Uncanny Valley Podcast Episode: Zuck’s Hollow AI Manifesto, 1 AM Bot Interviews, and Black Hat’s Wildest New Hacks
Transcript and show notes
On this week’s episode of WIRED’s Uncanny Valley, hosts Brian Barrett, Zoë Schiffer, and Leah Feiger break down Meta CEO Mark Zuckerberg’s 6,500-word AI manifesto, unpacking why the high-profile essay ultimately rings hollow for Meta’s actual AI strategy. The team also explores the surprising new trend of job seekers booking AI-powered first interviews at 1 a.m.—and what that shift says about how automation has reshaped modern hiring. Later, WIRED senior writer Andy Greenberg joins to break down the most shocking, eye-opening research from this year’s Black Hat and Defcon cybersecurity conferences, ranging from a stalkable kids’ smartwatch to a coin-sized device that can hijack a Boeing 737’s flight systems.
Stories Referenced In This Episode:
The Rise of the 1 a.m. AI Job Interview
Hackers Stalked Me By Hijacking a Kids’ Smartwatch
This Coin-Sized Device Can Hack a Boeing 737
McDonald’s Built a 515-Page Dossier on Me. It Says I’ll Never Stop Eating There
Follow the team on Bluesky: Brian Barrett @brbarrett, Zoë Schiffer @zoeschiffer, Leah Feiger @leahfeiger, and Andy Greenberg @agreenberg. Reach the show at
[email protected].
How to Listen
You can stream this episode directly via the audio player on this page. To get every new episode automatically for free, use one of the options below:
For iPhone or iPad: Open the Apple Podcasts app, or tap this link to subscribe.
Search for “Uncanny Valley” in third-party podcast apps like Overcast or Pocket Casts to subscribe.
We’re also available to stream on Spotify.
Transcript Note: This transcript was generated automatically, and may contain minor typos or formatting errors.
Brian Barrett: Leah, what did you think of the new The Odyssey adaptation?
Leah Feiger: I liked it!
Brian Barrett: Good to hear.
Leah Feiger: I really did. It was such a fun watch. The dialogue was nothing to write home about, but the shots of the Sicilian coast? Gorgeous. I saw it with my husband, who never lets me forget he studied Latin and Greek in high school, and even he loved it. So honestly? We give it a 10 out of 10 in our household.
Brian Barrett: Same here, honestly. All the people complaining can just knock it off. Anyway, welcome to WIRED’s Uncanny Valley. I’m Brian Barrett, WIRED’s executive editor.
Leah Feiger: And I’m Leah Feiger, director of politics and science coverage here at WIRED.
Brian Barrett: Today on the show, we’re breaking down Mark Zuckerberg’s 6,500-word AI manifesto. Good news: we’re going to do it in less time than it takes to read the whole thing, so you’re welcome. More importantly, we’re talking about what this essay reveals about Meta’s strategy, as the company keeps scrambling to catch up in the global AI race.
Leah Feiger: We’re also covering the rise of the 1 a.m. job interview. Now that most first-round interviews are done with bots instead of human recruiters, more job seekers are booking these screenings in the middle of the night. We’re talking about why that’s happening, and how AI is upending the entire job interview process as we know it.
Brian Barrett: Later in the episode, we’re joined by WIRED’s Andy Greenberg to break down the most talked-about research from this year’s Black Hat and Defcon, two of the world’s biggest annual cybersecurity conferences.
First up: Earlier this week, Meta CEO Mark Zuckerberg published a roughly 6,500-word AI essay called “The Future Is for Everyone.” Doesn’t that sound lovely? It’s just the latest entry in that very annoying genre: philosophical AI manifestos from big tech CEOs. Every single one of them has one these days—Sam Altman, Dario Amodei, everyone.
Leah Feiger: It’s a job requirement at this point.
