Why DeepSeek Disrupted Markets & How Tokens Are The New GTM | Roundup

This deep-seek moment was one where it compressed the timeline for the market's patience to start asking, when is the P&L gonna show the benefits of all this investment in AI? Someone asked me, what's the most overhyped sector right now? I said AI. And it's not because I don't believe the impact is gonna be massive. Don't get me wrong, I think it's hugely transformational. But like everything, and we see it in crypto, in extremes is there's always— it's never a question of how impactful a piece of technology is going to be. It's, it's more so the spread between the timing of that impact and when it's actually realized relative to baked-in expectations.

This perfectly encapsulates what I think is happening here, which is dog coins lead to cat coins lead to celebrity coins lead to AI agent coins lead to Trump coin lead to actual companies launching tokens.

Today's episode is brought to you by Geodnet. Discover how the network's native token, G-E-O-D, is powering the next wave of the robotics revolution by hitting the link in the description of today's episode.

What's up, everyone? Welcome back.

GM.

GM. I feel like GM went from being like the thing everyone said and then it was like super cringe.

And then, and then now GM is back.

No, look, okay, it outlasted We're All Gonna Make It. Like, you know, when Zuckerberg's sister, uh, put out that video, the—

oh, Randy.

Yeah, that was top cringe moment, and then that sort of killed it. But GM is— it's nice, you know.

Jim, good morning. Yeah, good morning.

We are recording in the morning. It's a good morning.

Yeah, early. 5:30, 6:30.

Yeah, 5:30 in the morning. Yeah, nice commitment to Blockworks.

To Empire, baby. To Empire. So let's see, we have some, some updates. Um, one is if you are a 1000X listener and you usually listen on the Empire feed, um, 1000X now has its own feed. So you no longer have to get stuck with Santi and me. You can just go listen to Jonah and Avi if you want it. So that's, that's one update. Just go, go find the 1000X feed in Spotify or Apple or wherever you do it. And yeah, go, go follow and subscribe and rate that show. Helps us out a lot. The other one is DAS. DAS agenda is live. We're about to crank ticket prices higher. So if you don't have your DAS ticket yet, go do that. I think I have a code. I think it's, I think it's YANO10 or EMPIRE10 or like—

It's usually like EMPIRE20, EMPIRE10, EMPIRE15, EMPIRE5.

I would, I think I would go on the DAS page and just start typing in EMPIRE50, 40, 30, 20, until you hit one. I have— we have an official code, um, and I think it's one of those. So go, go try it out. Uh, it's gonna be great.

I was looking over the agenda yesterday, and, uh, I'm gonna be there. It's, uh, looks like great lineup. And yeah, I'm gonna be there. It should be a very exciting conference. People ask me all the time, like, what is DAS known for? It's really— it's skewed historically a lot of institutions. But I think this year being like so pivotal, I think, yeah, I think it'll be very, very good.

Yeah. Yeah.

Comparing it to last year's in London, I'm excited to do that comparison.

Yeah.

Last year in London was by far our best one yet too.

I mean, we had 1,700 attendees, which for DAS is massive. And we did it at the Hilton in London, which I thought was like a fine venue. It wasn't anything amazing, but the Hilton shut us down. They basically said you have to stop. Fire code, you have to cap it. So we had 1,724. This year I think we'll have like 2,500 attendees. We had to get the— we had to do it at the Javits Center in New York.

Yeah, this is where I pick up the New York Marathon bib every year.

No, it's crazy. Like, I never thought we'd be hosting events at— especially not DAS at Javits Center.

The big venue. Yeah, we'll have to do— we'll have to do some cool Empire side or meetups.

Yeah, we've got— yeah, 100%. Yeah, it's great. We have Michael Saylor, Cathie Wood, Novogratz, Mohamed El-Erian, who's the chairman of PIMCO, Richard, the CEO of Binance. We've got like congressmen and we're close to confirming a senator, hopefully. So yeah, it's crazy lineup. Go get your ticket. Empire something is the code. Empire something. Empire something. I think those are the updates. Anything else? Cool. Oh, Santi, we finished a— we finished the podcast studio. So if you— Blockworks, we have an actual full podcast studio. We're gonna start doing more in-person recordings in New York. We are maybe going to do our first in-person recording next week. So TBD. But yeah, that's exciting. So next time we do a poker night, you guys can all see the studio. It's awesome. So, all right.

