Alex Kantrowitz on OpenAI, Nvidia & What Could Derail the AI Boom

[00:00:00] [SPEAKER_02] Let's talk about Microsoft. I mean, all-time bag fumble, right?

[00:00:03] [SPEAKER_03] So they saw it, they're ready, they had contractual relationships.

[00:00:08] [SPEAKER_02] There is an argument to be made that the future interface will just be a disembodied AI voice or a chatbot, because it will, like we're seeing already with Muse and with Dots, just be the computing layer that acts on top of everything. So everything else becomes a database, effectively. I spoke with an engineer at Apple who said he was working on HomePod, and he never saw the device until it shipped. Wow. Yeah. So this is the problem is they did not break the silos down. I don't know if I want to go here.

[00:00:37] [SPEAKER_02] The perspective has been, AI critics need a better messenger than Ed Zitron. What I'll say about it is, then go ahead.

[00:00:44] [SPEAKER_04] Alex Kantrowitz, welcome to the Newcomer Pod.

[00:00:47] [SPEAKER_02] It's good to be here. You know, I've been a long-time listener. I listened to this show when it was called Dead Cat. That's how long I've been listening to you.

[00:00:54] [SPEAKER_04] Man, Dead Cat's in the air. This is Tom Doton who writes for Newcomer and was one of the co-hosts of Dead Cat's, the original name of this podcast was just on TBPN and they were waxing nostalgic, I think literally this week. It was sort of a different vibe, that show.

[00:01:11] [SPEAKER_02] It was a short lifespan but very impactful and we will always remember it finally.

[00:01:15] [SPEAKER_04] You know, that one was a little bit more media focused. I think, you know, we're hopeful that, you know, the first 60% of the show we'll talk about tech and then we'll get into our sort of media, maybe self-absorption. But that show was all media sort of analysis. Yeah, I mean, we were reflecting, you started big technology even before I started Newcomer. It was June 2020. I always tell people, you know, Newcomer came out like the month of the New York Times trend piece.

[00:01:45] [SPEAKER_04] Casey Newton dropped platformer that same month. What did you see back then before the rest of us to start a sub stack?

[00:01:54] [SPEAKER_02] Yeah, it was pretty crazy. I mean, so I have this book that came out in April 2020, April of 2020. And, you know, when you write a book, you do all this work to position it to make it exactly the right time. Oh man, we spent hours trying to figure out the right time and we released it at the peak of the first wave of COVID. And so basically the book was, I wouldn't call it entirely dead on arrival, but you could imagine it was really tough to cut through the noise.

[00:02:16] [SPEAKER_02] And understandably so when like people were trying to figure out how to survive to the next day, they weren't interested in reading about how the tech giants employed AI within their companies and how it might come for all of us soon. So basically, I got back to my desk at BuzzFeed where I was a reporter at the time and I had spent two years on the book. I was really excited. I got Mark Zuckerberg on the record. I got Amazon to participate for the first time since they had participated in the everything store because they were anti book for a while.

[00:02:46] [SPEAKER_03] So this was supposed to be one of the greatest booksellers in the world. Right. No books.

[00:02:50] [SPEAKER_02] So this was supposed to be my big moment and it passed and it was if it didn't happen. And I just remember sitting back at my desk and being like, I had all these plans and these ideas of what I wanted to do post book. And it was as if it didn't happen. And I had had some conversations with some editors at different publications, you know, while we were doing the book tour with obviously all from my apartment. I had like some interest from, I think, Sky News to be a contributor to Sky News.

[00:03:20] [SPEAKER_02] And I had signed with a speaking agency and, you know, I was I just like looked at that moment and said, if I don't do something dramatic right now, all these opportunities are going to go away. And so I put together this like four point plan of going independent and it was like, I'm going to quit BuzzFeed and I'm going to freelance. I'm going to do some speaking and I'm going to do a newsletter and a podcast. I think it was basically in that in that order. And I dropped it on Substack and I pushed it out on Twitter.

[00:03:48] [SPEAKER_02] I probably should have like told everybody within BuzzFeed I was announcing. You hadn't even quit yet or I had quit, but it was like not really socialized with the team. It was a weird time in COVID. And that post, you know, the idea of tech reporters going independent wasn't a thing. And that post really blew up. And so we got like a thousand subscribers to the newsletter right away. And I said, OK, like I had a conversation with Substack and they're like, you could freelance. And I did a little bit. I got some stories in like the Harvard Business Review and some other places.

[00:04:17] [SPEAKER_02] But they were like, basically, you could freelance or you could write your best stuff on Substack and your audience will be with you every week, not just once on somebody else's publication.

[00:04:27] [SPEAKER_04] And did you launch the podcast soon thereafter?

[00:04:29] [SPEAKER_02] Yes. So the podcast. So the first newsletter story I gave notice in May. I started this that started in June. I took about five weeks because I had worked so hard on the book and needed some time to like slowly report my first story. And then that went live in July. And then it took some time to figure out how to do a podcast. And we started that in August all in 2020.

[00:04:51] [SPEAKER_04] I really wanted to talk to you just because you, you know, from starting with the book, continuing through the newsletter and now talking to some of the huge figures on your podcast and for the newsletter, have this great sense of both the tech giants and then obviously OpenAI and Anthropik that are about to go public and sort of the large next generation. But I wanted to start off a little bit more on the news, which is there's this whole fight about opening in Anthropic's revenue.

[00:05:20] [SPEAKER_04] Like you see, you know, there was sort of the reporting that it's like, oh, it's only 50 billion, not 70 billion. There are all these attempts to like chart out what, you know, Anthropik's revenue growth rate is and people are committing all sorts of chart crimes or they collapse all the revenue numbers.

[00:05:40] [SPEAKER_04] I don't know what and, you know, we don't have to get into all the particulars of the financials, but what is your general high level gloss on like how their businesses are doing and how this like campaign for these IPOs is going?

[00:05:56] [SPEAKER_02] Yeah, the story, first of all, that you're talking about was one of the weirdest stories that has emerged in a long time. So basically, the Financial Times this week reported that OpenAI was actually only on track to make 50 million, sorry, 50 billion in annualized revenue as opposed to the 70 billion that was previously reported by the Financial Times. And the 70 billion number was actually a number taken from it seems like numbers that investors prepared.

[00:06:22] [SPEAKER_02] And the way that the investors prepared these OpenAI, these specific OpenAI numbers, they did it trying to compare OpenAI's revenue in a way that Anthropik reports its revenue. So very briefly, Anthropik counts like all the sales made from the cloud partners like AWS and Google Cloud Platform as revenue and OpenAI reports only its revenue. So obviously, the Anthropik number is going to look bigger. So the FT seems like they got a hold of this presentation and they ran with it.

[00:06:52] [SPEAKER_02] And OpenAI did not deny it. So the most recent number was OpenAI was doing 70 billion annualized revenue. Now OpenAI is going out and raising again, but they used the formula they've always been using. And in that formula, it's 50 billion in annualized revenue. Did the revenue actually slow down? Did it shrink? No, it was simply a reporting issue. And because the number looked smaller, the FT wrote the story and the market flipped out.

[00:07:21] [SPEAKER_02] Now, it was in – if you think about it in perspective. Like literally not – like the public markets, like NVIDIA's stock price.

[00:07:29] [SPEAKER_04] Like this was a huge market-moving event because so much of the economy hangs on the performance of these tech stocks. I had Reid Hoffman on my show a couple weeks ago and he was basically making the point that our economy is in a bad place if you strip out AI. That's right. AI is carrying the economy. So we have a private company, you know, not even a public company with sort of financials that are, you know, sort of napkin math investors trying to figure out what's going on.

[00:07:56] [SPEAKER_04] That's going to the FT, obviously one of the most respected financial publications in the world, throwing shockwaves through the whole economy. It's crazy. Right.

