Subscribe
Share
Share
Embed
TBPN is a live tech talk show hosted by John Coogan and Jordi Hays, streaming weekdays from 11–2 PT on X and YouTube, with full episodes posted to Spotify immediately after airing.
Described by The New York Times as “Silicon Valley’s newest obsession,” TBPN has interviewed Mark Zuckerberg, Sam Altman, Mark Cuban, and Satya Nadella. Diet TBPN delivers the best moments from each episode in under 30 minutes.
We have a very special show for you today because we have Tyler Cosgrove guest hosting. He's here in Great to be here. Chair. And I know I'm gonna make mistakes today because I always throw it over to Jordy. Gotta remember this.
Speaker 1:Tyler, it reminds me of this video from Warren Buffett. We gotta play it to show what I'm going through emotionally today without Jordy in the TBPN UltraDome. Let's pull up this video of Warren Buffett throughout the years at the Berkshire Hathaway shareholder meetings. Warren Buffett always goes to Charlie after he gives his comment. He he he he gives his speech and then he kicks it over to Charlie.
Speaker 1:Charlie. Through the years. Charlie. Charlie. Charlie.
Speaker 2:Tearjerker. Yeah. Makes me wanna cry. It's emotional.
Speaker 1:Charlie, how do you feel about that? Charlie. Charlie, me. Charlie. Charlie.
Speaker 1:That's Greg Abel. So if that happens today, I apologize. But the first big story is, of course, Thinking Machines' new model has released. Miramaradi's AI startup released its first model in bid to loosen AI giant's grip. We're gonna be talking about open source, closed source, where the frontier is, national geopolitical model moves.
Speaker 2:Yeah. And and some people had an idea this was gonna happen. Some inkling Oh,
Speaker 1:they had an inkling.
Speaker 2:Yeah, that
Speaker 1:they're gonna I didn't have an inkling that they were going to jump into the open source.
Speaker 2:I think it's actually I think it makes a lot of sense given the Tinker API. Right? Okay. The whole business is is, you know, you're doing fine tuning on open source models. Yeah.
Speaker 2:It makes a lot of sense that they're gonna have their own Yeah. That, you know, you can easily
Speaker 1:It's sort of like are set up as a business to launch an open source model without it degrading any other piece of their business because the Tinker API, that fine tuning that they do, that integration with the customers that they have actually benefits from open source. Yeah. And then they can go to their clients and say, look, you know, it's the Red Hat model. At any time, you can leave because we are giving you the weights of the model open source. You can do whatever you want with them.
Speaker 1:Yeah. But keep working with us because we're helping you a bunch and we're making money in the process. So Yeah. Thinking Machines Lab, first the first model is an open weights model designed to chip away at the lead of OpenAI and anthropics, says The Wall Street Journal. Former open AI technology chief, Mira Maradi, is betting on more customizable artificial intelligence models to chip away at the lead that the frontier labs, such as her former employer, hold over the technology.
Speaker 1:TML, a company led by Maradi, released its first AI model Wednesday and did it with open weights, meaning that others can modify it with their data, called Inkling. The model has 975,000,000,000 total parameters, making it far smaller than estimates of the most advanced closed source models.
Speaker 2:So only, I think, the number is 41,000,000,000 of those are actually, like, active.
Speaker 1:At any moment.
Speaker 2:Yeah. So this is definitely on the on the bigger side of open source models. Sure. Yeah. But, like, that number, it's not these aren't, like, dense models like what you traditionally think of Sure.
Speaker 2:Of the models, like, four years ago.
Speaker 1:Yeah. Yeah. Muradi told the journal, We trained it to be a broad, balanced, foundational foundation model, strong across many domains, flexible enough to adapt. Inkling is not the strongest overall model available today, open or closed, which is a different frame of reference for many of these model launches. There's it's been everyone's been jockeying for the frontier, even if they're not world class at everything.
Speaker 1:Usually, when they launch, they say, oh, well, we're best at something or we're best at this. But a different tone, different communication strategy. And I think it's being well received. I think people are are having fun with the
Speaker 2:mean, I I think the main picture is that this model is, like, uniquely set up for the Tinker API. Yes. It's built to be fine tuned. Sure. Sure.