Nice.

News of the week. I mean, Deepseek You know, I know this is a crypto show, but tough to come on here and not talk about Deepseek and AI because it impacted the entire market, right? NVIDIA sold off 17% and lost, I think, $500 or $600 billion. It was the single worst stock loss in history. And that, you know, spilled over into all of the Mag Seven, which then spilled over into all of tech stocks, which then spilled over into crypto, which then spilled over into basically every asset got impacted by this. So I think we should talk about Deepseek. Maybe let me give— at the risk of being someone who knows really minimal about AI and pontificating about AI, I did end up reading a lot about it this week just because it impacted everything. And I think the best article I read was probably by Dario— I'm blanking on his last name, but the CEO of Anthropic, Dario Amodai, I think it is. We can link this article in the show notes, but I'll try, I'll try to summarize it here if people don't want to read the full thing. It's called On Deepseek and Export Controls, and this is by the CEO of Anthropic. So basically what happened, to give people a summary if they weren't paying too close of attention, is there's a— Deepseek is a Chinese quant hedge fund, actually, and very successful quant fund in China. I think they're in Beijing, and they've basically turned into more of an AI company. So they actually launched this thing called They launched Deepseek V3, which was a pre-trained model. They launched that about a month ago, and that was this actually really innovative release. And it showed that they were able to essentially come up with more efficient ways to train a model. And that was, I think, for people who are really deep into the AI world, actually even, even us, I think we talked about it here. Like it was a very impressive thing, this V3 model, but it was a fully pre-trained model. Then they came out with R1 last week, which I think was a— which is their first reinforcement learning model. And from what I've read, this was actually less impressive, but it kind of spooked— but it kind of caused people to realize what they had built with V3. And this is what triggered this market panic. Nvidia sold off 17%. And I think for probably two reasons, right? The initial gut reaction was A, holy shit, they just built this thing with, I think the headline number was like 5 or 6 million. So A, we don't need as many Nvidia chips as we thought that we were gonna need to build these things. So Nvidia sold off. And then B, the profit margins of like Google and Facebook and Microsoft are gonna be way less than we thought. Then people start digging in. And I think the realization, as we've been able to sit with the news for a little bit, is Deepseek's main achievement here was engineering efficiency, and they managed to do this like k-value cache, which I don't understand what that is at all, but they achieved performance. I think the actual performance is close to what some of the US models from like roughly 6 to 10 months ago look like, but at a much, much, much lower cost. The thing about the cost is that there are the $6 million saying that they spend $6 million, whereas US companies are spending billions, is very misleading. So the Claude CEO who wrote this article said Claude 3.5 Sonnet, which in my, in my opinion is the best model right now. He said we spent a few tens of millions. So let's call it $20 to $30 to $40 to $50 million to build this thing, not billions. Now there's a massive infrastructure build, right, where you have to spend hundreds of millions or billions. And he said that Deepseek's total infrastructure investment, where they have 50,000 Nvidia chips, they spent roughly— his estimate was about $1 billion in hardware and infrastructure alone. So that is comparable to US companies, right? So there's a natural cost reduction that comes from these training models. And he said, look, like you would kind of expect the industry cost reduction curve to where we're at in the the AI cycle to be kind of getting 4 times more efficient per year. Every year we should get about 4 times more efficient. And the Claude model is roughly, I think he said, about 8 times more efficient. So definitely more efficient, but just from some like great engineering efficiency improvements, not from some like they were only able to spend $6 million, not $1 billion. It's, it's comparing apples to oranges. So if you look at So anyways, I think that was the, I think that was the biggest takeaway is like we're at this crossover point where reinforcement learning for reasoning is becoming super crucial. Smaller investments can have bigger gains, but this isn't a, oh my God, sell everything, nobody's going to need Nvidia chips type of moment. The $6 million does not include the costs associated with, you know, research and buying all the Nvidia chips and all that kind of stuff. So yes, you can build a $6 million model if and only if a research lab has already spent hundreds of millions, maybe even billions on prior research and has access to these large clusters and stuff like that. So, but I will say it was still very impressive, right? It was the number one app in the App Store ahead of ChatGPT. Claude's never been able to accomplish that. Gemini has never been able to accomplish that. It is comparable comparable to O1, right? OpenAI's O1 from a quality perspective. It definitely lags O3 from what I've seen, but it, but it is as good as O1. And it seems like there were real algorithmic breakthroughs that led to it being dramatically more efficient to train and to inference. So my— anyways, my layman retail dumb take on this is that if this was put out by— if this had been done by a Google, a Microsoft, a Facebook, a US company, nobody would be panicking. Nvidia would not have sold off even 2%. There would be no panic in the market. But the real panic comes from this being a Chinese company and us saying, holy shit, we thought we were winning the AI race, right? Microsoft and Google and Facebook. But oh wait, there's actually this Chinese company that not only has been able to build this model that's great, but actually somehow they have their hands on, you know, 50,000 Nvidia chips, which there's all these export controls, which we can talk about if we want to. But wow, how did they actually get 50,000 chips without us even realizing this? So anyways, yeah, I did the, I did the retail thing and I bought some Nvidia on the 17% sell-off. And but yeah, man, I'm very curious to get your take because you spend probably more time than me in the weeds of AI.