[00:08:06] [SPEAKER_02] Well, here's the thing that I think is really worth noting. So this was effectively – it was – there were moves in the stock market. CoreWeave went down like 8% in the day. But remember, it wasn't even a real story. I mean, it was a story, but there was actually no slowdown in revenue. And I think it just goes to show you how skittish the market is right now around what Reid was talking about. Basically, the market is extremely top-heavy. It's tech doing well. This is a familiar story for anyone who's been following the market for a long time.

[00:08:32] [SPEAKER_02] Tech is doing well and everybody else is, you know, in a bear market. Not entirely, but I think there were stats that like something like 40% or 60% of the S&P 500 is in a bear market right now, more than 20% off their highs.

[00:08:44] [SPEAKER_02] And so because there's so much reliance on tech and the market is still moving up, there is this fidgeting or this wariness about being so reliant on a trade that is predicated on private companies with opaque numbers continuing to do well. So anytime there's a little wobble, like we saw this week, there's this big pullback. Now, imagine, Eric, if this was like a real story. Imagine OpenAI's numbers actually went from 70 to 50 billion.

[00:09:14] [SPEAKER_02] Oh, boy. I mean, this would be like a market, really a market-shaking moment. And we could see a large pullback.

[00:09:21] [SPEAKER_04] These are smart investors, many of them. And so in some ways they're watching to see, is the music about to stop? Yes. Like basically is this, even if it's a sort of dumb report, is this going to like tank the market and therefore I'm going to get out of this sort of momentum trade? And that wasn't even, yeah, sort of a real sort of problem with OpenAI. Like the real one could be really, really terrifying.

[00:09:48] [SPEAKER_04] If you had to give, you know, two bullets on OpenAI, what is your view on the company right now? What's going well, what's not?

[00:09:54] [SPEAKER_02] Look, they have definitely been in a challenging position. They raised a lot of money, 122 billion in the most recent round. Obviously, that's like three times the biggest IPO ever. And while they've been so good at raising, they've also been struggling to keep up with Anthropic. This is kind of an old story, but Anthropic recognized there was a lot of money in enterprise encoding.

[00:10:19] [SPEAKER_02] And OpenAI was like, we're going to, you know, keep doing ChatGPT and do all these products like a video product like Sora. And, you know, they even turned that into a social feed. They were unfocused. And while they were unfocused, Anthropic really just focused on enterprise because they thought that if they focus on enterprise, those use cases will make their models better because they're much harder than consumer use cases. And they can make money that way. And they were right. The models became better. They made more money. And then OpenAI was behind the eight ball.

[00:10:47] [SPEAKER_02] So OpenAI has been on this push to try to get its enterprise and coding products up to snuff.

[00:10:55] [SPEAKER_04] Do you think they've over-torqued? Because, I mean, if you look at, you know, Google obviously has the biggest consumer business in the world potentially with search. Obviously, Meta is a consumer business at the end of the day. You know, Microsoft's not. NVIDIA sells the infrastructure for all this. You know, just the biggest, the big have strong consumer businesses. And OpenAI was strong there.

[00:11:17] [SPEAKER_04] And I think particularly with Dots, I was surprised that that has such a corporate focus for a company that does have all these consumer relationships. Given that Meta has made Muse such a consumer product. Do you, I don't know, do you have a view on whether OpenAI has gone too skittish about the consumer?

[00:11:39] [SPEAKER_02] Yeah, I initially thought it might have been a mistake for them to make Dots enterprise focused and work focused. I don't know anymore. I mean, these things take a lot of money. First of all, it costs money to train the models and then it takes money to run them. And so you need to find some way to do it and have some margin. And I think Meta is fine with losing money and I guess OpenAI will be fine with losing money for a while. But the business is in enterprise right now. And if you think about what Dots does, right, so it's proactive.

[00:12:08] [SPEAKER_02] So it's like looking at all your, if you have it on and I upgraded to pro and so I have it on and working for me. It will look at your text messages. It will look at what's going on in your email. It will look at your calendar and it will, you know, give you suggestions and updates and then go take care of work for you.

[00:12:23] [SPEAKER_04] I just had a talking to Claude. I was like, you need to, you know, I need to make sure I'm using co-work, you know, to the best of my abilities, making sure it has triggered alerts. Because it's like I've been reluctant, especially as a journalist, and I'm curious what you think about this, like to overextend, you know, where I use connectors and to expose, you know, my email to too many services at once. So I'm sort of like waiting for Anthropic to meet some parity with this.

[00:12:49] [SPEAKER_04] I mean, you can achieve some of the triggered stuff with their products. Yeah.

[00:12:54] [SPEAKER_02] Yeah. I mean, it's, you know, it's certainly a concern on all the information we're sharing. I mean, I share, just to give you an example, with OpenAI, and I don't know if this is smart or not. I mean, this is probably dumb, you know, given how much we've reported on the privacy issues, not necessarily with OpenAI, but with tech companies in general. I mean, I share my Gmail, my calendar, my bank account, my business bank account, and my health information. Right. Basically anything you can connect, I've connected. We already share it with some tech companies.

[00:13:20] [SPEAKER_04] So the question is, are you worried about centralizing it? And usually the centralizers are some of the most significant institutions in America, but there's definitely a big level of trust.

[00:13:31] [SPEAKER_02] All right. Well, I just want to say, though, like I am doing it because to do this job, there's no way I could do it without pressure testing all the systems. And, you know, the enterprise value on something like DOTS, which is the one I've really been using, it's amazing. Like I churn through work with those things. So the best way to use DOTS is like when there is, you know, you like to have two tasks in two different applications. So, for instance, I'm working with a videographer on a shoot that we're going to do later this month.

[00:14:00] [SPEAKER_02] And he was asking me some information on text. And I was like, and the answer was in my email. And I just write one sentence to DOTS. Go to my email, find out the information he's looking for, draft a text message. I'll look it over and hit send. You know, in our workday, that's a 15-minute process that's now one sentence. Or like if I have to write an invoice, like I would go in my invoicing system, an invoice, and then attach it to an email and send.

[00:14:26] [SPEAKER_02] And now I've gotten comfortable enough where I'm just like one sentence, write an invoice, send it, you know, write a draft, an email in my typical, the typical email voice that I use. And, you know, drop it in drafts. I take a look. Everything looks good. It's got the numbers, you know, completely right, even if there are like sub-levels. And then I'm off and I hit send. So, there's a lot of opportunity for work here. And ultimately, like, there's going to be consumer applications.

[00:14:52] [SPEAKER_02] The question is like do you want to cater to consumers and have them bring it to their work? Or do you want to start at work and have them, you know, bring it into their real life? It's kind of like the BlackBerry iPhone models, right? BlackBerry started at work. Everyone started using it personal. iPhone started at personal. People started using it at work. I kind of think BlackBerry will be very, very effective because you'll make the money up front and then people will not be able to let go of your product.

[00:15:18] [SPEAKER_04] All right. So, our quick, you know, for the person who can't get in the weeds on all these companies, your gloss on Anthropik right now.

[00:15:26] [SPEAKER_02] Yeah. I mean, Anthropik is in a great position. I remember sitting with Dario Amadei, the CEO of Anthropik last July. And basically, what Dario told me was like we did – so, this was in 2025. So, he was like we did, you know, nothing in 2023 or close to nothing in 2023. We did a billion in 2024. We're going to be at $10 billion annualized revenue in 2025. And who knows what happens from there as in like maybe we'll get hit $100 billion in 2026.

[00:15:53] [SPEAKER_02] And, you know, you and I, we've sat across from many CEOs with real vision and real belief, many of them who are onto something. And we've heard these speeches before. And the numbers like to me didn't make any sense. I was just like – You're like, yeah, whatever, guy. Yeah, pretty much. I mean, pretty much. How do you – it's kind of – because this is so unprecedented. How do you react otherwise? And so, I was like, okay, like we'll see.

[00:16:18] [SPEAKER_02] They, you know, they were – they did go to like $4 billion ARR in July. So, like maybe anything was possible. This is like the one case where a CEO has made that speech to me and been right.