Speaker 2:That's the whole point.
Speaker 1:Got it. DD Das says, Thinking Machines just dropped the best open weight AI model outside of China. And, obviously, that is a big topic of conversation as business leaders in The United States have some policies and some reticence about using Chinese open source models. Even if they're not worried about the the dystopian, you know, Manchurian candidate hidden inside the weights, Maybe they just want to be aligned with a US based company for a variety of reasons. Inkling beats Nevo Tron three Ultra and benchmarks put it between Kimi K two point five and two point six.
Speaker 1:Of course, there's also news today that Kimi K three will be launching and is another jump forward. But there's back and forth between some AI researchers around what's going on there, how long that strategy will continue. So, Didi says, Many were contending to this throne, but Thinki has come out on top. Really solid release and will pair well with Tinker. So there are some benchmarks that you can go and dig into if that's your thing.
Speaker 1:There's another very bullish take from Jack Morris of Engram Labs. He says, people are under are underestimating what a big deal this is. This is the only open weight model that that's trained without distilling for OpenAI from OpenAI or Anthropic. Kimi distills, GLM distills, Quen distills, Nemotron distills, Kimi and DeepSeek, which counts. Basically, fully different tech stack.
Speaker 1:The first pure Open Frontier coding model. Very exciting. There's a community note on this. Can you break down exactly, like, where are they standing on the shoulders of giants? Where are they not?
Speaker 2:So I think this tweet is is not exactly true. In the blog post, they say, to bootstrap post training, we ran an initial supervised fine tuning Mhmm. On synthetic data generated by open weight models, including Kimi K 2.5. Okay. So I think that's like generally how people think of like distillation that they mean something related to this.
Speaker 2:Sure. So I think that that actually is is not that different than what people like, you know, Nvidia with with Nematron did.
Speaker 1:Sure.
Speaker 2:So this is not like very new, I think.
Speaker 1:But it's sort of like the lightest touch of distillation that could happen. Because it's just one piece of the pipeline Sure. One small amount of data. Yeah. It's not one these of scenarios where we're like, why is it identifying as Claude?
Speaker 1:Or why is it why is it
Speaker 2:Yeah.
Speaker 1:Just saying that it's ChatGPT?
Speaker 2:But it is funny because you can kinda say like, oh, well, if this is like kind of distilled on Mhmm. On Kimi and Kimi's kind of distilled on closed source
Speaker 1:Yep.
Speaker 2:Well, then maybe you get some kind two layer This
Speaker 1:sort of round trip loop. But Yeah. Yeah. At the same time, there's probably something to be said for the more layers of abstraction, the the the the more, you know, ingredients you pour in, like, the distillation becomes weaker and weaker.
Speaker 2:Yeah. And I I think it's also an important question of, well, okay, they're doing some level of distillation, like, why? Mhmm. Because you can either be, like, well, they're just doing it to save time, whatever. Yeah.
Speaker 2:Like, obviously, they have these capabilities Yep. But there's no point in in in, you know, doing everything over again. Sure. Might as well just just use what's out there already. Sure.
Speaker 2:Or is it because these capabilities that they get from this this, you know Mhmm. Distillation light, whatever it is
Speaker 1:Yeah.
Speaker 2:Are those actually super imperative to the model, like, being good?
Speaker 1:Mhmm. Ingram says our founder, Jack Morris, recently issued some unfounded claims that got community noted. We deeply apologize for the confusion caused by his original post, the follow-up post, and the follow-up to the follow-up post. Nevertheless, we stand by his conviction in his own takes and in strong open source models like Inkling. And there is a question of, like, distillation is a vague term where it's not a binary thing.
Speaker 1:Yeah. And if it's not in the pre trained data, does it count?
Speaker 2:I don't I think it's also very much this meme people love to talk about on x. Yeah. They like to kind of, you know, scapegoat. Oh, you know, it's all distillation. That's the only reason Chinese models are good.
Speaker 2:Yep. Is that actually true, probably?
