Ivy: A nearly seven hundred percent spike in security alerts... and it's mostly just noise.
Marcus: I'm Marcus.
Ivy: And I'm Ivy.
Marcus: And this is AI Unfiltered for Monday, September 14th. We've got a lot to get to today.
Ivy: On the show: The AI boom is creating a crisis for security teams just trying to sort signal from noise.
Marcus: Then, we go inside the secret Meta project to replace thousands of workers with AI... and why it was a spectacular failure.
Ivy: And we'll ask why a top VC is zigging while everyone else zags—betting against the AI giants.
Marcus: Alright Ivy, let's dive into this idea of AI's double-edged sword in the workplace.
Ivy: It’s less a double-edged sword and more a firehose of nonsense. A new report from The Hacker News found that companies rolling out AI tools are seeing a 685% spike in security alerts.
Marcus: Okay, but isn't that a good thing? The tools are working, they're flagging more threats.
Ivy: Are they, though? The report says the vast majority are false positives. It’s not finding more needles; it’s just making the haystack exponentially bigger.
Marcus: So you get alert fatigue. Security teams are just overwhelmed.
Ivy: They're drowning. Imagine your inbox going from 10 emails a day to nearly 700, and all but five of them are junk. That's what's happening to security ops centers right now.
Marcus: The report calls it the new 'signal-to-noise problem'. So what's generating all that noise? Is it just badly configured AI?
Ivy: Partly. It's employees using generative AI for everything—writing code, summarizing docs, hitting APIs. Each one of those actions looks weird to older security models, so bang, you get another alert.
Marcus: But that's how people work now! We have to use these tools to keep up.
Ivy: No one's saying don't use them. But rolling them out without updating your security to actually understand AI-native behavior? That's just asking for trouble. It's corporate malpractice.
Marcus: Ouch. So what's the fix? Let me guess... more AI? An AI to watch the other AIs?
Ivy: That's the pitch, of course. New platforms that can contextualize these alerts. But that's just another tool, another line item on the budget.
Marcus: So if you're a leader, just handing out AI licenses isn't a strategy. You have to actually plan for the chaos it creates downstream.
Ivy: Exactly. Otherwise you're just paying for a very expensive, very noisy distraction.
Ivy: Speaking of chaos, Startup Fortune is out with a wild story about a secret project at Meta to replace thousands of its own staff with AI.
Marcus: Okay, that sounds like a movie plot. What was the actual plan?
Ivy: Efficiency, of course. The project was codenamed 'Agora,' and it aimed to automate huge chunks of ad operations, content moderation, even internal HR.
Marcus: Wow. So not just automating little tasks, but entire jobs. Basically creating AI agents with specific roles.
Ivy: Thousands of them. The idea was for AIs to handle routine ad campaigns and first-pass content reviews, only escalating the tough stuff to humans.
Marcus: So... what went wrong? Why'd it fall apart?
Ivy: Humans, Marcus. Turns out, we're messy.
Marcus: Come on.
Ivy: The report calls it 'deep human friction.' The managers who were supposed to oversee these AI agents basically revolted. They didn't trust the AIs, and they spent more time correcting mistakes than it would've taken to just do the work themselves.
Marcus: So it wasn't a tech failure, it was a workflow failure. A massive human-computer interaction problem.
Ivy: Exactly. The AI could do the task in a vacuum, but it couldn't handle the nuance, the exceptions, the 'hey, can you just do this one quick thing for me?' that makes an office run.
Marcus: So Zuckerberg's dream of a hyper-efficient, automated workforce just... hit a wall.
Ivy: It was quietly dismantled. It proves you can't just drop an AI into a human system and expect it to work. You have to redesign the system, and you need the people to actually buy in.
Marcus: And if they don't, they'll just work around it until it's useless. The ghost in the machine was the ghost of the old org chart.
Marcus: Let's switch from building to funding. There’s a great interview on TechCrunch with Deven Parekh from Insight Partners. While everyone's pouring money into OpenAI and Anthropic, he’s going in a different direction.
Ivy: Which seems like the only sane thing to do during a hype cycle, doesn't it?
Marcus: Maybe, but the conventional wisdom says foundation models are a winner-take-all market. You have to back the biggest horse!
Ivy: Parekh clearly disagrees. Insight has ninety billion dollars under management, and they're explicitly not just betting it all on the big guys. They're investing in smaller, specialized players, even rivals.
Marcus: Right, he even mentions losing a deal to a competitor and seems totally fine with it. He's betting the whole ecosystem is going to be big enough for lots of winners.
Ivy: He's playing classic portfolio theory. Why own a tiny slice of one potential mega-winner when you can own a big piece of ten other companies that could lead their own categories?
Marcus: So he thinks the real value isn't in the big, general models, but in the application layer—the companies that use the models to do specific things.
Ivy: Exactly. The picks and shovels. He's happy if his companies use OpenAI, or Anthropic, or Cohere, or all of them. He's not betting on the power company, he's betting on all the things people build with the power.
Marcus: That feels very... old-school VC. Almost contrarian for today.
Ivy: It's only contrarian if you think two companies will own intelligence. It's common sense if you think AI is more like the cloud—a foundational layer. You don't just invest in Amazon Web Services, you invest in the thousands of companies built on top of it.
Marcus: So for any founders or investors listening, the message is: don't get blinded by the foundation model arms race. There's huge value in the layers above.
Ivy: And for the big VCs, it's a reminder that you don't have to catch every unicorn. Sometimes it's better to own a stable of very healthy horses.
Marcus: Alright, so to recap: Enterprise AI is creating a firehose of useless alerts for security teams.
Ivy: Meta's secret plan to replace its staff with AI crashed and burned against the messy wall of human reality.
Marcus: And one of the world's biggest VCs is betting on a diverse AI ecosystem, not just one or two giants.
Ivy: Before we go, Marcus, I found the perfect gadget for you. A company just launched an 'AI-powered smart salt shaker.'
Marcus: [laughs] Okay, I'm listening. Does it analyze my food's sodium content in real time?
Ivy: It claims to. You shake it over your food, a micro-camera does 'flavor-profile analysis,' and it dispenses the 'algorithmically perfect' amount of salt. But apparently, it just... dumps a ton of salt on everything.
Marcus: So it's just a really expensive, broken salt shaker. [laughs] I kind of want one.
Ivy: Of course you do. That's our show for today. We'll be back tomorrow with more.
Marcus: Until then, go easy on the algorithmically perfect salt. See you next time!
This show is made with AI: the hosts’ voices are synthetic and the scripts are AI-assisted. Every story links to its original source.