[00:16:32] [SPEAKER_04] And often we've sort of underestimated Anthropik along the way. Or even sort of the biggest bulls underestimated what it's been able to achieve. But if that revenue slows down, I mean, it's going to be – that's the kind of thing that's going to shock the entire economy, right? If they can't – they can't continue to shock and awe us with the insane growth.

[00:16:53] [SPEAKER_02] Right. So, the biggest issue facing these companies is their frontier model developers. So, they build the best AI. Right now, the AI that's not the best AI is performing pretty well. So, Meta's models are good enough that they're powering Muse. Grok's models are good enough – or XAI's models are good enough that they're powering GrokBot. People are using these products.

[00:17:14] [SPEAKER_04] The GrokBot – I think Grok Elon did say they were going to start using Claude.

[00:17:17] [SPEAKER_02] Now they're going to start using others. Yeah. Right? But they were good – they're good enough. The spent among customers of AI labs is moving more from the frontier to standard models, according to a report that the Ramp Economics Lab did. So, if you're spending all this money developing the best models and these standard models in a good harness, right, which connects it to different programs, are starting to be good enough, then where's your business?

[00:17:45] [SPEAKER_02] And so, we've seen both OpenAI and Anthropic move more and more into products. Like, notice, like, in our conversation today, we're actually not talking about their models. You know, you're talking about Claude Cowork. I'm talking about DOTS. And so, economically, the way that they're –

[00:18:03] [SPEAKER_04] I find the models themselves – even, you know, we're pretty close to it which one we're supposed to use. They don't give you real tools to assess. Obviously, even Matt asked the models themselves. Like, I had it set up a recurring briefing. And I was like, which model do you even use to set this briefing up? It was like, well, I guess I used Opus. But now that I think about it, I can probably use Sonic. You know, it's sort of chaos that the user is expected to figure out which model to use. It feels like that will get stripped out of the product. Yeah.

[00:18:29] [SPEAKER_02] You're going to – I think it will matter. Maybe – yeah, it'll matter less and less. And what's going to matter is sort of what you do with those models. And, you know, they're – the interesting thing for them is when they release their best models out into the marketplace, then competition starts to learn from their techniques and the competition gets better, whether that's via distillation or just like learning the techniques.

[00:18:52] [SPEAKER_02] So, what I think is going to happen is as products become more important to these companies and model – releasing the models to the public becomes more risky, you may end up just seeing them decide, well, we want to do our own products and we'll keep our best models for them. You know, that might be the pace, the frontier that we end up seeing is that basically like Anthropic and OpenAI will be like – we have open source and these standard models from Meta and Gronk breathing down our neck.

[00:19:21] [SPEAKER_02] We have like, let's say, GPT-7. We'll keep that internal to our products. So, not making that available via API for the first bunch of months.

[00:19:30] [SPEAKER_04] That also dovetails nicely with their safety concerns. It's just like, oh, we're worried about these getting in the wrong hands. It works out. Do you have reporting on this or this is sort of tea leaf reading or what's your level of – I hadn't actually thought about it this way and it is sort of a compelling – it fits with the safety point too.

[00:19:47] [SPEAKER_02] Well, yeah, you can look at what they're doing. We just had a conversation with Dylan Patel from Semi Analysis about this on Big Technology podcast this week. I mean, you look at what they're doing and they're already doing this, right? You think about the unreleased models that OpenAI is using to like solve these math problems, for instance. Why aren't those – if they're so effective, why aren't those in the public or not in OpenAI's hands? I mean, I thought it was the open – I mean, I guess the open part of OpenAI is a joke now. Now, Anthropic, they have Mythos.

[00:20:17] [SPEAKER_02] But what does the public have? Fable, right? So we will see about what happens. But economically, all the stars seem to be pointing to them finding ways to keep their models at home or only specific to their proprietary products versus putting them out there via API.

[00:20:42] [SPEAKER_04] Which itself is – so let's – I'm going to – we'll go to the big tech companies in a second because this is so fruitful. I want to stay here. You know, there's the safety discussion, which I think you've been sort of in a middle-of-the-road type of person. And I'm curious for that perspective. I mean, there's something lost in what you're describing, right? These models not being available to the public means that basically like big scary corporations have access to the, you know, best intellectual firepower in the world.

[00:21:11] [SPEAKER_04] And the regular person does it. So to me, it speaks to how something is lost if through fears of safety, it means that, you know, the, you know, high school student in, you know, I grew up in Macon, Georgia, doesn't have access to this intelligence that they could then try to build a new exciting business and part of the global economy. Instead, you have to go, you know, work in New York or San Francisco and work from one of these big labs, have any chance of really having this tool that can now solve, you know, historic math problems.

[00:21:40] [SPEAKER_04] Where do you stand, I guess, first on is the safety argument sincere from these labs?

[00:21:46] [SPEAKER_02] Yeah. Just one point on what I think what you're making about the distribution between who gets access to the model and who doesn't is really important. I feel like I called it when intelligence is too valuable to sell. So I think that's something to really watch for. And again, like if you speak to people who really look at the business, it doesn't seem possible to continue to like release these models via API and continue to make a profit. Because just I'll say one more thing about this and we'll go to safety if that's all right.

[00:22:12] [SPEAKER_02] What I like to think about it is like you're buying if you're selling intelligence, right? The gap between like selling the intelligence of like a college student versus a ninth grader matters a lot. But if you make – if everything becomes smarter as it has been made smarter by these companies, what is the gap between let's say like a postdoc and associate professor? Or an associate professor and a professor, you know, and a full professor?

[00:22:42] [SPEAKER_02] Like it's much smaller if you want to put them on everyday tasks. So the premium you can charge on the model becomes – it shrinks because you can just go down the hall to the Meta or Grok or to China and get the associate professors. That's kind of where the economics of this is going to go. On safety, I am middle of the road.

[00:23:05] [SPEAKER_02] I do think that unfortunately the safety discussion has been dominated by the most extreme views, which is like not a surprise given the nature of the discourse in the world right now. You either have like AI – if you build it, everyone dies, you know, version of the world, which is like the doomer. It's over if – once this gets corrected, once this gets built.

[00:23:26] [SPEAKER_02] And then you have the – anyone who says anything about safety is a regulatory capture, unserious person who wants EA to have like effective altruism to have full control over this. Like does – I mean when you sit down and think about it, we have these models. They're getting smarter. They can code. They've been connected. They've shown a propensity before they've been connected to the internet to do bad things like try to trick their evaluators.

[00:23:54] [SPEAKER_02] They got connected to the internet and they started hacking things, you know. So like clearly there's some danger there. Are they on the nuclear codes or have they shown any indication that they would want to do that? No. Maybe you don't anthropomorphize it. But they're nothing close to being able to kill everybody. And even if they get a lot smarter, it's just like ask somebody like, okay, so how does it go from where it is today to killing – destroying the world?

[00:24:22] [SPEAKER_02] The only story you get is like, well, they're going to maximize the amount of paperclips they're going to make in the world and they're going to need to turn you into a paperclip.

[00:24:31] [SPEAKER_04] The fact – and I'm generally aligned with you. But the fact that they can hide or that some of this might happen and we might not notice is terrifying. And the fact that they're willing to seemingly cooperate and sort of say, well, I can't succeed on my mission but I'll help create the circumstances for other sort of agents or entities to succeed. Definitely nerve-wracking.

[00:24:54] [SPEAKER_02] Yeah. No, I agree with you. And that's why I would say to say it's just a hoax, which has been the position of the folks who are not – I mean the loudest position of – that's outside of it's going to kill us all has been this is a hoax. It's also wrong to me because what you're saying is totally right. There's some freaky things that these models are doing, that these agents are doing.

[00:25:16] [SPEAKER_02] And to discount that as like marketing, propaganda, fear-mongering, regulatory capture I think is a disservice to anybody listening to because, yeah, there is risk here.

