Meta's Llama drama and how Trump's tariffs could hit moonshot projects

Meta dropped three new models over the weekend: Scout, Maverick, and the still-training Behemoth, billed as the next evolution of “open-ish” AI. But instead of excitement, the response was mostly shrugs. Critics called the release underwhelming, saying it lacked the edge expected in today’s breakneck AI race. Meta’s clear attempt to claw back some attention quickly turned messy. Accusations began circulating on X and Reddit around benchmark tampering, a mystery ex-employee, and large gaps between the models’ public and private performance. Today, on TechCrunch’s Equity podcast, hosts Kirsten Korosec, Max Zeff and Anthony Ha are unpacking Meta’s rocky rollout, the AI industry’s obsession with looking smart on paper, and why, as Kirsten put it, “creating something to do well on a test doesn't always translate to good business.” Listen to the full episode to for: A breakdown of Trump’s latest tariff push, what you missed and how companies are bracing for impact The secretive EV startup backed by Jeff Bezos, and whether it was Bezos’s Plan B Colossal Biosciences’ Dire Wolf discovery, and whether or not the breakthrough justifies the startup’s $10B+ valuation Equity will be back next week, so stay tuned! Equity is TechCrunch’s flagship podcast, produced by Theresa Loconsolo, and posts every Wednesday and Friday. Subscribe to us on Apple Podcasts, Overcast, Spotify and all the casts. You also can follow Equity on X and Threads, at @EquityPod. For the full episode transcript, for those who prefer reading over listening, check out our full archive of episodes here. Credits: Equity is produced by Theresa Loconsolo with editing by Kell. We’d also like to thank TechCrunch’s audience development team. Thank you so much for listening, and we'll talk to you next time. Learn more about your ad choices. Visit megaphone.fm/adchoices

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Anthony Ha?

  This episode of Equity is brought to you by Schneider Electric. Schneider Electric, your energy technology partner.

• 0:17Kirsten Korosec?

  Hello and welcome back to Equity, TechCrunch's flagship podcast about the business of startups. Today is Friday, April 11th. I'm Kirsten Korosec, Transportation Editor at TechCrunch.

• 0:28Speaker 02

  I'm Max Zepf, Senior Reporter with TechCrunch.

• 0:31Anthony Ha?

  And I'm Anthony Ha, Weekend Editor at TechCrunch. And there are all kinds of insane things we could talk about from this past week. And we'll get into some of them. But Max, I think you were going to start us off with something crazy that happened at the very beginning of the week.

• 0:43Speaker 02

  Yeah, so this week got off to a crazy start when on Monday, a news aggregator account on X, known as Walter Bloomberg, which has no connection to Bloomberg News, posted a false report declaring that President Trump was considering a 90-day pause on his controversial tariff proposal. The headline that he posted, which is since deleted, has it, "Trump is considering a 90-day pause in tariffs for all countries except China." This caused a huge surge in the stock market, but then when it was found out to be true, it plummeted. But this was all the more confusing when on Wednesday, Trump actually announced this exact thing.

• 1:20Kirsten Korosec?

  I was gonna say, I was gonna jump in and be like, was it false? Because in the end, it ended up being a 90-day pause in tariffs, which we're gonna probably talk about later. But I did want to note that in that surge, basically it was a whipsaw event. So extreme volatility in a short period of time, over about 30 minutes, $2.4 trillion or so was lost. So yeah, tweets matter, folks. Tweets matter.

• 1:47Anthony Ha?

  Which, I mean, yeah, we should talk more about the tariff side of it. But I also had another question, which is like, what is the deal with Walter Bloomberg? Because I see those tweets all the time, less than I used to, because I spend less time on X. But like, it just, it just seems to be like this, you know, very rapid-fire aggregator, but no one seems to know a lot about, or at least I haven't seen a lot of background of who it is.