Today's episode is sponsored by Geodnet. They're a leading DePIN protocol that's providing critical location and spatial intelligence services needed for AI-driven robots and drones. Discover how the network's native token, GEOD, is powering the next wave of the robotics revolution by hitting the link in the description of today's episode. Today's episode is brought to you by Unichain, an upcoming Layer 2 that's purpose-built for the next generation of DeFi applications with up to 95% lower fees than Ethereum and 250-millisecond block times coming soon after launch. Unichain is designing the future of cross-chain liquidity. Stay ahead of the launch by visiting unichain.org and giving Unichain a follow on X.

Yeah, I first heard about it over the weekend. Purnav was supposed to be on, like, wanted to talk about markets, and when he reaches out, there's something going on. Then I pinged a few other folks like Casey, which I'm likely going to record a podcast with her. Casey, we've done a few podcasts on AI. I'll share some more her thoughts because I spoke with her, um, and then I spoke with like a few other research analysts and whatnot. I'll, I'll add some more just color. So obviously, um, Mark Andreessen had a good tweet which is like, this is a Sputnik moment for AI. Um, you know, like Russians kind of beat us to the moon. This is like, well, are Chinese beating us in the AI race? Thought that was well articulated. Uh, as you said, the market was down a lot across the board. Um, the market coming into this was very nervous. I mean, the Magnificent Seven already accounted for like 38% of the S&P which is like a historic high. And so there's all these comparisons saying, oh, it's, it's never been that high of like a select few stocks. I think the prior moment in time was like in the '80s or like where Peter Lynch was a manager and like, yeah, but anyways, historically the data suggests that like when that happens, like there's a, there's a pretty big correction recalibration. I'll add a little bit more data points before I go into just my take because again, I'm not an AI expert. No one really saw this coming at all. Is there's a really good video from the CEO of Scale AI, Alex Wang, on Davos that people should go listen to. Maybe we can link on the show notes and he just— it's a 7-minute clip. He goes into his, his thoughts on Deepseek. And of course you need to just take that.

Well, I think his main takeaway from that interview was, hey guys, actually they have way more chips than people realize. Yes, they didn't go build this thing with like a dozen chips. They built it with like 50,000 chips. So, right.

Yes.

Yeah.

Yeah. So anyways, everyone has an opinion. Just add a little bit more context of why the market panicked. And the interesting thing is crypto started to sell off over the weekend, which is so interesting from a like a risk barometer perspective. So you just knew that as crypto started to sell off over the weekend, like stocks would— like the open on Monday was going to be pretty—

which is not— which is not— which is a very important thing. I think it's not to be dismissed. I think the all of traditional equities markets are now looking at crypto over the weekends because it's the only market that's open. So it's your best signal, right? You used to look at like pre-market and what are options doing? And like now, now it's—

I mean, I'm pretty sure like if you're a sophisticated futures trader in New York and Wall Street, like you're looking at like Bitcoin constantly.

Yeah.