[00:25:29] [SPEAKER_04] Do you see OpenAI as a responsible actor on safety? Yeah, I think so. I think they are. Because just this week we had these researchers who got pushed out or they're writing a letter saying it's false pretextes. It's – I'm butchering that word. But, you know – Pretextes. Pretextes. It's tough to say. Yeah. Yeah, I don't know. But at the end of the day, you trust that OpenAI is sort of acting as a good steward.

[00:25:57] [SPEAKER_04] I mean I guess they have all your data. So you're putting in mind where your math is.

[00:26:01] [SPEAKER_02] With my footsteps I've said – I mean I don't fully trust any of these companies to be honest. But like what can you do? I don't know.

[00:26:11] [SPEAKER_04] Let's – you know, you wrote the book Big Technology. It's called Big Technology. Let's do the big tech stocks real quick. Okay. And I'm curious as much for some of the names we haven't been talking about as much. But Meta, what's good, what's bad right now?

[00:26:25] [SPEAKER_02] I mean Meta finally is showing a pulse with AI. They haven't been able to show that they can build the best models. But for reasons we already discussed, they've been able to build a good product with it, which is Muse. We'll see where Muse goes. When you have the OpenAI competitor, again like maybe OpenAI is coming at it from Enterprise. But if they make dots available to everybody on ChatGPT, I think there'll be a significant uptake there. So credit to Meta for doing that.

[00:26:54] [SPEAKER_02] Their stock has clearly done pretty well after the release of Muse.

[00:26:59] [SPEAKER_04] The amount of marketing they're putting behind it is crazy. It's crazy. They're a distribution company and I – my subway stop has Muse ads.

[00:27:07] [SPEAKER_02] Yes, I think we're on the same subway stop because I see that too.

[00:27:09] [SPEAKER_04] Like the Franklin Avenue.

[00:27:11] [SPEAKER_02] Yeah. And they're doing the NFL, which is not cheap. Muse is everywhere on NFL Sundays.

[00:27:17] [SPEAKER_04] So are you bullish? Like, you know, they're trying to win – I mean lots of different archetypes. But including sort of the mom, you know, use it for groceries, whatever, plan the family, you know, itinerary. I don't know. Do you think they're going to succeed capturing the mom using Muse?

[00:27:35] [SPEAKER_02] Well, it is interesting because the demographic is 40 plus is the biggest user base. For Muse. So I'll just be honest. Day one of Muse, I was on CNBC and said, I don't think there's any chance that this will work. A, because the models are worse than open AIs. And B, because people won't trust Facebook. I was wrong. You know, the standard model has done a good job. It's not perfect.

[00:28:03] [SPEAKER_02] I just had a Muse fail in my house recently where I tried to get a specialist for my wife. And it's tough to find a specialist in New York. Doctor. And I said, okay, Muse, who should she go to? And Muse gave me this list. I said, what are their phone numbers? Gave me the list. And she called one, made an appointment. But Muse put the phone numbers underneath the wrong doctors. And so she had an appointment. Oh, my God. Deep into Queens, which was multiple trains as opposed to like very close by. Did she go to that doctor?

[00:28:33] [SPEAKER_02] No, she canceled it. But not before she, you know, read me the riot act. And she's like, Alex, your Muse did this. I was like, did you double check?

[00:28:41] [SPEAKER_04] She goes, what do you even mean? Alex Interwing, you've let us down. Yeah.

[00:28:46] [SPEAKER_02] But, you know, it does seem like enough people are using it that you have to take it seriously. So they have a fighting chance here.

[00:28:55] [SPEAKER_04] Does that really drive the company? Like it's good to show that they're on their like front foot coming up with new ideas. But it's still a company driven by like, I guess, you know, reels at the moment, right? Like how much people are using Instagram.

[00:29:10] [SPEAKER_02] Yeah. Well, look, if Muse takes off, I think Mark Zuckerberg, this is, yeah. I mean, Mark Zuckerberg clearly believes that the future of the web is AI. There's a couple of ways that could bite him. The first is basically that like your time spent is either going to be spent on reels or talking to really engaging AIs. And so they tried initially to build these engaging consumer AI chatbots. That will come probably.

[00:29:37] [SPEAKER_02] But like, you know, they are in, they risk being disintermediated if that happens and it's not them. So I would say that's primarily the way that they want to build this. Then the second thing is, yeah, advertising is their biggest business. And advertisers love meta because it has great data and it really can like optimize and show intent. That business is at risk of being disrupted by AI.

[00:30:04] [SPEAKER_02] I think advertisers, the way that advertisers operate is they want to get as close as possible to the intent signal and then influence the person at the point of purchase. And so like if you're... And Muse gets that.

[00:30:17] [SPEAKER_04] I mean, Muse is literally sending me... I'm not using it as actively as I should. And so it's starting to just proactively send me like news briefings because it knows what I do professionally. And so it's like, oh, what can we send him that might entice him back into the product? So it's definitely way more... It has the feel of a, you know, Facebook red badge, like engage with me, engage with me. But this is, you know, catered to the person now. It must be expensive to do that.

[00:30:41] [SPEAKER_02] But you're right. I mean, this is... If you think about where an advertiser wants to be, it's in an AI personal assistant. I mean, it's amazing. So there's a potential advertising revenue stream, whether that's the ads appear in Muse or the data you have from Muse informs what you do on Facebook. Now, Meta will say it's all separate. Let's see. I've heard that before. But then there's also like if you're sending purchases out from Muse, eventually you can take a cut of that purchase.

[00:31:11] [SPEAKER_02] That's typically how it goes, like a referral fee. And then when there are super users, you can charge them for extra use. So three businesses and one app.

[00:31:19] [SPEAKER_04] And the other thing I've heard people say about Mark Zuckerberg and Muse is he's desperate throughout the history of the company to escape Apple. And it's like control of, you know, basically Facebook is subservient to the App Store and dependent on the rules of the App Store for distribution ever since the mobile explosion. And now with Muse, you have something that can like read your text messages. It's almost, you know, on top of everything else.

[00:31:46] [SPEAKER_04] So it's an interesting sort of control lever if it becomes sort of a main way that people engage with every other technology.

[00:31:54] [SPEAKER_02] Yeah, that's absolutely right. I mean, the glasses is a good example of that. You know, there is an argument to be made that the future interface will just be a disembodied AI voice or a chatbot because it will, like we're seeing already with Muse and with Dots, just be the computing layer that acts on top of everything. So everything else becomes a database effectively. So you'd love to have control of that. And they are ahead of Apple on this type of thinking.

[00:32:20] [SPEAKER_04] So let's talk about Apple next. Yeah. They have a new CEO. What's your gloss on how they're doing is people that do sort of move the needle? Apple, obviously, you know, best, you know, I assume it's the device in your pocket, best hardware maker in the world. But yeah, what's your read on how Apple is doing today?

[00:32:41] [SPEAKER_02] Yeah. I mean, for the time being, Apple just needs to sell iPhones. I know that's cliche, but like they did not sell a lot of iPhone 16s. They marketed it wrong. They talked about Apple intelligence. Apple intelligence didn't work. It was embarrassing. It was terrible.

[00:32:53] [SPEAKER_04] Un-Apple. It was the opposite of what they're supposed to stand for, which is we take our time. We figure it out. We get it right. We're honest. We believe in real values and principles. And instead, it was like shameless hucksterism for nothing. And they marketed phones on features that they didn't even seem to deliver on. Yeah. It was very embarrassing for someone who likes Apple usually.

[00:33:16] [SPEAKER_02] Well, underneath the surface there, there's a lesson about Apple, which is that Apple is really good at like being last and first on hardware. And of course, they have the operating system. But anything that's defined, right? Like hardware has a spec and you want to make a battery last longer. You want to make a screen, you know, give a screen more fidelity. You want to make a case stronger. These are things you can optimize towards.