• 2:08Speaker 02

  Yeah, I mean, it's an anonymous account and I think most, I would say 90% of the posts are just kind of taken from the Bloomberg terminal, just kind of those little snippets that appear on there, if you've ever used one. But then there's other kind of, he takes kind of headlines and he kind of looks at interviews that are happening and, or maybe it's multiple people, but yeah, and that's what happened here. It was just kind of an aggressive interpretation of something. But we'll talk more about that later on the show.

• 2:33Kirsten Korosec?

  Yeah, well, speaking of the show, on the show today, we have deals of the week from a secretive EV startup backed by Jeff Bezos and some genetic tinkering by colossal biosciences. Then we're going to get into all things tariffs and then talk about what Meta's Llama Four says about the state of AI.

• 2:54Speaker 02

  Kirsten, you mentioned at the top that Jeff Bezos is backing a secretive EV startup. You and Sean worked on this scoop for a very long time. Can you tell us more about what went into that?

• 3:05Kirsten Korosec?

  So by a long time, it was about a year, coming on a year. And it all started sort of in a twofold way. Sean was looking into this startup that he stumbled upon because he was looking into another company and saw that someone had recently gone over to this company called Slate, started digging into it. Around the same time, I tossed out a few calls and emails to some sources and got a tip that Jeff Bezos was involved. And so there it started. And really, it's been an on again, off again, sort of investigative

• 3:45Kirsten Korosec?

  story over the past year, where we've just been like slowly digging into things, talking to people, trying to verify. And we kind of had an aha moment at the end of last year that finally kind of confirmed the Jeff Bezos piece. A lot of the other stuff had kind of, had kind of already been confirmed through a bunch of different sources that we had.

• 4:10Anthony Ha?

  And I mean, it sounds like it's not just, I mean, so Bezos is an investor, but it sounds like this is a company that also has some like deeper connections to Amazon as well.

• 4:19Kirsten Korosec?

  Oh yeah, absolutely. So the interesting thing is, a couple of weeks ago, we were talking about Rivian's spin out, that sort of micromobility spin out called Also. This is in some ways reminiscent of it, with the exception that we're not talking about a publicly traded company. There's a company called Rebuild Manufacturing, using that same prefix that Amazon loves to use, like Remars and, you know, all of those other RE:monikers. It's so ugly. Yeah. And it sticks in your brain, right? So we noticed it right away that, that what appears to have happened is that Rebuild Manufacturing, which is attached to a bunch

• 5:00Kirsten Korosec?

  of Amazon folks, including Jeff Wilkie, in the confines of Rebuild, Slate was born, and then Slate then spun off. But they were both kind of secretive companies. So a lot of people just didn't realize the connection or the roots. But yes, Amazon DNA throughout Rebuild in particular, and now over at Slate. Slate also has a lot of folks from kind of some unexpected places, not your typical tech places. So Harley-Davidson, Stellantis, which is the maker of Jeep Chrysler, which is, again, Stellantis connected. So really interesting sort of manufacturing folks coming in, which, once you learn about the

• 5:45Kirsten Korosec?

  business, makes a lot of sense.

• 5:47Speaker 02

  Yeah, I had a question about that because, I mean, you write in the story that they're building this two-seat electric pickup priced around $25,000, which first off, I want that. But secondly, that seems like such a kind of unique approach for an EV startup to just, yeah, go with this kind of affordable, mass-produced kind of item right out of the gate, which you talk about a little bit in it. And it's very different from Rivian and Tesla. It feels like something that only Amazon would really make. Like, that's something that would only come out of a company like Amazon. But can you talk about kind of like how this ties into, you know, Bezos's larger vision?

• 6:27Kirsten Korosec?

  Well, first of all, we haven't seen the truck yet. So, and, you know, nothing has been produced. So I always like to put those, like, Clarifying points on there. We'll see what happens, but they aspire to do this low-cost pickup. And the interesting piece about this is that it is very different from what almost every other EV startup has tried to do, which is usually enter at the top of the market with a high-priced, low-volume luxury vehicle. And then as word gets out, brand awareness, locking up supply chain, then move into a higher-volume, lower-priced vehicle. Tesla's

• 7:11Kirsten Korosec?

  done this. Rivian is attempting to do this. Lucid is still very much in the luxury market, but they are trying to be higher volume. This is different, but with also an interesting sort of twist. The idea here is that the core vehicle will be incredibly cheap, but where the margins will come into play, or where they'll make money, is on all the afterbits, the accessories. So think how Harley-Davidson makes a lot of money based on its apparel, or Jeep maker, when you buy a Jeep, it can be very basic, but then Mopar is the aftermarket product stuff. So this is their sort of business model. We'll see, we'll see how it is actually executed, though.