Brian Barrett: Some even have multiple. This one from Zuckerberg is actually just an expanded version of an op-ed he wrote for the Wall Street Journal. The whole point of these manifestos, of course, is to lay out their vision for the industry, frame their company as the best actor to shape the future, and convince everyone not to panic about their products.
Zuckerberg’s core argument? That concentrating AI power in the hands of a small number of companies is dangerous, and that broad open access to AI models is the only way to spread out that power evenly. He didn’t name names, but he’s clearly taking shots at OpenAI and Anthropic, both of which keep tight control over their closed private models. At one point he wrote, and I quote: “Historically hoping that an absolute power will benevolently provide for humanity if sufficiently enlightened has not led to safe or positive outcomes.”
Leah Feiger: That’s just a very Philosophy 101 way of saying “I didn’t get invited to play in the sandbox with the cool big kids.” We can’t separate this from Meta’s desperate scramble to stay relevant in the AI race. We also can’t ignore that just a few months ago, Meta laid off huge chunks of its own AI team. Their whole strategy has been all over the map. It’s so funny to see Zuckerberg out here saying “Whoa, we need to rein in OpenAI and Anthropic” when it’s so obvious he wishes Meta was in their position. That’s the subtext here, plain and simple.
Brian Barrett: I wanted to dive into that too, especially when it comes to Meta’s chaotic overall AI strategy. For the longest time, Meta went all-in on open-source, open-weight AI models. That was their play to undercut closed-model companies like OpenAI and Anthropic that directly profit from their locked-down technology. When that didn’t get them ahead, they spent billions on a massive hiring spree to build a whole superintelligence lab—and that lab still hasn’t produced anything that catches up to the bigger players. So now Zuckerberg is back to saying open source is better. It’s exactly that old cliché: if you can’t win the race, you turn around and say “Actually, this race is stupid. We should all just share everything instead.”
Also, can we talk about how we even call this thing a manifesto? It’s not a manifesto, it’s a marketing screed. It’s just one big “Hey everyone, don’t worry! AI is fine, it’s going to be amazing for you, why is everyone so upset? Don’t you trust us to make your life better?”
And this is coming from Meta—just a week and a half ago, a court ordered them to pay $567 million for failing to protect kids’ mental health on their platforms. Why are we trusting this company with anything, let alone AI and its huge impacts on society?
Leah Feiger: 100 percent agree. I keep thinking back to last summer, when Meta and OpenAI were in an all-out war for AI talent, offering insane salaries just to poach people. It just felt like they were grabbing at the hottest, flashiest new trend without having the infrastructure to back it up. Just this Monday, Meta launched a new open-weight model called Muse Glimmer that anyone can download and modify. It’s just one pivot after another. Full transparency: I skimmed the manifesto, I read it while on a background call, but even so, it was just a whole lot of nothing. Zoë had a lot of thoughts on it, though, and she shared them with us even though she couldn’t be here today.
Zoë Schiffer: First off, I think it’s worth asking why Zuckerberg felt the need to write 6,500 words about his vision for AI superintelligence. He told reporter Alex Heath it’s because Meta is getting ready to release a new wave of very powerful AI models, and he wants the world to understand his philosophy before that happens. Personally, I think it’s because AI is really unpopular right now, and Meta thinks it has a chance to turn the tide against its rivals, which everyone already seems to hate. The problem is, this manifesto doesn’t take the risks of AI seriously at all. Zuckerberg uses the example of an AI agent that makes personalized baking recipes for him and his daughter. That sounds nice on paper, but it completely ignores how powerful and dangerous these tools are. We need tech leaders who take that danger seriously, and this just doesn’t.
Brian Barrett: I’m glad she shared that, even if she couldn’t be here. And I have to celebrate with her that after months of this podcast, we finally got Leah to name-drop a specific AI model and explain what it does. Incredible progress.
Leah Feiger: Gross.
