Ivy: New York City just banned AI... to protect a student's right to struggle.
Marcus: I'm Marcus.
Ivy: And I'm Ivy.
Marcus: And this is your AI daily briefing for Tuesday, September 8th. We've got a packed show for you.
Ivy: First up, we're digging into why New York City public schools just banned generative AI.
Marcus: Then, the founder of Opendoor is out of stealth with a new company, using AI to tackle the massive labor shortage in construction.
Ivy: And finally, we tested Meta's new, supposedly 'more efficient' coding model. Spoiler alert: it's not great.
Ivy: So, the nation's largest school district, New York City, just banned generative AI for all students. And their reason why is... pretty fascinating.
Marcus: Fascinating... or completely backward? Banning it feels like trying to ban the internet in 2005. You're just setting these kids up to fail.
Ivy: The official reason is to protect—and I'm quoting here—'the human process of struggle and discovery in education.' The idea is that learning how to think is more important than just getting the right answer, fast.
Marcus: But isn't that the whole point of AI? To be a Socratic partner, a tireless tutor? You're not outsourcing thinking, you're scaffolding it.
Ivy: That's the sales pitch. The reality for a 14-year-old on a deadline is 'write my Great Gatsby essay.' The city's worried we'll end up with a generation that can't solve a problem without an AI holding their hand.
Marcus: Okay, but a blanket ban? Not guidelines, not integration... just a hard 'no'? That seems like a massive overreaction.
Ivy: Is it, though? Or is it a powerful statement that some parts of human development are too important to 'disrupt'? Learning to deal with frustration and dead ends is the skill. They literally want kids to be 'productively frustrated.'
Marcus: Productively frustrated? Or just... frustrated? I get the argument, but the world these kids will eventually work in is full of AI. This is like teaching them to use a slide rule when everyone else has a calculator.
Ivy: So if you're a parent in New York, your kid's school just took a very strong, and I'd argue brave, philosophical stance.
Marcus: And if you're a student... you're about to get very good at clearing your browser history. This just pushes AI use underground, where teachers have no oversight at all.
Ivy: Maybe. Or maybe this forces a national conversation about what education should even be in the age of AI. I'm with the city here. The struggle is the point.
Marcus: Alright, next up, let's talk about something a bit more tangible—we're talking steel and concrete. Eric Wu, the founder of Opendoor, has a new company called NavigateAI.
Ivy: Opendoor, the iBuyer that used algorithms to buy and sell houses? That Eric Wu?
Marcus: The one and only. He's been in stealth mode since leaving Opendoor, and TechCrunch just got the scoop. His new company, NavigateAI, is building AI copilots for construction workers.
Ivy: Copilots for construction? What does that even look like? A robot in a hard hat?
Marcus: Think of it as a voice in your ear. The AI gives workers real-time, hands-free instructions straight from the digital blueprints, through an AR headset. It'll say, 'Okay, install bracket 7-B, it's three feet to your left.' No more rolling out giant paper plans in the mud.
Ivy: So it's basically an audio guide for a construction site. I gotta admit, the hands-free part is smart.
Marcus: It's brilliant. It's tackling a huge problem: the massive labor crunch in construction. There just aren't enough skilled workers. This AI could essentially upskill a new worker on the fly, making them more productive and safer, faster.
Ivy: In theory. But a construction site is pure chaos. It's loud, plans change by the minute, and a mistake isn't a software bug—it's a wall in the wrong place. How can you trust an AI with that? What if it hallucinates a support beam?
Marcus: That's why they're starting with lower-stakes stuff first, like electrical, HVAC, and plumbing. But with a founder like Eric Wu, you know there's big money and even bigger ambition here.
Ivy: That Opendoor connection makes me wary, though. That model worked great... until it spectacularly didn't. I'll believe this works when I see it handle a rainy Tuesday with a delayed concrete delivery.
Marcus: So if you're in the trades, AI might be coming for your toolbelt—in a good way. And if you're watching the space, this is a huge play in a massive, old-school industry that's ripe for disruption.
Ivy: Okay, last up: Meta. They just released a new coding model, Muse Spark 1.3. The big promise is 'efficiency'—it supposedly uses 20% fewer tool calls and 25% fewer tokens.
Marcus: Which, if you're a developer, sounds amazing. Fewer tokens means faster responses and lower costs. Fewer tool calls implies the model is thinking smarter, not harder.
Ivy: 'Supposedly.' MarkTechPost put it to the test and found that all this 'efficiency' comes at a steep cost: it's not actually reliable.
Marcus: Wait—what do you mean? What broke?
Ivy: For simple, one-off tasks, it's fine. But throw a complex, multi-step coding problem at it—the kind of thing you'd actually want an AI for—and it just falls apart. It gets stuck in loops, can't fix its own mistakes, and sometimes just... gives up.
Marcus: So they made it more 'efficient' by making it... dumber? What the heck? That's not a trade-off, that's just a regression.
Ivy: It looks like Meta optimized for a single metric—'fewer tool calls'—without checking if the model was making the right calls. The reviewer found the older version, while chattier, was much more likely to actually solve the problem.
Marcus: As a builder, this is so frustrating. You see these flashy announcements about '25% fewer tokens!' and get excited, but under the hood, the core capability is actually getting worse.
Ivy: It's a classic case of teaching to the test. The lesson for developers? Don't just read the release notes. Test everything yourself. 'More efficient' might just mean 'less effective.' Meta's marketing team gets a win, but the devs get a headache.
Marcus: So, to recap: New York City schools are banning AI, hoping to force students to struggle and think for themselves.
Ivy: Opendoor's founder has a new AI copilot that wants to be the voice in the ear of every construction worker.
Marcus: And Meta's 'efficient' new coding model is a painful reminder to always test things yourself.
Ivy: And before we go, a quick cautionary tale from the world of AI recipes.
Marcus: [laughs] Oh no, what is it now? Did someone try to make AI shrimp scampi again?
Ivy: Worse. A Redditor asked an AI for a 'surprising and delightful cookie recipe.' It confidently delivered instructions for 'savory oatmeal raisin cookies' that called for a full cup of soy sauce and two tablespoons of black pepper.
Marcus: A cup of soy sauce? Get out of here! That's not a cookie, that's a biohazard. Please tell me no one actually made them.
Ivy: [dry] The post included a photo. They looked like glistening pucks of tar.
Marcus: Alright, that's our show. Thanks for listening.
Ivy: We'll be back tomorrow with more. Until then, maybe stick to recipes written by humans.
This show is made with AI: the hosts’ voices are synthetic and the scripts are AI-assisted. Every story links to its original source.