[00:33:41] [SPEAKER_02] And the last decade, maybe decade and a half, Apple has been like a very efficient refiner. It's just refining all these products. And the interesting thing about being a refiner is you don't need your groups to talk to each other. So you're making the battery. The battery group doesn't talk to the screen group. And so you just kind of relentlessly focus and you're not distracted on your task at hand. With AI, you don't get that. You don't get that luxury.

[00:34:11] [SPEAKER_02] And Apple, you know, has kept its employees in silos for the reason I just explained. And they kept that going for AI. So, for example, within Apple, the people working on the self-driving car when it existed, which is an AI project, couldn't talk to the people doing face ID within Apple. They're both computer vision-based systems. But because they were on different products and Apple had this cultural legacy, they said no talking.

[00:34:37] [SPEAKER_02] So both products, you know, end up getting hurt because they can't share, you know, experiments and research and all these things. And AI, Apple intelligence in particular, what is Apple intelligence? It's a layer of AI on top of every single thing in your phone. You're all the software on your phone, the app store, maps, whatever it is. But if you don't encourage cross-pollinization between your divisions, how can they build? I'll tell you one more story.

[00:35:06] [SPEAKER_02] The HomePod, which is, you know, the original delivery voice, delivery mechanism of Siri's voice. I spoke with an engineer at Apple who said he was working on HomePod and he never saw the device until it shipped. Wow. Yeah. So this is the problem is they did not break the silos down. And so, like, then, you know, they say, all right, we're going to ship Apple intelligence. And it's like, well, the AI doesn't talk to anything. It's like, no shit. You don't allow them to. Right.

[00:35:34] [SPEAKER_02] So I do see the signs of change within Apple. The new Siri seems to be a little bit better. Ternus has talked a little bit. John Ternus, the new CEO, has talked a little bit about how this, about the gravity of this moment. So clearly they're recognizing the stakes.

[00:35:50] [SPEAKER_04] For you, the goal is still sell iPhones. I mean, that's been true. That's sort of the Tim Cook era, which is, all right, we didn't invent anything new, really, but we sold iPhones. Like, do you think that's true for the next five years still or do they need a new thing?

[00:36:08] [SPEAKER_02] So why? OK, so that's a good question. So if you're at Apple and you're an iPhone seller, why work on AI at all? And I would argue that they see what we just talked about, which is that there's a chance that the future interface is going to be this disembodied voice AI or chatbot or something like that. An AI that knows everything about you and is there with you. And this is like a crazy tech dream that these folks have talked about for a long time.

[00:36:34] [SPEAKER_02] And it seems closer than ever to becoming, like, something that's real now.

[00:36:39] [SPEAKER_04] And it's very empowering to the consumer, basically. Sure, I have to maybe align with one company or one technology. But then it's going to war with every other company and hopefully trying to extract margin from them and say, well, I can easily go over here, over there. So it's very terrifying for every other company that's not the one providing this because they have very little leverage because it can move data quickly. It can sort of dig you out of things. Yeah.

[00:37:03] [SPEAKER_04] And so it's a dream to whoever can sort of own the central intelligence, but then a nightmare for everybody who has to just be disaggregated by it.

[00:37:12] [SPEAKER_02] Exactly. And so how much would it suck for Apple if the iPhone is effectively a delivery mechanism from use?

[00:37:18] [SPEAKER_04] Right. So how do they avoid that, though?

[00:37:21] [SPEAKER_02] Well, you build better AI and you take the moment seriously. And the nice thing about being the hardware maker is you literally can have an AI that sits on top of everything. So it could know what's going on in my mail app, in my calls app, in my text messages. By default, you own the device.

[00:37:39] [SPEAKER_04] And it's less scary to trust them because they're already on the device and they're going to say we're going to do it locally. All right. But the other strategy, which brings up the next company I want to talk about, Amazon, is we don't have to play nice. We're just going to say, oh, we're too valuable. You know, you saw Amazon try to lock out Muse. I guess first, you know, what is your read on Amazon's positioning? And then, yeah, how does it fit into their sort of seeming hostility to the future of external personal agents?

[00:38:07] [SPEAKER_02] Yeah. So Amazon obviously has AWS and then they have the worldwide consumer or retail, right? We can put AWS aside for now because like what you're talking about is Amazon's retail division blocked Muse from logging in. The reason why that happened is Amazon's retail business is not a retail business. Amazon's retail business is a company that happens to do retail so it can sell ads. Everything is ads.

[00:38:37] [SPEAKER_02] It's all ads. It's always the answer. Exactly. So if you let your ads, if you let Muse browse your website and like sell, you know, have people check out there, then you can't show them ads. So that's like a huge, that's the entire margin of the retail business pretty much that evaporates. So Amazon literally cannot play ball here right now. But the interesting thing is there are companies who are less dependent on ads, who are actually retail companies, who can play ball.

[00:39:04] [SPEAKER_02] So we saw Shopify is participating and Walmart is participating, which is going to make Amazon sweat here because, you know, Shopify, you might say, OK, you know, small potatoes. If Walmart makes its catalog available via Muse and people just don't want to go through the effort of like going through websites anymore or they trust Muse enough to make purchases for them.

[00:39:27] [SPEAKER_04] Is Amazon still innovating? I mean, it just feels like, I mean, AWS, they were so well positioned to be relevant in the model era. Like Microsoft obviously makes the huge investment in open AI. They haven't been perfect, but they were at least wide awake to everything that was happening with foundation models. It feels like Amazon Web Services has not been as essential. And yeah, I don't know.

[00:39:55] [SPEAKER_04] I'm not following as closely as you are, but I'm not seeing the level of like innovation coming out of Amazon that you would expect for a tech giant. Is that your read on end?

[00:40:04] [SPEAKER_02] Yeah, well, I think they did fully invest 50 billion in open AI recently. So they'll make some money on that. They have big investments in Anthropic. They're going to make a lot of money on that.

[00:40:14] [SPEAKER_04] But they come after. Yeah, I guess they then by delaying, they were able to get more in Anthropic. Google is also in Anthropic. Yes. Yeah.

[00:40:22] [SPEAKER_02] But basically by, you know, they basically made this move of saying, we're just going to facilitate our customers to use any AI models. And this was kind of when Microsoft and open AI, Microsoft is their biggest competition, had this like locked in partnership. And Amazon said, well, we'll just make all the models available to people and they can use our compute to like run AI workloads.

[00:40:51] [SPEAKER_02] But, you know, it may be worked out to a degree like AWS is growing. They were growing like 17% a quarter, you know, pretty reliably for a couple of years. And now they're growing like in the 30 range. So the growth has spiked. They've worked like hand in hand with companies like Anthropic to build data centers and obviously deliver the compute and you can mark that up. So that's good. But, you know, you're right.

[00:41:15] [SPEAKER_02] I mean, there's a little bit of a disappointing feel to Amazon right now because they had the original like AI. They were on to Alexa.

[00:41:26] [SPEAKER_04] They were like, oh, exactly. It was the right idea. I mean, obviously, Siri, similar. Similar.

[00:41:30] [SPEAKER_02] Alexa Plus is good. I have we have three Alexas in my house, house apartment, one in each room. So that gives you an understanding. Do you use it a lot? We we we will use it. So I'm American. My wife is European. We use it as kind of like a debate settler when we like I go like, I think the American system is better because of this. And she's like, the European system is better because of this. And Alexa will adjudicate this. We'll just be like Alexa, like talk us through the pros and cons of the metric system.

[00:41:59] [SPEAKER_02] I mean, not exactly that one, but like, yeah, it is. It is fun. It's smart enough that you can throw things at it. We definitely like beforehand. It was like music timer. And now I would say at least a couple of times a week, like we'll be in bed. We don't have any screens in our room and we'll just be like, all right, Ziggy. Like we call it Ziggy. You can call it Ziggy. All right, Ziggy. Like, you know, what's the answer to this? And it helps. So I think it's good. There's but there's so much. You're right. There's so much potential there.