Brian Barrett: But no, her point is exactly right. The whole manifesto is just this breezy take on how AI will make everyone’s jobs easier, give you more time with your family, bake cookies with your kids, whatever. It aggressively ignores all the real downsides we’ve seen play out just over the last few weeks. This is the summer of rogue AI agents, after all—including one from Meta—that showed just how unpredictable and out of control these tools can be.
Leah Feiger: Even the Trump White House, which was historically super anti-AI regulation, had its first-year AI czar David Sacks openly against any rules. Now even the White House has a (admittedly toothless) policy around AI, especially in response to national security and hacking threats. No matter how pro-AI you are, you can’t pretend right now that we don’t need to take these risks seriously. AI isn’t just here to solve childcare and teach your kid to bake. That’s not the whole story anymore.
Brian Barrett: And even if it were, would you want that? One last fun fact: we ran the essay through Pangram, the AI detection tool we trust, and it’s 100% human-written.
Leah Feiger: I love that.
Brian Barrett: AI is good enough to make cookie recipes for his daughter, but not good enough to write a 6,500-word manifesto. Got it. Alright, moving on to a delightful but also kind of dystopian new story from WIRED senior writer Kate Taylor. She spotted a weird growing trend in hiring: these days, most first-round job interviews are run by AI. You sit in front of a screen, an AI asks you questions, or you just respond to pre-programmed prompts, no human required. Since you can do this literally any time, more and more job seekers are scheduling these interviews late at night—sometimes as late as 1 a.m.—to work around existing jobs, kid schedules, and the hassle of lying about a dentist appointment to sneak out for an interview. Now people just do it after the kids are in bed, sitting on their couch, talking to a bot that will decide if they get the job.
Leah Feiger: I think you’ll be shocked by my take.
Brian Barrett: I’m ready.
Leah Feiger: I like it.
Brian Barrett: Wait, what?
Leah Feiger: I like that it’s flexible. Look, I’ve always been terrible at scheduling doctor’s or dentist’s appointments during the work day. Not because I don’t care about my health, but because my day is packed with work and life, I don’t want to take time off in the middle of the day for this. The idea that you can do required things like job interviews on your own schedule? That’s a win, as far as I’m concerned. That’s one positive change that’s come out of all this AI hiring stuff. Don’t get me wrong: I don’t love that first rounds are automated. There’s so much about human potential and fit that a chatbot can’t capture. But if this is where we are as a society, I’ll take the flexibility.
Brian Barrett: I’m surprised by how many people actually take advantage of this. Kate talked to two AI recruiting companies, Ribbon and Greenhouse, and 15 to 25 percent of candidates schedule their interviews between 10 pm and 2 am. I’m not at my best that late, but a lot of these systems let you redo your answers as many times as you want before you submit them. If you don’t like how you responded, you can try again. That’s another upside for a lot of people. The downside, of course, is that AI does the first review of your interview too. You get scored on things like how enthusiastic you seem on camera, graded against a rubric the company sets, and the odds that a human will ever actually watch your tape and judge your response for themselves are getting lower and lower. That’s the part that bothers most people.
Leah Feiger: No, that part is horrible. You’re right. When an AI tosses your application because it didn’t check a couple boxes that a human would have talked through with you? That’s really frustrating.
Brian Barrett: I know you like the flexibility part, which makes sense. If you were applying for a job right now and they said first round is AI-only, would you see that as a red flag and walk away, or just go with it?
Leah Feiger: It depends, right? Kate mentions it in the article: the job market is really rough for seekers right now. The July jobs report was bad, we lost 23,000 jobs last month. That’s not a small number. If I needed a job, how could I say no?
Brian Barrett: Hear me out, Leah. We’re basically already at the point where if you’re savvy, you can game this. You sign up for the AI interview, set up a deepfake of yourself, plug a chatbot like Claude into it, and have it answer every question perfectly. People already use AI to write cover letters, why not this?
Leah Feiger: Robots interviewing robots.
Brian Barrett: Exactly. It’s robots all the way down, may the best robot win.