• 7:57Anthony Ha?

  It's something that feels like, oh, that's a great idea. Like, that's like, I would like to buy vehicles that way if it's done right, but at the same time, it really runs the risk of feeling like really nickel and dimed, where it's just like, well, like, you say it's $25,000, but actually, like, what I get is like a piece of trash for that, and like, I actually need to spend another $25,000 to actually have a good experience.

• 8:17Kirsten Korosec?

  Yeah, so that's where the execution part is going to be very interesting. You know, before we move on, I will say this. If you look at the timing of the deal, so the Series A was in 2023, which means discussions were probably happening a little bit before that. The company itself, Slate, was imagined and was created around 2022. And that's interesting because Rivian and Amazon are quite tied together. Amazon is a major shareholder of Rivian. They are a customer, making vans. And so when I was speaking with sources, at that time, Rivian was kind of struggling, and some people speculate, and I want to emphasize speculate here, that this was Jeff Bezos' Plan B.

• 9:04Kirsten Korosec?

  If, you know, things went sideways with Rivian. Very different, obviously, but it is, the timing itself is quite curious. And so we're going to see how it all plays out. But I actually have another deal I want to talk to you about because it's sort of wild about colossal biosciences. I'm not going to use the terminology that that company is using, but essentially they do some genetic tinkering and created something close to a dire wolf.

• 9:32Anthony Ha?

  I'm going to, I'm going to use the word they use not necessarily because I fully co-sign it because, but because I think it is part of, part of understanding sort of the debate over the company and their latest announcement, which they talk about de-extinction. And we've talked about this on past episodes of the show as well. The example people bring up a lot is Jurassic Park and the idea of let's get dinosaur DNA and bring them back. In the colossal biosciences case, they're, I think they're kind of white whale, if you will, is, is the woolly mammoth. And to get there, they've kind of announced a few different experiments. The latest one being what they're saying is a dire wolf, which is basically wolves that have had certain genes that are, you know, turned on.

• 10:12Anthony Ha?

  And it's led to these two six-month-old wolves called Romulus and Remus, who are sort of mythological figures nursed on a wolf. Although they are also clear about the inspiration from Game of Thrones because, I mean, A, dire wolves really popularized recently in Game of, recently-ish in Game of Thrones. And then also they have apparently another wolf named Khaleesi. But, you know, sort of circling back again, like any, any piece of like press and then, you know, colossal is a very hyped company right now. Anytime they talk about it, people talk about this idea of de-extinction. And I think a number of scientists are like, um, I don't know that this really counts as, as de-extinction. What do you guys think?

• 10:51Kirsten Korosec?

  Well, I, I was wondering, so the question posed here is, this is a company that they raised 200 million in a series C at a $10.2 billion valuation, which is big. And I'm wondering, Max or Anthony, like kind of put it back to you, does this justify their valuation? Or is this a very elaborate stunt that isn't going to turn into anything?

• 11:19Speaker 02

  I mean, there's a certain element to this where it's very much a stunt. Like dire wolves have become so popular because of Game of Thrones in recent years. Like if you're going to bring back an animal, that's going to get attention. Makes sense to do a dire wolf. The company kind of says that part of the reason they are focusing on dire wolves is to help with red wolves, which are a very endangered species in North America. But the questions about what are the real world applications of this? And more kind of pressingly, like what are the real world applications of this that can make this company billions of dollars over its lifetime? I think those are kind of some real questions that aren't quite answered because to me,

• 12:02Speaker 02

  if the ultimate purpose of this company is to help with the conservation movement, then I don't really understand how this is a $10 billion company.