Speaker 1:I mean, Anthropics head of national security policy, Tarun Chabra, accused ZeePu, z dot a I, of distilling both Claude and OpenAI models for GLM 5.2 at the Aspen Security Forum earlier this week. This is from Vincent Chao, senior AI reporter at SCMP. He said it's the first time that they've named Zepu specifically after previously calling out DeepSeek, Alibaba, Moonshot, and Mini Max join the join the club at this point. Yeah. Also accused they also accused DeepSeek of continuing its adversarial campaign of distillation.
Speaker 1:Anthropic is now shutting down distillation accounts on the order of millions accounts of per week. That is crazy scale. You have to I I mean, you always think about it as like, oh, there's like, shut down that one company or shut down that one block of IP addresses. But when there's a really, really distributed attack, We've even heard about whole companies that just, like, resell clawed tokens or Yeah. Or GPT's 5.6 tokens.
Speaker 1:And that looks like a reasonable business because it's just a wrapper company. Of course, you wanna work with them, but then you don't realize that on the other side, who are their customers? Why do they why did they get to a 100,000,000 run rate so quickly? Well, maybe it's a lab that's trying to distill through this pass through entity. And Yeah.
Speaker 1:Of course, it's hard to, like, watermark the tokens once they go out the API and they get passed through some other system. And they can go through other countries, all sorts of things. So millions per week, that is crazy. That's got to be really difficult to it's a game of whack a mole. They say, GLM is, quote, probably the most advanced Chinese model on the market now, which poses significant cyber security challenges.
Speaker 1:They hinted that Anthropic will expand access to mythos to ensure fair fight for cyber defenders. And they said that distillation challenges real in shrinking US lead in AI, suggesting the US government could do more to clamp down on Chinese model adoption globally by working with allies similar to trusted telecom efforts like Huawei and ZTE. So, obviously, a hot topic and people will be debating how how how exactly how heavy of a hand the government should be pushing.
Speaker 2:Yeah. I I think this this release is also makes a lot of sense in the I think it was a week ago, there was that article about, like, Beijing is looking at curbing overseas access to Chinese top AI models. Yeah. Right? So you're not gonna be able to access the Chinese open source.
Speaker 2:Right? It makes a lot of sense to to start doing American open source, Western open source.
Speaker 1:Yeah. It really does feel like there's a It's like pretty wide
Speaker 2:very well timed.
Speaker 1:Yeah. There's it seems like there's a pretty wide gap with at least what's reported preferences from Beijing from the actual government and the companies. The companies are like Yeah. Send us all the NVIDIA chips. Let's distill everything.
Speaker 1:Let's and then let's open source these models and compete internationally. And Beijing's like, hey. Maybe we need, like, an indigenous supply chain here. Maybe we need to, you know, lock down these models, keep our lead over here, go work internally. I don't know.
Speaker 1:This was an interesting post from Grace Lee. She's she asked the question, how did OpenAI Soul finally learn design taste? She projected a thousand websites by GPT 5.6 Soul into a design manifold and discovered big holes. These holes were where GPT 5.5 previously generated outputs with bad AI smell. So, there you know, there's these tells in any AI model that it's not this, it's that, the em dash.
Speaker 1:Once people start identifying those as, we don't like that, it's too AI, it's too generic, one way it appears to actually sort of beat that out of the model is to actively avoid those specific things. Then she calls out three particular areas that have been avoided as anti patterns. One, the bento box layout in dashboards. Two, large typefaces and hero images. I did realize that that sometimes you would ask for a website and you would just get a massive block of huge text and that's just not the way you when you land on a beautiful website.
Speaker 1:It's usually there's more wordsmithing. There's more
Speaker 2:Yeah.
Speaker 1:Yeah. Terse language.
Speaker 2:Well, you know, you you make your first website with with five, six, or whatever, and it looks really good. Yeah. And then you make 10, they're like, oh, okay. There there's actually a lot of patterns I'm I'm seeing.
Speaker 1:Totally.
Speaker 2:Totally. And you can start clocking them. Like, everywhere see. A lot of like claudisms whatever on Yeah. General designers.
Speaker 2:Especially You see them everywhere.
