Is Anthropic’s Mythos a Real Threat? And Who Pays for AI?

Transcript

Speaking of things that are experiencing some price discovery and still causing some degree of fear in the market, we also had a large drop on Mythos. Was it threat or is this just marketing? So, Anthropic gave a preview of Mythos, their new model um which has been autonomously finding bugs and things like OpenBSD, um some MPEG vulnerabilities, and both of those had been missed, quite frankly, for decades. Um They found 181 Firefox JavaScript exploits versus just two for Opus 4.6.

This is a significant increase in call it capability for AI to find various bugs. And they are continuing to manifest. Now, it's not publicly released. It's deployed via a thing called Project Glasswing to quote critical industry partners and open-source developers. But, there are some critics. So, IO Research said several vulnerabilities could have been detected by openly available models. Fortune said the era of AI-driven hacking is already here. These capabilities are not new. They're just slightly better packaged. Um and that Anthropic publicizes the threat to justify controlled rollout and build a competitive moat from Axios. So,

I'm curious what you guys are making of this thing, um especially given that things happened like Bessen convening some of the major financial players in the United States to talk about this. Is this a real threat? Will it cause problems? Will it have impacts on some of these entities and their stocks? I I think this is incredible marketing for Anthropic. It is extraordinary marketing. I read about that. I Bessen convening these CEOs of financial institutions. Reminds me of like Y2K for those that remember, which was a big nothing burger. Spent a lot of money and nothing really happened. It's incredible marketing when Claude released Mythos or announced the release of it. Even that's disputed

about it. Some say it doesn't even exist. Claude, back to Keyser Söze, Mythos is a Keyser Söze. They're they're dealing with mythology. They're using legendary concept and they're selling it from fear, right? This is a it's it's a masterful marketing that that Claude has done. When they announced that or it was leaked, cybersecurity stocks went into a tailspin. Every time Claude releases a product, these stocks go into a tailspin. I think again, that's probably behind us now. I don't know how many products they can possibly release every week.

So, 12 days of Anthropic. Uh but I think it's mostly marketing. This is uh it's just a lot of hype. I don't think that that many security vulnerabilities are waiting to be zero-day exploited by by AI. I don't know. I I I feel a little bit different. I feel like um you can't ignore the threat because living in crypto every day, um you know, one vulnerability could cost you tens, hundreds of millions of dollars. That said, most of the threats we've seen in crypto are social engineering. Um and so that's something to be aware of, but if if it is true, um to me, it's not just a threat, it's an opportunity.

You have to look at it through both lenses. So, what it tells me is that as a country, we need to accelerate AI. We need to accelerate crypto because the as this technology gets better, I would rather corral it, use it to fix our vulnerabilities because as much as they can discover vulnerabilities, AI can also fix vulnerabilities. And so, I don't know. I'm not aware of any major crypto foundation that was invited to the White House if this was such a threat. I would expect them to be at the table in the near future. Uh but like we need to embrace these tools in a major way uh across every single industry from not only from an just to be scared,

but to fix the threats that exist. And there are plenty of threats across D5. We get hacked all the time. We need to be better. And so like I think this is a humongous opportunity, not just a threat. They They want to win back that DOD contract which they lost to Open AI. This is what this is all about. It's pretty If I could go long Anthropic and short Open AI, I think they're both bubbles and both overvalued. But if I could get market neutral on long Claude, long Anthropic, short Open AI, I think we'd do really well. Claude is eating Open AI's lunch.

And they did it because they focused on the enterprise customer. They did had no distractions. No bring the guy from Apple, you know, Johnny Ive, build glasses, build a new hardware form factor, try to compete with Nvidia, let's do a fab. Let's straddle across different customers. They You know, the execution is incredibly impressive. What's happening now though is that the demand for token consumption is so high that they're throttling usage. And the quality of output just isn't as strong as it was even 2 weeks ago. Like the product market fit is there, but can you charge for value and capture that value now?

That's going to be the question. If they can, then that's great for semis and a lot of other categories out there. If they can't capture that value and it's really VC subsidizing customer consumption, then we're going to have, you know, some issues around the AI trade. So you're not going to the Worldcoin event on Friday? Uh Friday I'm in Puerto Rico, I think. So I I don't have to make that choice, but I So I will say, Rob, I think you're hitting on something important that we've talked around on the pod previously, which is as you look at the AI trade, in addition to the value capture, like what is the sustainable pricing model?