• 12:11Kirsten Korosec?

  Yeah, I mean, one of the things right now is that they don't charge governments for helping with endangered animal conservation, although the company has suggested that it might. They've also spun out a bunch of companies. I believe they've spun out about two companies and have plans to spin off three or more businesses over the next two years, one of which will be for its artificial womb technology. So I suppose there is this idea that the main company is where all the R and D and big thinking happens. And then if there is something that turns into something real, whether it's an application of some kind that it can license this technology, then it spins out and becomes a business.

• 12:54Kirsten Korosec?

  That is intriguing and interesting. Of course, there is the whole ethical question of whether there should be genetic tinkering and this very heavy-handed human element in conservation. But I'm not sure we're going to resolve that debate today.

• 13:12Anthony Ha?

  I mean, I think another element of that debate, I think I've seen when people, you know, start to get a little skeptical of that sort of de-extinction language that they use is that I think, I can imagine a scenario where like bringing certain species back is part of a larger conservation strategy. I think sometimes it can come across more like parts of the tech industry, like instead of like saving what we have now, they're just focused on these sort of tech solutions that maybe will mitigate the damage decades in the future. Like, I think it's a sort of similar conversation to the conversation around like going to Mars, right? Like, which I think is, you know, a worthy goal, but sometimes it feels like people are essentially saying that, being like, "Well, like, we screwed up this planet.

• 13:56Anthony Ha?

  It's too late to like worry about saving it. We're going to Mars." As opposed to being like, "Yeah, we can go to Mars, but like, this is our planet. This is the one we're stuck with. Like, and we can talk about Mars too, but that, you know, that's not going to solve the fundamental issues." And I think, I kind of feel the same way about these kind of genetic de-extinction ideas.

• 14:13Speaker 02

  Right. There's a lot of kind of long-term problems that Colossal is trying to solve, but we have to move on. We are going to take a short break, but after the break, we're going to talk about a very near-term problem, which is the tumult of Trump's tariffs.

• 14:31Anthony Ha?

  What moves our world forward? Energy. History has been shaped by it, revolutions won with it, and progress dependent on it. For 190 years, Schneider Electric has been a leading force in this story, creating the technologies that harness energy and transform the competitive landscape. And today, in the dawning era of energy and industrial intelligence, Schneider is helping customers build more efficient, resilient, and sustainable operations across every sector, from data centers to homes, from buildings to industry. Schneider Electric, your energy technology partner. Discover more at se.com.

• 15:08Kirsten Korosec?

  And we're back here to talk about all things tariffs. And if there's a theme here, I think it's just uncertainty. Because by the time this recording comes out, it may have all changed yet again. But to catch everyone up and kick off the conversation, there were what some people have, I feel incorrectly called reciprocal tariffs, but tariffs nonetheless, a baseline of 10%. Then specifically very high tariffs for certain countries. That is now being paused for 90 days. But the important thing here to remember, and what I'm really interested in hearing from you, Anthony and Max, is that there is a 90 day reprieve on tariffs.

• 15:52Kirsten Korosec?

  However, that baseline 10% is still in place. And more importantly, the levies on Chinese goods are now even higher, I believe, to about 125%. Now it feels like we're sitting in a like legitimate US-China trade war.

• 16:12Anthony Ha?

  Yeah, it's weird to be in this situation where that seems more reasonable than, or at least slightly less crazy than what was about to, you know, take effect. But yeah, it's like when you talk about, oh, it's only China. I think we're about to see like how widespread only China is because it seems clear that like that's going to be a huge swath of goods because it's not just, you know, products that maybe have the made in China label on them, but there's also like parts and, and all kinds of different things where, you know, that sort of economic relationship now has, as far as I can tell, really just been, been shattered.