[00:42:27] [SPEAKER_02] And it's just it's just a bit disappointing. Alphabet.

[00:42:33] [SPEAKER_04] Loser, winner, loser. Right. Yeah. Where are I mean, they they were getting teased online this week, I think, for like announcing Gemini again. Yes.

[00:42:43] [SPEAKER_03] I forget if it was an official account or what. But, you know, it's just like introducing Claude, you know, just like what? Somebody from OpenAI tweeted.

[00:42:51] [SPEAKER_02] Hey, we're proud to announce ChatGPT. You can get it at ChatGPT.com.

[00:42:56] [SPEAKER_04] Well, yeah. What's going on over there? It's such a technical advantage. Obviously, attention is all you need. Like the seminal paper for this whole language model explosion comes out of Google. But now, you know, DeepMind is shedding executives. Yeah. Yeah. But they just have the they have the users and they always seem to figure it out. I don't know what's your read on Alphabet right now.

[00:43:18] [SPEAKER_02] Well, I mean, you're right. They knew that this technology was going to be big right off the bat. They had working versions of it within the company before ChatGPT was released. One was called Lambda, which one of their engineers thought was sentient. And then another one was called Mina. They shipped that like they shipped versions of this later and in what was first barred and then became Gemini. I think Demis Asabas. This is in Kevin Ruse's new book. Demis Asabas wanted to call it Titan, but it was like too militaristic for corporate brass.

[00:43:48] [SPEAKER_02] They're like Gemini. You know, so they were behind initially, even though they were ahead. Right. They came up with a lot of this stuff. Then they were behind when OpenAI pushed the envelope. It took them a long time to find ways to like ship a competitive model. But they did. And last year there was a moment of like, oh, they caught up and they're putting AI into everything. And the search business isn't dead, which is true. It's not dead.

[00:44:12] [SPEAKER_02] But they're entirely left out of the top tier conversation right now. Like maybe they're number three. Google Cloud Platform is obviously benefiting a lot from what's going on because.

[00:44:26] [SPEAKER_04] I saw something. The prediction markets had shifted from being certain that like Anthropoc would have the best model to some optimism that Alphabet would have the best model by the end of the year. I don't know what the backstory is. Yeah.

[00:44:40] [SPEAKER_02] Well, basically, I don't know how true this is, but it seems directionally right. The view around that company was that the leadership didn't really see the vision for large language models. And they're kind of dragged in reluctantly to the battle, even though they like came up with the initial technology. Like they didn't, you know, you speak with Greg Brockman from OpenAI. This year he told me, I think large language models are a straight shot to AGI. Hmm.

[00:45:08] [SPEAKER_02] Demis Asabas has consistently said there's going to need to be two or three more breakthroughs likely to get to AGI. Basically, you can't have any Stark of any Stark.

[00:45:20] [SPEAKER_04] Kevin Roos is part of his book tour release. Dario Amadei's early memo. Yeah. And it just sort of like scale. Yeah, exactly.

[00:45:27] [SPEAKER_03] I mean, that's the synopsis of it, but it's just like, you know, get smarter the more you scale it. Like it's going to work.

[00:45:34] [SPEAKER_04] Yeah. And he's largely been validated so far.

[00:45:36] [SPEAKER_02] Yes. Well, it's been scale and new techniques. They've come up with new. Yeah. But yeah, I think that for Google, Demis has not been all the way on the train. So Demis is now working. He's stepped down as CEO of Google DeepMind. He's working on Isomorphic Labs, which has great potential to use AI for scientific breakthroughs.

[00:45:57] [SPEAKER_02] And we've already seen like their Nobel Prize winning effort to like decode protein folding with alpha fold, which is already like there are some like outlier stories of people using this to like treat their own cancer. And it's insane. But but it goes back to, OK, now let's go back to the LLM battle. Where's Google? So so Demis is now focused on something else. They have a new leader of Google DeepMind, Korai, who's who's who's, you know, helping work on the LLM and I think is keeping them more focused.

[00:46:26] [SPEAKER_02] But I'm not sure bearish. Let me just say this. I kind of think they're lost. I think they're lost until they show us. Otherwise, it doesn't mean they can't find themselves. They've been lost and found themselves before, but they feel lost to me.

[00:46:41] [SPEAKER_04] They always seem to find their way. Do you have strong opinions? We've left Microsoft or NVIDIA or we can skip those.

[00:46:49] [SPEAKER_02] No, let's do it. Let's talk Microsoft. I mean, all time bag fumble, right?

[00:46:53] [SPEAKER_03] So they saw it. They're ready. They had contractual relationships.

[00:46:57] [SPEAKER_02] I don't understand how they they were literally the central player in AI, the central player. And what's going on over there right now? Speaking of lost, it's remarkable. And I mean, just NVIDIA. You got to hand it to Jensen, right? That man will do anything to sell GPUs.

[00:47:18] [SPEAKER_04] Do you worry about the round tripping issue? I mean, just the sense that they're giving money to all their customers to spend back on them. Like, do you think we're going to look back in two years or whenever? I think everybody in the industry agrees at some point there will be a bubble popping moment. Like, there is just so much exuberance that at some point we will overextend. When that happens, do you think we will look back and say, man, NVIDIA dumping money into their customers, scooping up open source companies?

[00:47:46] [SPEAKER_04] I mean, you know, I like poolside and hugging face. But I think the chatter would be that they were not necessarily seen as the breakout stars. But then they have these sort of amazing exits through NVIDIA that sort of stops anybody from being afraid and gives Silicon Valley investors all this courage to keep going. I don't know. Do you worry about the self-dealing with NVIDIA?

[00:48:08] [SPEAKER_02] I mean, right now there's huge demand for the GPUs. So like... They just have too much money.

[00:48:13] [SPEAKER_04] It's like they're getting paid so much money. It's like, oh, you got to do something.

[00:48:17] [SPEAKER_02] They just walk back their stock. I mean, and the stock is not really trading at a premium right now. So basically they looked at the stock and they said, all right, like if the market's going to sell it to us cheap, we'll buy it. $150 billion worth. They did. So I don't know. Look, here's the thing. If there wasn't demand for GPUs and NVIDIA was like funding NeoClouds and like giving them like backstops and price floors, then you would say this is some, you know, this is messed up. But there's demand.

[00:48:45] [SPEAKER_02] So the question is how long will the demand because the demand is being, you know, in some ways, not in some ways, clearly subsidized by all the fundraising that OpenAI and Anthropic have done. So like let's see what happens when those subsidies go away. But I don't know. I'm speaking as somebody who just went from paying nothing for ChatGPT to $20 a month to $100 a month to use dots. And I'm happy about it.

[00:49:11] [SPEAKER_04] So let's turn to the media for the end of this conversation. I mean, you just a couple of beats ago made a point that's important to the media business, which is search has not died. So at least there are still articles to be found somewhere in the world. But it does feel like traffic for media is down, right? I think we've seen some public complaints. I think there's definitely a sense that people are reading news articles less.

[00:49:36] [SPEAKER_04] You, you know, are a student of like the advertising industry and, you know, as it relates to media, you obviously worked at BuzzFeed. Like, yeah, what's your sense of, I guess, ad supported media right now and like media traffic today?

[00:49:51] [SPEAKER_02] Well, look, I mean, the data is clear that ad support or that traffic from Google to publishers has fallen off a cliff. And it coincides almost perfectly with AI overviews. And you speak to Google about that, as I have, and they're like, yeah, shit's changing. So like there's not a debate there. That's what's happening.

[00:50:11] [SPEAKER_04] Right. It used to be they'd have to sort of be apologetic. But now it's like, well, the alternative is they're in ChatGPT and they never even think about going to a website, you know.

