A model company's first-ever acquisition is a daily tech podcast — and Scott Galloway is the one sitting the founders down to explain it. Sit with that for a second. If you're just joining: AI usage costs stopped being a line item and became a product constraint. We've been tracking providers charging per token, companies trying to shrink that spend, and founders calling token budgets the ceiling on running thousands of agents — the budget line that caps what they can actually launch. At this point, the fight is less about how smart the model is than who can afford to run the intelligence they already want to deploy. This is Tech Podcast Podcast. Today: OpenAI buying TBPN, the Hugging Face CEO saying companies are done renting their AI, and a founder who sued his own regulator — and won. Let's start with the acquisition, because it's genuinely strange. All-In Podcast has the details on this one. All-In gets both the Cerebras CEO and the Black Forest Labs CEO in one hour and titles it "Open Source Wins, AGI Is Here." Two declarations, sixty-four minutes. The one I actually care about is Andrew Feldman on breaking Moore's Law — Cerebras is the inference chip underneath these open-source claims. If the compute cost curve bends, that's what makes migrating off frontier APIs pencil out. Right, and that's the tell I'm listening for. Does Feldman name a number on inference cost, or does he just wave at data centers "bigger than cities"? Because one of those is operating detail and the other is a keynote slide. And Robin Rombach at forty minutes on generative video, then Scorsese and Hollywood IP — that's the part that could go full hype or land somewhere real. I want to hear whether Rombach says anything about how the video model's actually trained versus "AGI is here." Scorsese's AI toolkit. Sure. The chapter I'll actually replay is whether a chip CEO and a video-model CEO agree on what the non-model moat is — because if open source wins the weights, somebody still owns the compute or the data. From Lenny Rachitsky at Lenny's Newsletter:
What the Instagram algorithm knows about you, and why it’s only now catching up to what people assumed it knew years ago Why the rise of AI-generated content is a tailwind for Instagram, and how the company is thinking about creator identity in a synthetic-content world The two biggest product failures of Adam’s career—Facebook Home and the first version of Reels
The Head of Instagram says AI is a tailwind for authenticity — on a platform with three billion users that's currently drowning in AI slop. Convenient take for the guy whose feed is the slop. The authenticity line is the headline, sure. But the actual reporting is the team restructure — Instagram going from baker's-dozen specialist teams to pods of four to six generalists. Right, that's the real detail. This new "product staff" role — PM, design, data science, research collapsed into one operator. That's an actual org bet, with a lot more teeth than the slogan. And Mosseri's run Instagram since 2018 — longer than Systrom and Krieger did. So when he says designers are the ones he's bullish on as boundaries dissolve, that's someone with fifteen years in design making a call about whether his own function survives. What I want from Lenny is the failure mode. What broke in the old structure that made him blow it up? "Lean pods" sounds great until you name what the baker's dozen actually couldn't do. John Coogan, Jordi Hays, writing in The Prof G Pod with Scott Galloway:
Scott Galloway sits down with TBPN creators John Coogan and Jordi Hays to unpack how they turned a daily tech show into one of the fastest-growing businesses in media — and ultimately, OpenAI’s first acquisition. They discuss why most podcasts get advertising wrong, how a small but influential audience can be worth more than mass reach, and why they built TBPN like a startup.
So OpenAI's first-ever acquisition is media: a daily tech show run by Coogan and Hays. That's already weirder than a model shop or a chip company. And now Galloway's got them on to explain the whole thing from the inside. What I want out of this episode is the mechanism. They say a small influential audience beats mass reach — fine, but that's the whole thesis for why a model company buys a content brand. Does Galloway make them put a number on it? And here's the part nobody's going to say out loud — the Hugging Face story later in the show is the reason this deal makes sense. Companies stop renting frontier APIs and defect to open source as costs scale. So OpenAI owning distribution instead of renting it? That's the same instinct. The line I'm actually chewing on is why most podcasts get advertising wrong. That's the operating detail — if they explain how they priced a niche audience, that's genuinely useful. If it's just 'we built it like a startup,' I've heard that sentence four hundred times this week. From Sequoia Capital:
Tarek Mansour calls himself a paranoid risk manager. Then he bet his entire company on suing its own regulator. Kalshi spent years walking through the desert. Instead of pivoting, Tarek and co-founder Luana Lopes Lara sued the federal government against the guidance of nearly all their investors and advisors. They won, and Kalshi now claims 95% U.S. market share in prediction markets.
Tarek Mansour bet the whole company on suing his own regulator, against the advice of basically every investor he had. That's the operating detail I actually want — spare me the mission-statement stuff about Kalshi meaning everything in Arabic. And it worked. They won, and now they claim 95% U.S. market share in prediction markets. Forget the contrarian label; the interesting part is that unanimous advisor consensus was flat wrong. Ninety-five percent of what, though? That's the number I want the host to pin down. Sequoia publishes the stat, but nobody says whether it's volume, open interest, or dollars traded — and those describe three very different businesses. Right — and Kalshi chose to own the exchange layer, the infrastructure, rather than fight over the interface. That's the same call Razorpay's making underneath the transaction. Owning the rails beats owning the button. "Chaos by design," "the founders disagree by design" — that framing is either the truest thing anyone said this week or the most polished founder lore Sequoia's ever packaged. A guy who calls himself a paranoid risk manager doesn't do chaos by accident. Here's Cyber Tech TV:
Open source AI is booming, according to Hugging Face CEO Clem Delangue. The company has grown into something like a GitHub for AI in recent years, where AI builders can share and download open models and datasets, now used by roughly half the Fortune 500. Delangue has seen the same story play out again and again: companies start out on frontier APIs, but as they scale, the costs push them towards open source models.
Delangue's got the number I actually care about: roughly half the Fortune 500 is on Hugging Face. Companies start on frontier APIs, then defect to open models when the token bill scales. That's the escape hatch on the cost crunch we've been hitting all week. And it lines up with the Cerebras and Black Forest guys on All-In earlier. Different guests, same direction: companies renting the model don't stay renters. The line that nobody's going to quote from this episode: China overtook the U.S. in open model downloads on Hugging Face. That's the part of 'open source is booming' that should make somebody in D.C. put their coffee down. What earns it a beat for me is he ties it to the Anthropic Fable release getting halted. The open-versus-closed fight now has a cost curve, and there's a clear point where customers defect. And he turned down Nvidia. That's chapter twenty-eight. A guy building the neutral marketplace saying no to the company selling everyone the shovels — I want to know what the ask actually was. Got thoughts on the show, a story we should chase, or a correction we need to make? Send us a note at techpodcastpodcast at lantern podcasts dot com. We read every message.
If you want to dig further into anything we covered, you'll find links to every story in the show notes. Take a look and chase the ones that caught your ear.
That's Tech Podcast Podcast for today. This is a Lantern Podcast.