Speaker 1:Yeah. Especially if you don't come with You know,
Speaker 2:it's like the the the, you know, high border radius on the edges. There's a little Yep. Color on the
Speaker 1:Yeah. Side. Yeah. Yeah. Especially if you don't come with like an opinion.
Speaker 1:If you come like we made a whole vibe coded website in Codex for just the latest episode of Nick Bostrom on Joe And I wanted it to look like a UFC fight card and a fight promotional website. And it doesn't look like any, like, normal AI slop. I mean, there's still, like, AI generated images. It looks like AI, but it doesn't look like, oh, yes, that's the bento box layout or that's the offset layout or that's the purple or it's stealing from linear. It it it's a completely different style.
Speaker 1:So if you at least inject, like, one reference point, you'll usually land somewhere.
Speaker 2:Yeah. Mean, it is interesting though. This makes it seem like the new model is not necessarily it doesn't have, like, higher variance
Speaker 1:Mhmm.
Speaker 2:With outputs it gives, but it we basically just found, like, oh, there there's certain examples that people really don't like. Let's just remove those. Mhmm. But you're not necessarily like making the model more creative by your by removing these, like, patterns it always comes to.
Speaker 1:Yeah. Well, you're giving like the the flavor of creativity and maybe that's
Speaker 2:Yeah. But you can imagine It's too bold. If we kind of keep the same model for six months, we'll just notice new patterns. Totally. And and you'll have this kind of
Speaker 1:at the same time, like, mid journey had like a very distinct look and people like that look at least
Speaker 2:Yeah. Some people.
Speaker 1:And so, if you can if you can quickly personalize and customize and land in a place where someone whose job is designing dashboards is happy every time with the layout. Like, there is somewhat of a platonic ideal for some of these design patterns. And Yeah. At the same time, if you're, yeah, working on certain like, there there there are certain designs that are just like solved. Like, you know, make the call to action green, blue, not red.
Speaker 1:Right? Yeah. And so some of those like do need to be consistent. And then also I imagine that many folks who are using these tools like in enterprises are doing even if it's not a fine tune, they're uploading a reference for everything that they're designing. So, it's consistent with the brand Yeah.
Speaker 1:That they've designed. Yeah. Anyway, California Forever lost a $3,200,000,000 shipyard project from defense startup, Cyronic, after the company chose the Port of Brownsville, Texas over Solano County. I was, oh, no. You're not supposed to clap for that.
Speaker 1:We got a Texan in the studio who's happy about that. This is bad news for California. We want California to have a whole bunch of amazing stuff. Brandon Corral, wrote the newsletter, tpppn.com, was very disappointed about this. The automated ship shipyard, known as Point Alpha Port Alpha, is expected to create roughly 10,000 permanent jobs along with thousands of union construction jobs.
Speaker 1:Supporters say California's lengthy approval process ultimately cost the state one of the first marquee tenants that California forever had pointed to as evidence its planned city could anchor a new era of American shipbuilding. Joshua, executive director for the California Alliance for Jobs, said California failed to move with the urgency the product required, quote, while Texas moved quickly and aggressively. Thank you, Jackson. California could not provide clear expedited approval process needed, he said, calling the decision an enormous loss for Solano County, California workers, and our state's manufacturing economy. Earlier this year, California forever signed a forty year construction labor agreement covering seventy thousand acres, and labor groups later backed legislation to fast track environmental review and permitting for the proposed shipyard the legislation has yet to advance.
Speaker 1:Instead, Texas approved a $211,000,000 tax abatement package in June to secure Ceronix investment at Brownsville, roughly 20 miles from Starbase. Labor leaders said they warned that without expedite expedited approvals, the project would leave the state, and that is exactly what happened. A project insider told the San Francisco Chronicle that California forever itself remains on track, but acknowledged that losing a major defense contractor sends a powerful signal about the state's ability to compete for large industrial investments. Very disappointing. But I like Yan.
Speaker 1:I like the California forever project, I'm excited for where he takes it next. I'm sure he's on the hunt for the next major tenant.
Speaker 2:I'm talking about TSMC.