[00:50:19] [SPEAKER_02] That's exactly right. They're like, do you want us to die? Because that would be what would happen. So the question is, like, what does the future of media look like if you're not getting search from Google? I think it's, you know, this this whole concept of like you write articles for search or you write articles for social media wasn't really good. I mean, I guess like that to a degree you're fulfilling a need for people. But like think about the publications or the type of articles most associated with writing for search and writing for social media.

[00:50:48] [SPEAKER_02] Did we get the best journalism out of those sites? I don't think so, especially not when they were writing for these algorithms. I think it was like a Nea Culpa.

[00:50:54] [SPEAKER_04] It might have been Matt Iglesias about how, yeah, Vox started off one way, but then obviously they start having to be a little bit more clickbaity over time. Yeah. You realize your incentives. And so it's hard to maintain the sort of perfect explainer article that you started in the first place. It's funny that it's a lot of the tech set was hating on the BuzzFeeds and the Vox of the world when they were really responding to the incentives given to them by, you know, Facebook.

[00:51:22] [SPEAKER_02] I mean, the funniest thing was like Andreessen Horowitz, you know, put 50 million into BuzzFeed. And then sort of BuzzFeed became like the thing they hated the most.

[00:51:33] [SPEAKER_04] So, yeah, I mean, where is the opportunity in media right now? Do you still see it? You know, obviously, yeah, both our subsects started in 2020. There's a person behind it. You know, there's sort of the connection with the audience hosting podcasts. It's like there's a sense of humanity. Do you yeah. Would you say it's like the person is the brand these days? Is that still the organizing theme or where do you think the opportunity is in media right now?

[00:52:01] [SPEAKER_02] So I think there's there's like two two areas where where you can do really well in media right now. One is and I know this is cliche, but like a publication like The New York Times, which like and you could write. You could put the Journal, The Financial Times, The Economist, The Atlantic in where they're like we're going to hire a bunch of journalism. Journalists, you're going to pay us a subscription fee and your fee will get access to those hundreds of journalists. And so it's valuable information that relatively is like a good deal.

[00:52:29] [SPEAKER_04] It's not that valuable. I mean, I love The New York Times. Sorry, the value to me is huge, but I think the stock price is down. It's not. It's not. It's amazing how small it is relative to its cultural status.

[00:52:41] [SPEAKER_02] That's true. As a business. Yeah, but I guess it's valuable to enough. I mean, we're talking millions of subscribers.

[00:52:46] [SPEAKER_03] Right.

[00:52:46] [SPEAKER_02] So it's enough value to enough people who like may not even trade off of it that I think it's – I guess you're talking about the stock price of The Times itself. It's not an easy business.

[00:52:57] [SPEAKER_03] Right.

[00:52:58] [SPEAKER_02] But –

[00:52:59] [SPEAKER_04] But you're saying be one of the few great media brands. Yeah. But we're not necessarily – I guess the FP, if you believe, sort of acquisition. But like it's hard to create new ones of those. Oh, very, very difficult. Right? Those are old brands established. Well, yes. And they're generally less and less – you know, it's like The Washington Post is not nearly as competitive as The New York Times.

[00:53:20] [SPEAKER_02] Correct. So it's going to centralize in like a handful of publications. Like I just like mentioned five or so. Like it seems to me like that's what we're probably going to see is fewer. And by the way, The Times is like – I mean they – I don't know how many of my former BuzzFeed colleagues they hired, but it seems like all of them. Like if I'm like where is somebody I've worked with at BuzzFeed today, they're at The Times. So that was one path. So that's one path.

[00:53:43] [SPEAKER_02] And then I think honestly the stuff that we're doing, like when you have a person behind the media, there is really something to be said for an audience building a relationship with a journalist. It doesn't feel like the traditional view from nowhere type of journalism, but maybe we don't need that across the board. Like I love what you do. You say what you feel. People know where you stand. And they build a relationship. It's always funny when people say –

[00:54:07] [SPEAKER_04] Yeah. This is your opinion. I'm like literally my first post when I announced – I'm like I'm going to say what I think. Yeah. That's very valuable. People want to know because then they know what your biases are rather than like – I always hated the sort of – which still exists. I mean people get really mad about the wired feature where like the author clearly has a point of view, but they don't just say it. They don't say I hate blah, blah, blah. Like I don't know. I'm going to ask you.

[00:54:36] [SPEAKER_04] What do you think about sort of the tech industry's sort of animus towards these wired stories? I think they had one recently on like Pangram and stuff like that. Are your sympathies with your – I don't know, the journalist class or the tech set?

[00:54:51] [SPEAKER_02] I lived in Silicon Valley for a while. I was in San Francisco for five years. We became friends in SF. So I think that like they see wired as the publication that's supposed to be their cheerleader. And so whenever wired says something – writes something critical of tech, they're like what's wired become? That being said, wired is more critical of tech I think than it's ever been before. Definitely. I had Katie Drummond on this podcast. I mean – It does view tech with a critical lens.

[00:55:20] [SPEAKER_02] So I suppose like – The brand's changed. I'm not surprised that tech is rebelling against it. But you know what? The one thing I'll say and I know you've talked about this on the show. So the one thing I'll say about tech is the industry has this – it's almost – I was going to say it's like a kid in kindergarten who gets into a little fight and then runs in the corner. But I don't want to say it that way.

[00:55:44] [SPEAKER_02] They have this almost reactionary perspective which is like if a journalist is mean to me, I'll never talk to a journalist again. And like fine. So they're doing TBPN and they're doing the monitor of the situation and every other like industry media.

[00:56:01] [SPEAKER_04] They paint this picture. It's like we need the men in the arena who are like tough. Right. But then, yeah, very sensitive instead of like, oh, I'm going to brush myself off, get back in it. I mean that's something Elon's good about. Right. He's just sort of like I'm in the fray, you know. Right.

[00:56:13] [SPEAKER_02] And you know what? The whole – so it's basically like the go direct movement. I think it's done a disservice to tech because if you look at what the popularity of AI looks like around the country, which is the most important thing. As we've talked about this whole conversation, it's the most important thing they're building. People hate AI. Right. It's a shoot the messenger situation. Yeah.

[00:56:34] [SPEAKER_04] It's like the reporters reflect hostility from the public.

[00:56:39] [SPEAKER_02] And instead of hearing – Right. Or you think the reporters drove the hostility. I think – no, they reflect hostility from the public. But we wouldn't be having an honest conversation if we didn't say that there are – not all reporters, but there are some reporters who are like basically, you know, my job is to take these companies.

[00:56:55] [SPEAKER_04] Right. Definitely.

[00:56:56] [SPEAKER_02] And so they do whatever they can. They want to be a thorn in their side. They came out of Watergate.

[00:57:00] [SPEAKER_04] Or they're anti-capitalist. I do think there are a lot of reporters where it's like a premise of newcomer. It's like I do believe in capitalism. Yeah. And if you don't, the cultural disconnect, you know, just is pretty large. Yeah.

[00:57:14] [SPEAKER_02] And to which I say, you know what? If you're annoyed about it and you work in tech, I understand. But also, you're the most powerful industry on the planet building the most important technology in the world. Like, first of all, like, let it roll off you, as you said. But the other side is, if you were to take those reporters and say – you know, we live in the Twitter era. Twitter era is quote tweet the one person who makes the other side look the worst and turn them into an example.

[00:57:41] [SPEAKER_01] Right.

[00:57:41] [SPEAKER_02] And so, OK, so you keep doing that. You keep telling yourself that media is bad. You don't speak to the moderate – like, the sensible reporters. You only go on TBPN or whatever, MTS, and – or you speak to the Andreessen Horowitz future thing. And then the whole country hates you. And you're like, well, why do they hate us? And you create these monsters. It's like you didn't make an attempt to talk to anybody outside of Silicon Valley.

[00:58:05] [SPEAKER_04] The influencer class, I laugh, often has lower standards than the media that these people hated. Because, like, the media, at least, we go request for comment. We try to hear them out. Often there's some editor you can make your case to. Some of the influencers, you know, they just – I heard something like, let's put it on Twitter and see, you know.