• 16:48Speaker 02

  Yeah, I think there's a, I heard an interesting quote from an analyst earlier this week that said, you know, there's always uncertainty in the markets. There's no such thing as real certainty, but it's a lot different when you wake up the next day and you're not sure if it's going to rain and whether you think there's a tornado that's going to come through your town. So this is really just like an uncharted level of kind of uncertainty that we're seeing kind of around tariffs. I think this has played out in a lot of kind of interesting ways for the AI chip world specifically. AI chips were given a cutout in Trump's original tariffs, but people still expected the other parts of building data centers to kind of dramatically rise and all the costs around that. But now it's, it's still kind of an open question around what's really going to happen there.

• 17:32Kirsten Korosec?

  Yeah, I mean, I think the thing that's important to remember and that might be missed in the conversation, but we'll see happen in the coming months, right? Right now, we're all reacting to the stock market's volatility because of this idea of around uncertainty. But the actual effect on businesses and then the trickle down effect to consumers, because pretty much every economist agrees that tariffs, you know, the cost comes down to the consumer. There's not a lot of debate there. I don't think we understand really, and it's gonna, we're going to learn soon, how much of a web Chinese goods are in all of our even American assembled products.

• 18:15Kirsten Korosec?

  So in the case of the data centers, which is really interesting, there is this carve out or exemption for semiconductors. But if you look at what a data center is filled with, it is filled with electronics hardware, many of which are made and are assembled in China. So think of server hardware, like motherboards and, you know, network interface cards. Then there's the power equipment. Then there's the cooling. A lot of this stuff comes from China. And so you see analysts already talking about whether this is going to compel tech companies to slow down on this AI data center building boom. And I'm wondering, Max or Anthony, if that's sort of what you're hearing as well, or if

• 19:00Kirsten Korosec?

  you think that the demand for it is so high that they're just gonna push forward anyway?

• 19:06Speaker 02

  Right. I mean, I think that there was a certain element to the data center build out story that some companies were already slowing down building out data centers before this happened. I mean, Microsoft is a good example of this. They've canceled a lot of their data center projects reportedly in the last few weeks. And a major reason for this partially is just because Microsoft was building so many data centers to begin with. Their relationship with OpenAI is kind of getting more complicated, and it seems like they might not be on the hook for supplying all their compute in the future. But I mean, these companies are investing hundreds of billions of dollars over the next few years on the data center build out. And when you're talking about that scale, I mean, even just if, you know, prices increase,

• 19:51Speaker 02

  you know, a few percentage points, I mean, it's billions of dollars flying out the door. So I think this creates a lot more risk associated with those massive investments they're trying to make.

• 20:01Anthony Ha?

  Yeah, to me, it's also interesting to think about this in the context of sections of Silicon Valley, like, really going hard for Trump in the last election. And 404 Media had this interesting piece, I think it was over the weekend, basically making the argument that folks like Mark and Jason, like, really went hard for Trump, and then it turns out they've basically, like, endorsed a D-Sell in chief. And, like, this is very much not what they wanted. And so if you thought it was gonna be, and because obviously, if you can look at other policy areas like crypto and things like that, where it seems like the Trump administration is very much kind of going with what that sort of rightward, libertarian-leaning side of Silicon Valley wants, this is not that. And you can sort

• 20:43Anthony Ha?

  of see that, I mean, A, there's just things that Trump himself believes in once, and then also, like, that there are clearly very different factions within the administration who are pushing for very different things.

• 20:53Kirsten Korosec?

  Yeah. I mean, I think what's going to happen is the reality of the cost of capital that is triggered by tariffs is going to be incredibly concerning for large segments of Silicon Valley and other tech leaders that may have been, you know, aligning themselves with Trump several months ago. And when you think about this, like, Max, to your point earlier about how very small percentage points can cause billions of dollars. So if you think about, like, the data center AI building boom, or even investments into deep tech, that was fueled by tech companies with lots of cash in a solid and fairly certain business

• 21:35Kirsten Korosec?

  environment. So now you have companies like Apple and others facing this, like, extreme uncertainty, and then they have to deal with this uncertainty and the pressure on their core business, and that could cause them to pull back, not just with data centers, by the way, but other investment areas. And that is a reversal to Anthony's point, a pulling back. And I, you know, what was top of mind for me was, how is this going to affect some of the really big moonshot style projects like Stargate?

