brian Barrett: There’s a timely little detail I wanted to share that ties right into all this. Just this Wednesday, a story came out: a researcher named Kenneth Payne from King’s College London pitted GPT-5.2, Claude Sonnet 4, and Gemini 3 Flash against each other in hundreds of war game simulations. Ninety-five percent of the time, the AI models chose to use nuclear weapons.
Zoë Schiffer: No way.
Brian Barrett: They couldn’t get enough of nukes, honestly. Now that doesn’t mean that’s what would happen in a real crisis, but when DOD is complaining about “woke AI” that won’t do what they want… I guess we’ll find out if not nuking people is woke now.
Leah Feiger: That’s the title of Brian’s new book, out Fall 2027—
Brian Barrett: Yep.
Leah Feiger: —Woke AI Won’t Nuke Your Enemies.
Zoë Schiffer: I’d read that, I’d click that for sure.
Leah Feiger: Wait, but let me walk through this for a second, as someone who doesn’t work with AI models every day. That actually sort of makes sense to me, right? If you tell an AI “win this conflict by any means,” of course it’s going to conclude the fastest, most definitive end to the conflict is just using nukes. It checks all the boxes for “win.” Does that actually check out? Why would the models even reach that conclusion?
Zoë Schiffer: Yeah, that’s exactly what AI alignment is supposed to prevent. The whole point of alignment is to train the model to align with human values, not just optimize for its own goal. But these days, alignment itself has gotten labeled as “woke,” so companies are only implementing the bare minimum of guardrails they can get away with, because it’s politically toxic to do anything more.
Brian Barrett: And when we say “align with human values,” the big question is always whose values, right? That’s the whole crux here. The Pentagon is saying “no, alignment has to be with DOD’s values, full stop.”
Zoë Schiffer: Exactly. I have a question for you two, let me just throw it out there then unpack it.
Brian Barrett: Hit me.
Zoë Schiffer: Do you think you’re agentic or mimetic?
Brian Barrett: Wait, no, take it back, I take the question back.
Zoë Schiffer: Is that because you’re mimetic?
Brian Barrett: Can you explain what those even mean before I answer?
Zoë Schiffer: Right! Silicon Valley is completely obsessed right now with being “agentic,” hiring only agentic people, and it’s the direct opposite of being mimetic. An agentic person is action-oriented: they just get stuff done, they have internal drive, they make things happen, they don’t second-guess themselves or the process. A mimetic person, on the other hand— and I think we can all agree I fall into this category— they’re more cautious, they weigh pros and cons, they wait to see what other people do before they make a move. These questions are popping up all the time in job interviews at AI labs right now. They’re trying to vet candidates: are we hiring someone agentic, or mimetic? The thinking goes that in a world where AI agents take over huge chunks of the economy, agentic people will get ahead and mimetic people are out of luck.
Leah Feiger: This just feels like Silicon Valley rebranding something that’s existed forever. It’s like asking if you’re a go-getter or a people-pleaser, these conversations have been around since we were kids on the playground. Of course Silicon Valley took a basic BuzzFeed personality quiz from college, added two fancy new buzzwords, and suddenly they’ve “invented a new way to rank people.”
Brian Barrett: And every single person who buys into this classification thinks of themselves as agentic, right?
Leah Feiger: No question.
Zoë Schiffer: Oh, 100% sure of that. But here’s what I find funny about this whole trend. Our great AI reporter Maxwell Zeff wrote about this for WIRED this week. This whole conversation blew up after a viral essay in Harper’s by Sam Kriss, which profiled a few people that embodied agentic traits and talked about this new cultural shift in Silicon Valley. My big takeaway is that media people tend to overinflate the importance of niche internet figures like Slate Star Codex and the founder of Cluely, and the average rank-and-file tech worker has no idea what they’re even talking about. These people are totally irrelevant to most of the industry. At the end of the day, this is just “we like hiring people who get stuff done,” which… don’t we all? I’m a little confused what the big revelation is here.
Leah Feiger: Are you a self-starter, Zoë? That’s all this is. Ugh, seriously.
Brian Barrett: This also isn’t the first time Silicon Valley has gotten obsessed with sorting people into silly new categories. Who can forget the frenzy in early 2022 when everyone was either a wordcel or a shape rotator? Does anyone remember that, or am I the only one?
Zoë Schiffer: I genuinely think you’re making this up to test me. I have no idea what you’re talking about.
Brian Barrett: That’s exactly what a wordcel would say.
Zoë Schiffer: Oh no, I’m mimetic. And a wordcel. Two for one.
Brian Barrett: Yeah, they’re actually pretty similar concepts, similar breakdown. Wordcels are— I might get this wrong, it was a while ago— but as the name suggests, they’re people who work more with language, talk through problems, whereas shape rotators… they’re all action, just get stuff done.
Zoë Schiffer: Where do those words even come from?
Brian Barrett: I have no idea. I think wordcel is derived from incel as a pejorative, but I’m just guessing there. I don’t remember the exact origin, but even Sam Altman was tweeting about wordcels and shape rotators back in the day.
Zoë Schiffer: Well, according to that Harper’s essay, Sam Altman is chronically online, for what it’s worth. The essay has a whole section about an X/Twitter troll who went after people, made fun of them, demanded they buy him stuff. Sam Altman made the whole thing mainstream when he actually bought the troll a gaming laptop and sent it to him. Then all these other big name tech guys felt like they had to buy this random guy stuff too, and it became a whole trend.
Leah Feiger: I hate that this is a cultural talking point now. This is just one more silly way to divide people, that’s all it is.
Zoë Schiffer: Right? It’s like when people put “pro Oxford comma” in their Twitter bio. Who cares?
Brian Barrett: Wait, that means you’re wrong, the Oxford comma is great.
Zoë Schiffer: No, no, I like the Oxford comma too! I just think it’s weird to build your personal brand around it. That’s all.
Brian Barrett: Wow.
Leah Feiger: She hates the Oxford comma, and she’s also mimetic.
Brian Barrett: She’s a wordcel.
Leah Feiger: And she’s a—
Zoë Schiffer: Sorry guys, moving on!
Brian Barrett: She’s a mimetic wordcel.
Leah Feiger: Okay, let’s move on to the other big news this week: President Trump’s State of the Union address, delivered Tuesday night.
[Archival Audio: Donald Trump]: Members of Congress and my fellow Americans, our nation is back— bigger, better, richer, and stronger than ever before.
Leah Feiger: It was the longest State of the Union in U.S. history, clocking in at almost two hours, and it was exactly what you’d expect. Trump bragged about his first term performance, doubled down on his hardline immigration stance, claimed the U.S. economy is booming,
