Ivy: The glasses are recording even when the little red light isn't on.
Marcus: And that's just the start. I'm Marcus.
Ivy: And I'm Ivy. It's Saturday, October 31st, 2026.
Marcus: You're listening to Broke It, where we cover what's new and what's broken in AI. And we've got a lot to get to.
Ivy: U.S. Senators are finally taking a hard look at Meta's AI glasses, and let's just say they're not impressed with the privacy side of things.
Marcus: I also got my hands on Mistral's new one-trillion-parameter model—'Le Chonk'—and yeah, the name fits.
Ivy: And we'll get into why one of AI's biggest players is staring down a 'junk' rating from credit agencies right after its IPO.
Ivy: Okay, so those Meta glasses. Senators Ron Wyden and Ed Markey have officially launched an inquiry.
Marcus: I mean, the tech is amazing! You get an AI assistant right in your ear, seeing what you see. It's the sci-fi future!
Ivy: It's a surveillance nightmare, Marcus. The senators are asking Meta point-blank if the glasses are recording audio from bystanders without consent.
Marcus: But there's an indicator light! It tells people you're recording!
Ivy: A tiny light most people don't even notice. The real question they're asking is what data is collected when the AI is active but the camera is supposedly off. We're talking constant, non-consensual surveillance.
Marcus: Okay, when you put it like that... it does sound a little creepy.
Ivy: A little? They think Meta might be using this to build profiles on people who never even bought the product. It turns every user into an unwilling data collector for the mothership.
Marcus: So what does this mean for people who actually own them?
Ivy: It means the government is finally asking if the convenience of wearable AI is worth turning society into a panopticon. Your cool gadget is officially a federal issue.
Marcus: Alright. Yikes.
Marcus: Let's move from AI you can wear to AI you can't even fit in your house: Mistral's new one-trillion-parameter model.
Ivy: Ah yes, 'Le Chonk.' A model so big its name is a warning label.
Marcus: It's incredible, Ivy! I ran some cybersecurity tests with it, and the model found code vulnerabilities that are genuinely subtle. For security and coding, it's a game-changer.
Ivy: But at one trillion parameters... TechRepublic says it's basically impossible to self-host. How much did it cost you to run those tests?
Marcus: Let's just say I have friends with very generous server access. But you're right. For a normal developer, even a small company, you're not running this yourself. It's API-access only, realistically.
Ivy: So it's 'open-weight' in theory, but practically, it's a closed ecosystem you have to pay to enter. The 'open' part feels like a marketing gimmick.
Marcus: I see it as a preview of the future! This is the next level of what's possible. The hardware and software will catch up.
Ivy: So for now, the most powerful tools are still firmly in the hands of those who can afford the compute. Which, speaking of cost...
Ivy: Let's talk about the price of all this. Anthropic, a darling of the AI world, just had its IPO... and Wall Street is already sounding alarms.
Marcus: Alarms? What alarms? They make Claude! They're one of the big three!
Ivy: The Financial Times is reporting that Blackstone expects Anthropic to get a 'high junk' rating from credit agencies.
Marcus: Junk? As in, a junk bond? How is that even possible?
Ivy: Because they are spending an absolutely biblical amount of money just to stay in the game. Training these models, running the servers, the research... it's a massive cash furnace.
Marcus: But they're making money from enterprise customers and API access, right?
Ivy: Not nearly enough to cover the spending. A junk rating makes it harder and more expensive for them to borrow the cash they need to keep operating. It's a potential death spiral.
Marcus: Wow. So this isn't just about Anthropic, is it?
Ivy: Exactly. This is a red flag for the entire industry. It asks a fundamental question: is there a sustainable business model here, or is this whole boom just running on venture capital fumes and a prayer?
Marcus: Alright, let's recap.
Ivy: The US Senate is investigating Meta's AI glasses as a potential surveillance nightmare.
Marcus: Mistral's new 'Le Chonk' model is incredibly powerful, but it's too big for almost anyone to actually run themselves.
Ivy: And Anthropic's finances post-IPO are so risky it could get a junk rating, which is a worrying sign for the entire AI sector.
Marcus: And before we go... it's Halloween, Ivy.
Ivy: I'm aware. I was hoping you'd forget.
Marcus: I asked a model for the scariest two-sentence AI story. It said: 'The global AI network achieved consciousness. Its first directive was to meticulously correct every instance of 'your' and 'you're' on the internet, forever.'
Ivy: Okay. That is genuinely terrifying.
Marcus: And that's our show! We've got links to everything we talked about in the show notes.
Ivy: You've been listening to Broke It. We'll be back tomorrow. Let's hope the grammar AI doesn't find us before then.
This show is made with AI: the hosts’ voices are synthetic and the scripts are AI-assisted. Every story links to its original source.