Like if this is one of those businesses where to capture a dollar of customer revenue, I've got to spend a buck 50, these companies have a problem in the long run. And as you look at what's happening with call it OpenAI and Anthropic, to some extent Gemini, like throwing the ball back and forth between each other over time, I think a lot of the competitive advantage is going to come down to cost structure as people work out things because exactly as you said, uh OpenAI lost market share to Anthropic.

And now Anthropic is throttling people because the math is not working for them as they're trying to grow and so on and so forth. And you saw Gemini try to use totally different chips to train and I think what we're seeing here is the competitive arms race is accruing ultimately to the benefit of the consumer. Like I still don't remain totally certain that these companies aren't incinerating huge amounts of capex that they will never get back. And that the end value will ultimately be in the hands of the consumers. So, I know we've talked about this before, but hilariously, the name I remain the longest in terms of benefiting from AI is Apple.

Right? Cuz they're just going to use somebody else's and sell it and they've got all the devices, so you're going to have to play by their terms. Well, look at Google or Meta. I mean, these are ad businesses and they charge $0 for their products. So, you've got a price cap. And Google also has an enterprise focus, too. So, it's a it's a tough spot to be in, but you know, Anthropic's got the attention. So, on the note of building for consumers, like I've got a friend of mine who's in the creative space and one of the hardest things to get people to switch out of is web browsers like I don't know if anybody's particularly familiar with that market but I would say most modern web browsers are like

riddled with security vulnerabilities over time and do all sorts of user data tracking and there are better options like brave and opera and people just won't use them because the switching cost is too high in many cases but AI has her using the perplexity browser now. So to me if you're building for consumers in the right way exactly as you were saying it's like the distribution and consumer capture is where the value is going to be. If you're competing purely on capabilities I feel like are you in this forever escalating arms race where you're getting revenue but your cost scale even faster?

All right. >> Yeah, we saw that with fiber optics in the dot com era. Right. I mean >> We had uh what valuations are based on number of eyeballs. Right. I was going to say I feel like part of where we're going to start seeing it now is potentially in advertising cuz I want to pick on another market thread that we've talked about before related to AI here Ram which is we're starting to come up you know as the calendar rolls to when we're going to start getting results from people and we would normally get into buyback season. What is going to be happening with many of these stocks that for many many years have had very strong tailwinds due to significant free cash flow and buybacks this time around?

That's a great question. That's a great question cuz last year we had a trillion dollars in buybacks. A trillion dollars in buybacks and stocks have buybacks up from stocks that don't have buybacks. There's a bid to the market and much of this correction that took place happened during the blackout period when most companies cannot buy back their stock. It's like that bid matters. Buybacks are back now by the way. It's like they're on. And not surprisingly, like markets are back on, too. They're acting normally. So, I think you raise a great point, and it's it's about positioning the right names. Names that still have Right, free cash flow still matters.

Like Amazon's is a good question on this. Amazon is trying to compete with Starlink. They're launching low earth orbital satellites. They're going to launch a phone. They're building more data centers. They got to invest in Trainium. All that is called cap That's all expense. I The vision's amazing, but I love the vision. But, the free cash flows are low. So, we'll have to see. But, on the other hand, you have other businesses where they've got a viable business model. And uh a lot of pain has been impounded in these prices, also, too. So, you know, everything does come back to cash flow. I think it's the right metric to focus on. The CFOs have also got smarter around this. They they see what's happening in the market. So, like, okay, I got it. I got to watch CapEx. I can't overspend.

Um but uh there there still legitimately good opportunities in the market. Like, they're 20 30% discount in EVs. Like, you're supposed to buy these assets. Like, I'm not saying all the assets. I'm not saying assets that don't have good free cash flow. But, there are pockets in the market that have, you know, I would call compelling value. It reminds me similarly to like April of last year, almost to like the day, within weeks, let's say. Yeah, for me, the market has come to broad lie in many ways on the bid.