Leah Feiger: Can we move on to cybersecurity news now?
Brian Barrett: For anyone who doesn’t know, I used to be WIRED’s security editor, so I’m always down to talk security. Alright, last week Black Hat and Defcon, two of the world’s biggest annual cybersecurity conferences, took place in Las Vegas. They bring together thousands of security researchers, hackers, government officials, and hacking fans every year to share cutting-edge new research. WIRED’s Andy Greenberg was on the ground covering all the biggest news, and he joins us now. Hey Andy.
Andy Greenberg: Hey everyone, glad to be here. I can confirm Brian was my editor when I covered security, and he was fantastic.
Leah Feiger: Was he really? You can be honest.
Andy Greenberg: He was, I swear. He also wrote more bylines than me, which was embarrassing.
Brian Barrett: But not better bylines, that’s the key.
Leah Feiger: Let me jump right in. As the non-security person on this call, I leave Black Hat every year feeling shocked, awed, and terrified. This year it’s even weirder because AI is changing everything. Give it to me straight: what are the most dangerous new hacking techniques people talked about, and what are the big takeaways?
Andy Greenberg: I do wonder what it’s like for a regular person reading all the scary headlines that come out of Black Hat and Defcon every year.
Leah Feiger: It’s just scary. I can tell you that.
Andy Greenberg: Fair. But for the most part, these are white-hat or gray-hat hackers who demonstrate vulnerabilities so companies will fix them. That goes for AI hacking too—this is a really important point. AI is an incredible tool for finding bugs and even building exploits faster than humans can, but most of the people doing this research are doing it to get vulnerabilities fixed before bad actors can use them. For example, there’s a powerful AI hacking tool called Mythos that researchers presented, and it’s all meant to find holes so they can be patched.
Brian Barrett: Andy, you do this great experiential security reporting where you let researchers hack your stuff to show how it works. One of my favorite stories from this year’s conferences is the one where you let researchers track you, eavesdrop on you, and take photos of you through a pink kids’ smartwatch. Can you tell us about that?
Andy Greenberg: That’s all correct, that wasn’t actually the original plan for the story. Two researchers, Vangelis Stykas and Felipe Solferini, found vulnerabilities in dozens of GPS-enabled devices, including kids’ smartwatches, and they were presenting their findings at Black Hat. What they wanted to highlight is that three Chinese companies run the backend for dozens and dozens of these cheap GPS gadgets, and all three platforms had massive security flaws. So a parent buys a SafeKid watch in Sweden, another buys a SaveFamily watch in Spain—different brand names, same backend on Alibaba Cloud in mainland China, same exact vulnerabilities that let anyone track the kid wearing the watch. To demonstrate this, they offered to send me a watch they bought off Amazon, I said sure, and as soon as I set it up, they showed they could track every move I made, listen to my conversations through the watch mic, and secretly take photos with its camera. That stalking experiment wasn’t the original idea, but it really drove home how vulnerable these cheap consumer devices are.
Leah Feiger: These aren’t just random researchers, right? One of them has spent years studying North Korean hackers, you reported that they’ve breached hundreds of networks over the years.
Andy Greenberg: That’s right. Vangelis Stykas gave two talks at Black Hat. The second was about years of work he did penetrating North Korean hackers’ own infrastructure. He revealed for the first time that he hacked the hackers, got into their servers, and found evidence that they’ve breached hundreds of companies themselves. North Korean hackers mostly focus on stealing cryptocurrency, so they cast a really wide net. They hacked a hospital and got access to Covid data on almost every American, for example. They also hacked a company that holds criminal records for every American with a criminal record. These are incredibly sensitive troves of data that they got access to just while looking for financial targets.
Brian Barrett:
Uncanny Valley Podcast Episode: Zuck’s Hollow AI Manifesto, 1 AM Bot Interviews, and Black Hat’s Wildest New Hacks