• 22:07Speaker 02

  Right. It's, it's interesting because companies like Apple and Tesla and a lot of Elon Musk's ventures are reliant on this kind of global infrastructure that they have. And to kind of implement these tariffs and to try to onshore jobs into the United States. I mean, there's just no world where an iPhone is assembled and built completely in the United States. Like it just doesn't make sense without the cost, you know, increasing two or three times. I am curious about kind of, Kirsten, what this means for Tesla in particular, because they are so exposed to China tariffs.

• 22:42Kirsten Korosec?

  Yeah, I mean, it's funny because when the 90 day pause was first announced, a lot of folks immediately reacted. We saw Tesla's stock go up and it was like Tesla's great. But they're actually exposed. They are one of the most quote unquote, "American-built" car companies out there. They have a lot of their assembly facilities in the US. They do a lot in-house, but they still rely on componentry. I mean, a story several years ago was about how they have one of their own lanes on the Texas-Mexico border for that very important, by the way, trade partner when it comes to automotive components, Mexico, very huge.

• 23:26Kirsten Korosec?

  So they're exposed and specifically because it's with China, I think that they are going to get a lot more pressure than people are realizing. And actually, I think they're realizing and now a little bit more because their stock price is down again. Tesla is absolutely exposed, but other automakers are as well. Mexico's and Canada are both huge trade partners. And while it's not 125% tariff like we're seeing in China, you still have that baseline 10%. And then, of course, some of the like cabling and other critical materials, battery materials are coming from China as well, which exposes really any company that is any tech company or any company that is trying to build EVs or any in-car tech.

• 24:11Kirsten Korosec?

  That's not going away, even though there is this 90-day reprieve. But we have to move on to one of my other favorite topics, which is AI. I'm very curious what Max thinks of Meta releasing this Llama 4.

• 24:27Speaker 02

  Yeah, so Meta did have a bit of Llama drama in the last week. They, no, it was good. It was good. And then, so last weekend, Meta released Llama 4. There are three new models in total. There was a small, medium, and behemoth model. And these were kind of Meta's next generation of AI models to really kind of push their open source effort forward. Not quite open source, but they call it open source, but that's a conversation for another day. Broadly, the industry was kind of underwhelmed by this release. You know, we're kind of getting used to a pace in the AI industry where Google releases a new model, and OpenAI releases a new model, and Anthropic releases a new model.

• 25:09Speaker 02

  Every month, we're seeing these releases that are just more impressive and pushing the capabilities forward. And Meta's release of Llama 4 really just didn't show any new capabilities exactly. It didn't quite push the edge of what AI models can do. It didn't quite push the edge of what open source AI models can do. DeepSeek still seems to have an edge on Meta here. So it really was kind of a confusing release. But what made matters worse was that there started to be these rumors flying around that Meta had kind of gamed some of its benchmarks. Usually Meta would never kind of respond to something like this, but one of their executives actually explicitly denied rumors

• 25:54Speaker 02

  that Meta tried to game benchmarks, that they specifically trained their AI models to boost benchmark results. These are these kind of AI tests that everyone loves to compare in the industry. But the thing got even worse when Meta released a version of Llama on this benchmarking site called LM Arena. And this is a place where people really care how their AI models perform against each other. But then the LM Arena, the folks who run that benchmarking site, said that Meta hadn't quite been as straightforward as they should have been, that they actually released a very different model from the one that was released more broadly. And this was a model that was basically trained to be very good at chatting, to

• 26:38Speaker 02

  perform well in this specific domain. So it was not a very impressive release, and it seems like Meta was kind of caught cheating to try to make it look better than it was.

• 26:48Anthony Ha?

  Right. Although, we should say, my sense is that one of those benchmarking things that you were mentioning, LM Arena, like, it seems pretty clear that they did use a specific, different model there. There are all these other kind of unsubstantiated rumors as well. So there's a little bit of a, you know, a mix of some things that are substantiated and some that are not.