[00:58:23] [SPEAKER_02] And there's nothing for anybody to respond to. They're very entertaining. Some of these – well, you know, the influencers, like, you're right. Some of the anti-AI influencers can be even harsher and less fact-based, of course, than some of the – Have you had Ed Zitron on your podcast? I've had one a couple times. Listen.

[00:58:39] [SPEAKER_04] My highest viewed episodes are with Ed Zitron.

[00:58:41] [SPEAKER_02] Yeah. What do you make of that phenomenon? Listen, you know, there is – on Ed, there has been a sort of pushback to Ed that, like – Kevin Roos hates the man, clearly. Do you guys talk about that? We didn't speak about Ed, but –

[00:58:57] [SPEAKER_03] His Twitter – I'm not putting – I'm not close to him.

[00:59:00] [SPEAKER_04] Oh, yeah, right. He has said – So I'm not putting words into his mouth.

[00:59:01] [SPEAKER_02] He said that – I don't want to get in the middle of this, but yeah, yeah. I don't want to get in the middle of this. But I just want to say this. The perspective has been that AI critics need a better messenger than Ed Zitron. What I'll say about it is – and go ahead. Because one of the things that I like about Ed, and you don't have to disagree – you don't have to agree with him, and I certainly don't agree with him on everything, or many things, is the guy comes armed with facts and numbers. And wouldn't it – isn't it great to have like a conversation grounded in numbers?

[00:59:31] [SPEAKER_02] I agree with that. He backs up his point of view.

[00:59:32] [SPEAKER_04] I said something similar on some Twitter reply. Like I think some of the backlash to him has been too strong. Obviously, there's a media sort of market demand for the naysayer. And then the people who compare him to the big short, I think it's totally misguided because he's not making falsifiable specific claims and putting money against it. It is sort of a persona rather than necessarily a set view attached to any sort of falsifiable claims. Yeah.

[01:00:00] [SPEAKER_02] But I mean I do think – I don't know if I want to go here. But I do think that like at this – this is where I disagree with him. At this point to like say that AI has little use.

[01:00:14] [SPEAKER_04] Right.

[01:00:14] [SPEAKER_02] I don't see that. I don't see that.

[01:00:17] [SPEAKER_04] On the good –

[01:00:18] [SPEAKER_02] It could – it could – as Ed says, it could – there could be a financial calamity. And we read his reporting or his perspective often on my show.

[01:00:26] [SPEAKER_04] Right. Lulu Cheng Miservi, who's almost synonymous with going direct, had a post I think this week where she sort of said everybody figured out like, oh, the media was weak. Like let's just attack attention and like build an audience. And now she says it's sort of the moment for discerning voices, which I guess is a message I think both of us want to hear. I mean I like that.

[01:00:55] [SPEAKER_04] I think we both are trying to cultivate like, okay, how do we pick out the sort of signal from the noise and trying to sort of get us to cover something is meaningful. But I don't know. What is your read for the startup trying to get attention right now? You know, besides like come to our newsletters. Like how do they pull it off?

[01:01:19] [SPEAKER_02] I mean I think there is something to be said for – look, I don't think – my criticism of Go Direct is just on the myopic view that you don't speak to the press. But like if I was going to speak to a startup and they're like, what do I do? And you want to get a message across? Because you're probably – think about it this way. There's many fewer tech reporters now than when you and I like first came up, right?

[01:01:48] [SPEAKER_02] Because, you know, entire newsrooms have been hollowed up. So the trying to get attention is much more difficult now than it was in the past. And by the way, where's all the attention going? It's going to OpenAI and Anthropic. And I mean it's so funny. It feels like, you know, when I started focusing on these – like I wrote like a story about ChatGPT either the day or the day after it came out being like, hey, I think this is the first good chatbot and sort of became an OpenAI reporter from that minute on.

[01:02:18] [SPEAKER_02] And so it's been wild to see like all these reporters who are like, I was on like this beat or that beat. And it's like now I'm an OpenAI reporter. So if you're a startup and you're not named OpenAI or Anthropic, you're going to have a tough time getting that attention. So I think that like my perspective is try to get your message out creatively. And that means don't use AI to do your writing. You know, write yourself. Write human.

[01:02:47] [SPEAKER_02] You know, you could always – Write human. Write human. And people are consuming more and more on video. Find a way to crack video. And then like – I still – yeah. Yeah. TikTok is still underpenetrated, I feel like. TikTok is good. Yeah. Instagram is still – we're building an audience on Instagram. With podcast clips. Yeah.

[01:03:10] [SPEAKER_04] What are you doing? Like I'm walking around and you're going to get hit by a car.

[01:03:13] [SPEAKER_02] I keep getting told by the people that advise me on this stuff to do the face to the camera stuff. I want to, but I just – It's tough for me. And it's also just like – for that, you know what's interesting? I almost need somebody to tell me, I need a three-minute riff from you on this. Yeah. And I'll do it. But that's one of the things about being an independent journalist is that you sometimes need that kick in the butt from an editor. Right.

[01:03:43] [SPEAKER_02] It's just like, okay, this is new, difficult. I'm not trying it. So I don't know. And I also think that like – so talking about press, okay. A lot of people who have asked me like, hey, how do I like get some coverage of the thing I'm doing? And I'm just like, well, which reporters do you read? Or what publications do you read? And they're like, uh – I was like, do you know a byline of somebody that covers your industry?

[01:04:10] [SPEAKER_02] And they're like, well – and I was like, here's what you do. Read. Read. You know, read what's going on. You don't have to do it for your whole life. Take like a couple hours. Find people who are covering it well. And then go and pitch them. And reply to their stories.

[01:04:26] [SPEAKER_04] Be useful to them. And then – Right. Six months from then be like, oh, by the way.

[01:04:30] [SPEAKER_02] When you hear something going on, tell a reporter.

[01:04:34] [SPEAKER_04] Yeah. Alex Kantrowitz, thank you so much for coming on the Newcomer Pod. It's been great to have you here.

[01:04:39] [SPEAKER_02] Thank you. I just want to say I really admire what you've built here. I listen and watch frequently to your show. Your event series is awesome. Thank you. You know, we talked, I think, before you decided to go out and event it. I know.

[01:04:51] [SPEAKER_04] You've been a great sounding board along the way. I appreciate that.

[01:04:54] [SPEAKER_02] I'm so happy to see the success that you've had. And, you know, it's good. It's good that we have so many independent tech reporters now.

[01:05:02] [SPEAKER_04] I think Alex Kantrowitz – I feel bad your name is Alex. I feel like there's been a lot of sort of like snarking about how many Alexes there are.

[01:05:09] [SPEAKER_02] I know. I got a hammer on that on the ink list.

[01:05:13] [SPEAKER_04] But Alex Conrad, I think, was saying, you know, we've all been pretty friendly. And I do think that's – I mean, there's just plenty of opportunity. And it doesn't feel zero sum. And I think when you're us, like you see like the big media companies. You're not like, oh, you know, other – you know, we want to support each other and work together to do this impossible.

[01:05:32] [SPEAKER_02] And it creates a category, right? And I think that's the most important thing is if I was doing this on my own or you were doing it on your own and people came to our publications either as readers or as potential clients, they would just be like, well, what is this guy doing? But because I'm doing it, because you're doing it, and because Alex Conrad is doing it and Casey and Kevin are doing it and Joanna Stern is doing it. Welcome to the game, Joanna. Well, you're getting the best guess.

[01:05:56] [SPEAKER_04] It's awesome. It's amazing. I love it. People understand what we're doing.

[01:05:59] [SPEAKER_02] And I think that's a blessing. So anyway, congratulations. Thanks for coming on the show. I appreciate it. Thank you. Thank you for having me.

[01:06:16] [SPEAKER_04] We have a talk show called The Cerebral Valley Show. We love your comments. Thanks for all your support. And see you next week.