Speaker 1:Yes. TSMC. TSMC both beat earnings and raised their CapEx guide. They're spending a lot more money
Speaker 2:and Pledged to invest an additional 100,000,000,000 in The US. Yes. Plans to spend a record amount cementing its position atop the global semiconductor supply chain.
Speaker 1:Yes. But and and yeah. And they're they're they're investing another 100,000,000,000 in in Arizona fabs, but people are worried about overspending. The news is that the Nasdaq dropped 1% on TSMC's spending plans offset by strong results. Very, very odd story that in in a time when even TSMC, which was not a particularly AGI company for a long time since they've been through the smartphone boom, the so many booms and busts, so many cyclical build out cycles, that when they are finally like, yes, now is the time, people are, oh, I don't know.
Speaker 1:It's too They're they're skeptical. In creator world, there is some news from Colin and Lexus is now the official car of Colin and Samir. What does that actually mean? They made four ads for them that roll that roll out across YouTube. They're sponsoring four videos on their channel.
Speaker 1:It's the first of its kind deal that represents a broader shift taking place in media. The aperture of what it means for a brand to work with the creator is changing quickly. It's very cool to see because obviously they've been on YouTube for a long time. They've done a lot of like host read ads, mid roll ads, but this is a much deeper integration and something that I think will be hopefully replicated all over YouTube and be a new source of revenue for creators of all kinds. So I was excited to see this.
Speaker 1:In other entertainment news, Jake from Economic says, this is almost hard to believe. Disney spent $129,000,000,000 acquiring Marvel, Star Wars, Pixar, ESPN and Fox, which is $182,000,000,000 in today's dollars. Throw in all their legacy assets in the entire company's market cap today is a 169,000,000,000.
Speaker 2:Wow.
Speaker 1:Do you know what this picture is missing?
Speaker 2:Which what do you mean?
Speaker 1:So so they're saying they acquired all these assets and the company is only worth a $169,000,000,000. What's missing from this analysis? The cash that's been returned to shareholders. Disney across dividends and buybacks has returned like 70,000,000,000 maybe more to shareholders, which is I don't know. I just thought for shareholders.
Speaker 1:And and it is it is it is an interesting angle because they have spent a lot acquiring and the company is not worth more than what they acquired. So there's this question of like, were those were those acquisitions accretive or destructive or dilutive. But there is a whole separate picture, which is that a lot of cash has been returned to shareholders throughout this journey.
Speaker 2:Yeah. Also, mean, that's the mechanism with with which those acquisitions were funded also Yeah. Should
Speaker 1:Yeah. It matters a lot. I don't know. It was sort of interesting. Sean Frank has a pitch.
Speaker 1:He says you should move to New York City. Bro, you gotta move to NYC. The weather, horrible. 100 degrees. Easy.
Speaker 1:AC? F that. Taxes, So high. Rent? Highest in the country.
Speaker 1:Air quality, some of the worst in America. Tech, bro, we banned Do they really ban Waymo in New York?
Speaker 2:I believe so. In New No Waymos.
Speaker 1:Wow. That's very wild. Yeah. If you can make it here, can make it anywhere. So, I don't know.
Speaker 1:Do you ever have aspirations to move to New York City?
Speaker 2:At some point, it seems
Speaker 1:You've been to New York City? Yeah. What do think? It's a nice city. That's the thing is that all of this is true and it's still a great city to hang out in.
Speaker 1:It's so fun, so dense. You can see so many people walk around. It's beautiful. It's just like, I don't know, it's unlike anything else. Still great, but
Speaker 2:yeah. You never lived in New York
Speaker 1:City. I've never lived in New York City. But I've spent like a lot time there. So, I've had a good time. Thank you to everyone who tuned in in the chat.
Speaker 1:Thank you for positive reviews of Tyler. Let us know what you think of Tyler. Leave us a review on Apple Podcasts and Spotify. Write us an email. Tell us how he did.
Speaker 1:I think he did fantastic.
Speaker 2:Have the best Thursday of your life. Yes. There you go. That's good good impression. Impression.
Speaker 1:Thank you. Sign up for newsletter at tbpn.com and we will see you on Monday.
Speaker 2:See you.
Speaker 1:Goodbye. Boeing flash bang. Oh, we got the flash bang. There we go.