• 27:06Speaker 02

  Yes, yes. The part we know is that they did release this fine-tuned model that would perform well in this setting, where people, the way it works is that people compare, you ask a question, and then the arena shows you two chatbot answers, one from, say, ChatGPT, and one from Llama's new model, and you pick which one you like the best. And Meta released this model there that basically starts off every answer with, "What an interesting question", or, "Wow, that's fascinating", with like a bright, like, like an iHeartiemoji. It's really weird. Apparently, that does the best in that kind of setting, was what they found. But that's not how the actual model works. If you use it, if you download it on Hugging Face, totally different thing. So it was very confusing, and it raises a lot of questions

• 27:51Speaker 02

  about what's going on at Meta's AI division.

• 27:54Kirsten Korosec?

  Okay, I do have some questions. And I'm wondering, is this, aside from being sort of a little shady and kind of underwhelming, is this Meta just getting caught at what every company is doing right now? And is there a across-the-board sort of dilution of progress, and that nothing that, like, everything that we're hearing from every company working on this should be looked at with a sense of skepticism? Or is it Meta stands out the one and only bad actor here?

• 28:30Speaker 02

  I think the answer is yes to both, like to all of the things you just said. Like, I think that there's a certain element of this where Meta has a history of gaming, like metrics across the board on social media for kind of how they measure monthly active users, how they measure all these kinds of like clicks and whatnot for advertisers. I mean, they have a history of doing this. But everyone, to a certain extent, is doing this across the tech industry. Like, it's not just Meta, especially in the AI world. I mean, every company is so focused on winning and getting their model to be the top of this LM arena kind of website. But all these benchmarks that they have out there, there's these kind of tests for coding

• 29:11Speaker 02

  and math and science that they're all trying to kind of get their AI model to score one percentage point better than another. And to me, it represents really a lack of innovation across the industry where everyone is basically just trying to beat OpenAI, beat Anthropic and beat Google on these very specific domains, when I think there's a lot more interesting, kind of larger questions we could be asking AI models to solve. Like, okay, how do we get this AI model to really automate, you know, my entire like email responses? Like, how do we get AI models to kind of, you know, get better at talking to people on the phone to automate customer, you know, service assistant jobs?

• 29:52Speaker 02

  I mean, there's this kind of broader, bigger picture things that people seem less, these AI apps seem less focused on solving, and they're really just laser focused on these tests.

• 30:02Anthony Ha?

  I guess to play a little bit of devil's advocate, would the argument be that like the models, these core models, they're not necessarily going to be focused on a specific use case, and really that's up to sort of maybe startups, whatever, to like maybe then take advantage of Llama or OpenAI or whatever, what have you, and they're going to build these more focused applications?

• 30:20Speaker 02

  Right. I mean, that's, that's the idea, definitely. Like, if you build a more capable model, people can do more with them. But I think that what we're seeing is that these models, while they're performing better at these tests, that in some ways, that's the only thing they're getting better at. Like, like with this Llama 4 model, it looked like it was better on all of these tests. But the reality when people used it was like, this is... Really nothing new. Whereas other models, I will say, like Google's Gemini 2.5 Pro that they just released, people have started using that in like their, you know, coding workflows. People I've talked to a lot of people who have used that and they say, this AI tool just we plug

• 31:01Speaker 02

  it in and it solves things that we couldn't solve before. Like, like it just makes things work. So to some extent, like, there is some credibility to what you're saying. But I think across the board, we're seeing too much focus on these benchmarks and it makes them not a very reliable indicator.

• 31:18Kirsten Korosec?

  Yes. Creating something to do well on a test doesn't always translate to good business. Unfortunately, we have to go now, but we'll be paying attention to your coverage, Max, on all things AI. And thanks again for joining. But Equity will be back next week. Until then, you can find Equity under the handle at EquityPod on X and threads.

• 31:41Rebecca Bellan?

  Equity is produced by Teresa Locan Solo with editing by Kel. We'd also like to thank TechCrunch's audience development team. Thank you so much for listening and we'll talk to you next time.