A $22 billion media deal, a Stanford researcher saying superintelligence can't act like a person, and a VC calling the next consumer gold rush — all on the same Wednesday. No, it doesn't really make a tidy theme. This is Tech Podcast Podcast. Today we're bouncing around — open-source candor from Evan You, a falsifiable AI counter-thesis out of Stanford, and whether consumer is actually back. Let's start with the deal that has nothing to do with tokens. After a week of compute-cost math, TBPN leads with Fox and Roku. That's either the most interesting non-AI story of the week, or somebody's chasing a headline. Let's find out which. Follow the show and the next briefing lands in your feed on its own. This one comes via TBPN. So today's TBPN headline is Anthropic drama, Meta token mining, and Fox dropping twenty-two billion on Roku. After a week of token-cost math, somehow the lead story is a content-distribution deal. And that's the one I actually want to sit with — twenty-two billion for Roku is a legacy-media-meets-streaming bet, and it has nothing to do with agents. Fox is buying distribution here; intelligence isn't the asset. Right, it's either the most interesting non-AI story of the week or TBPN chasing the biggest number on the wire. The Meta token-mining bit is the one I can't tell if it's real or a segment title. The Anthropic part is still just a headline today — there's no new figure attached, so I'd hold off treating it as fresh. The Roku number is concrete enough to chew on. Thirty-three minutes in Diet form. Twenty-two billion dollars, three storylines — I'll forgive Coogan and Hays if even one gets a real follow-up instead of a reaction shot. Apple Podcasts, with Joon Sung Park:
Joon Sung Park, founder and CEO of Simile and creator of Stanford's "Smallville" generative agents study, argues that simulating human society requires a fundamentally different kind of model. He frames today's frontier models as the "CPU of intelligence"—rational, superhuman at problems with right answers—and Simile as creating the "GPU of intelligence," built to encode the diversity of people's values, preferences, and tastes.
Joon Sung Park — the Stanford Smallville guy — has the cleanest counter-argument I've heard to the whole 'scale gets us everything' story. His frame is simple: frontier models are the CPU, great at problems with right answers, and human society needs something else entirely. Okay, but CPU-versus-GPU of intelligence is exactly the kind of grand metaphor I came in ready to be unimpressed by. Does he actually name what breaks, or is it vibes? He does. He names a number: they simulated 1,000 Americans, and predicted their behavior 85% as accurately as people reproduce their own answers. That's the falsifiable part — there's a benchmark you can argue with. 85% of your own test-retest reliability. That's actually a real claim — and CVS is running concept tests on it, with some customers simulating their own earnings calls. That last one's either genius or a great way to get sued. The big swing is the 'CERN of human society' — modeling bank runs, climate cooperation, early signals of a democracy collapsing. That's where I get nervous: the 85% result is on opinion-survey replication, not predicting a riot. Right. The gap between 'I matched your survey answers' and 'I called the bank run' is enormous. But after a week of everyone insisting bigger models fix everything, a founder saying flatly that they don't — and pointing at a study — that's the one I'd queue. Here's Colossus:
Clay has become one of the fastest-growing software companies of the last few years, valued at over $4 billion. It helps companies find their customers and reach them at scale. But this conversation is about a lot more than Clay. Kareem is one of the most original founders and thinkers I’ve met recently.
Okay, the setup here is Patrick saying Kareem Amin is one of the most original founders and thinkers he's met. That sentence right there is a yellow flag for me — usually it's the wind-up before a very gentle interview. It might be. But Clay's at over $4 billion, and it's a go-to-market automation company — find your customers, reach them at scale. That's the layer where agents either work in production or quietly fail, so there's a real operating conversation available if Patrick takes it. Right, and instead the menu is statues. Truth, justice, and courage — he keeps statues. After a week where we finally got Uber's four-month burn and a sixty-grand-a-month token bill, I'm a little allergic to founders answering with a value system. I'd push back a little — the way he frames computation as the new material is at least a specific lens on why Clay took off when it did. The question is whether he tells you what actually breaks when you automate outreach at scale, or just narrates the two-chapter arc. That's the tell I'd listen for. If Clay's sending a million messages a day, somebody's agent is making bad calls in production. Does he name the failure mode, or do the statues just stand there looking principled? Kate Holterhoff, writing in RedMonk:
In this MonkCast Conversation, RedMonk analyst Kate Holterhoff talks with Evan You, the creator of Vue.js and Vite and founder of VoidZero, just days after VoidZero announced joining Cloudflare. Evan clarifies what the deal does and does not include, as well as the history of how it came about. He’s also candid about the motives that inspired VoidZero to look for partnership with a cloud provider.
Okay, this is the one I wanted. Two weeks ago we sat here and wondered if getting bought by the next infrastructure layer was the only exit for an open-source toolchain. Now Evan You is on the record. And the tell for me is what he says Cloudflare is not getting. He apparently draws a line around the deal — the operating detail lives there, not in the press release. Right, and it gets sharper because he's candid about the motive — why VoidZero went looking for a cloud provider in the first place. Last time that was a rhetorical question; now there's an actual answer to interrogate. This is a guy who built Vue and Vite — two of the most-used frontend tools on the planet — and my honest takeaway is that independent infra still couldn't stay independent. Pull on that, not just the headline. And he's still not convinced Vue needs a foundation. So even after selling, he's skeptical of the governance answer everyone reaches for. That's the candor I came for. From Maggie Nye at TechCrunch:
They talk about why AI startups are reaching unprecedented revenue levels with remarkably small teams, what’s driving today’s soaring valuations, and why Chi-Hua believes many infrastructure businesses will eventually face the same commoditization pressures seen in previous technology cycles.
Chi-Hua Chien at Goodwater is calling the return of a consumer gold rush — which cuts across almost everything we've been chewing on this week. We've had all this infrastructure and bottleneck talk, and his bet is that the winners live at the application layer. And his evidence is basically Facebook. He sourced Accel into Facebook as a young associate, so now every platform shift looks a little like Facebook to him. It's not nothing, though — his actual claim is that infrastructure that looks dominant early gets commoditized, and value migrates up the stack. You can test that pattern. Sure, but what's the mechanism this time? He lists hyper-personalized entertainment, women's health, voice agents. Is consumer back because of AI UX, or is he just pattern-matching the last platform shift onto this one and hoping? And here's what he glides past: if apps win by riding cheap infrastructure, somebody's still paying the compute bill. After a week of sixty-grand-a-month token math, 'small teams, huge revenue' only holds if the power and inference costs cooperate. Right. The 'remarkably small teams hitting unprecedented revenue' line is the most quotable thing here, and it's also the one I'd push hardest. Show me the gross margin after the model bill. Got thoughts on today’s episode, a story we should chase, or a correction we need to hear? Send it to techpodcastpodcast at lantern podcasts dot com. We read it all.
Links to every story we covered today are in the show notes, so if one caught your ear, you can dig in there.
That's Tech Podcast Podcast for Wednesday, June seventeenth. This is a Lantern Podcast.