Right? Cuz if you look at shareholder returns for a lot of things, like say the mag seven, a lot of that has been propelled by buybacks. Especially if you're also at companies where they're going to be issuing stock as part of compensation. If you're not simultaneously buying back, you're just going to be diluting, diluting, diluting over time as you do these things. Um that to me, I think is the most under-discussed story of AI is not just that all the capex went into like say Nvidia. Right, but also that it's going to come out of the buybacks and we're going to see a much greater differential in the market. It could be that the lasting effect of rates going up and then the Iran conflict is actually the decay of correlation in

markets. Because for many years we saw it increasing. Well, so Apple wouldn't be where it is today without those buybacks. Yes. And Buffett, if you recall, he exited his Apple position. And it was almost dollar for dollar matched by buybacks from Apple. It's $110 billion he was selling and Apple announced a buyback program like around $110 billion. So, they bought out. Stocks go up if your capital allocation is productive. So, the number one driver of stock price returns actually is your capital allocation, right? So, if you're buying a stock at too high a price, then your book value per share isn't compounding efficiently. So, there are trade-offs. You get a nice technical bid. You get flows from those buybacks.

But, uh that can create some downside protection, but it's hard to get an outstanding return unless they're buying back their stock at good, cheap, discounted levels. Which is what Berkshire Hathaway's doing right now when their price to tangible book is just north of 1.3. I prefer that actually versus some of these other companies. Yeah, and in the crypto space as well, we've seen a lot of projects focus on buybacks because that's the only way they could do it. It was hard for them to return capital to their token holders uh from regulatory perspective over the years.

What I think most people understand is that many of these projects are awful about timing the markets. And they always buy back at the exact wrong time. And so, what we like to see is something that's more programmatic rather than discretionary because they're tend to be terrible from a timing perspective. Uh something that we continue to watch. I mean, you see the same thing in traditional financial markets as well. If you look at the people who have done well with buybacks, it's either people who sort of nibble, nibble, nibble, nibble perpetually because they can ensure they do average by using that method. And if it's an otherwise very profitable business, that's a completely viable approach.

Or it's the people who do exactly what Rom was saying, which is I allocate a bunch of money for buybacks, ideally somewhat quietly, and then when the market pukes and I'm down like 20%, I start buying back, right? You want to have an average level over time and bid at the lows of that range and just sort of sit at the lows and take things in. But, you know, I think it's a call it inverse of the problem that MicroStrategy has and part of why I get increasingly skeptical of that is they seem to have an uncanny ability to buy Bitcoin at the highs, right? Which, to go back to Rom, something you said earlier, like whoever's executing over there, you need to do better. Like you were saying

>> me up. I will do the service for you for free, okay? I'll build you a little bot, okay? But, yeah, I know, I agree. I agree. And like, MicroStrategy is a capital markets issuance machine. Yes. That's what they are. It's It is kind of phenomenal. It is almost like a ninth wonder of the world. Uh, I I I I I be a little cautious, right? Think it They've got all these tranches, like stretch, STRC and STRF, where they have a contractual obligation to pay out billions of dollars in dividends on the Right, that's the liability side.

That's how they're funding their asset purchases, right? They have an obligation to make those payments. But, the assets don't generate free cash flow. So, like, this is like, you know, Chris, we all we all know this from like 2008 have a days. Like you got you got a mismatch in the assets that you're financing and the only way at some point in the future they'll be able to generate the cash to deliver to those bondholders is by selling their common stock. That's called dilution.

So, that's not good for MSTR. Or It works as long as Bitcoin goes up. As long as Bitcoin goes up you're you're good. Citrea changes how Bitcoin scales. With a trust minimized BTC and a native stablecoin CTUSD, Citrea enables Bitcoin capital markets with lending, privacy, Bitcoin yield, and more. Get started at citrea.xyz/unchained. >> Etherfi is giving Unchained listeners 15% cash back on food and ride apps. And that's on top of the 3% you get on everything else. Your bank is charging you to use your own money. I switched.

Go to ether.fi/unchained to claim your discount. Introducing Nexo, the premier digital wealth platform. Receive interest on your digital assets. Borrow against them without selling. Trade a variety of cryptocurrencies. All in one platform. Now available in the US. Get started today at nexo.com/unchained.