To infer some signal. But yeah, you're right. Like, look, companies like— what's a big deal? The big deal is there's all this hype in AI. Companies spending huge amounts of CapEx. Like OpenAI on Tropic spent over $100 million just on computing power. What they do is they acquire massive data centers with thousands of GPUs, and it's like a huge, huge factory to run these models. And then Deepsea comes along and says, uh, what if we can do this with $5 million, not $100 million? Which basically, as you said, uh, and they did that. Um, and people were like trying to piece together and scrambling like how they did it. Uh, there's, to your point, a little bit of like, we just don't know, right? But, um, there's a couple of things that I'll just add color from Casey, JP Morgan, a few other people. Like they, they reimagined everything from scratch. Casey described it as, and I'll quote, it is an interesting showcasing of how top US A-labs like OpenAI have deprioritized architectural innovation in favor of scaling compute. Uh, meaning like Deepseek was like, maybe with the current constraint of not getting their hands on enough GPUs or just resource constraints, instead of throwing just much more compute at a problem, they're like, architecturally, can we optimize this thing? And I thought that was an interesting observation. And she said, it's logical to me OpenAI has little incentive to reduce the cost of training transformer-based models as their financial resources are part of their competitive moat, with only a handful of organizations able to afford such training. And so OpenAI benefits from maintaining this barrier. Uh, given this, it's unsurprising that this innovation didn't emerge from OpenAI or Anthropic. So it's actually an interesting, like, take of not oftentimes the most resource, um, able company or competitor is going to innovate. And sometimes having a resource constraint, constraint, not having enough money, can lead you to think in, in a different dimension that can unlock more output out of a particular kind of thing, right?

And then I'll go on that, Santi. There's one thing that from, from this article that I think was important. Let me see if this is— if I'm sharing the right thing. So I thought, I thought this guy Dario laid it out really nicely. He said there's three dynamics of how AI gets developed. There's basically three scaling 3 basic dynamics of AI systems. There's scaling laws, shifting the curve, and shifting the paradigm. And I'd never actually seen it laid out like this, but it was very helpful for me to understand. So scaling laws. So all else equal, scaling up the training of AI systems leads to smoothly better results on a range of tasks. So for example, a million-dollar model might be able to solve 20% of coding tasks. A $10 million model could solve 40% and a $100 million model could solve 60%. So, you know, you've got these factors of 10 corresponding to the difference between like, I don't know, might be like, I don't know, an 18-year-old like freshman in college to a PhD skill level. Like, so the more money that you're spending, the better, the better these things get, right? That's the scaling laws. But then you get shifting the curve, which is what I think Deepseek did here. So Shifting the curve, he describes as like just making things more effective or more efficient. So instead of like if an innovation is a 2x compute multiplier, then it allows you to get instead of the, you know, what do we say before 10, you know, you spend $10 million to get 40% or $100 million to get 60%. Now maybe you can spend $5 million to get 40% or $60, $50 million to get 60%. Instead of the $100 million. So he said every frontier AI company discovers a lot of these compute multipliers all the time. Oftentimes they're very small, like 1.2x or 2x, and very, very, very infrequently they're a 10x. And what Deepseek might have done here is actually figured out a 10x better way to actually train these models. And then, and then the third one is just shifting the paradigm, right? And that's you know, 2020 to 2023, the main thing that was being scaled was pre-trained models. And then in 2024, last year, obviously we came up with this idea of reinforcement learning. So that's like, that's shifting the paradigm. So anyways, that's kind of to Casey's point, like they were able to do number 2, which is really shift, shift the curve here and come up with a compute multiplier.

Yeah, there's a couple of two other things I want to add. There is like they introduced a multi— talk about shifting the curve or the paradigm is they introduced what is called the multi-token system. And for context, normal AI reads like a first grader. It's like every single word, like, the cat sat, and it just tries to predict the next token. And DeepSeq kind of processes an entire phrase at once, making it twice as fast and 90% accurate. And so when you're dealing with billions of words, it makes a significant difference. So I was like, point number one that I find quite interesting, this multi-token system, like reading an entire phrase at once, not single word or token. Um, and then they develop what is called an expert system. So instead of— think of like, instead of having one massive AI trying to know everything, and the analogy there would be like having, uh, one person be a doctor, a lawyer, and an engineer, and you never know at any moment's notice you need to be all, you know, one or the other. Um, they use a specialized experts at are activated only when needed. So, so to put a finer point on that, like traditional models have all 1.8 trillion parameters active all the time. So you have to be— again, all the different professions have to be ready on call at any moment's notice to be able to kind of like surface insights. Um, DeepSeq uses 671 billion in total versus 1.8 trillion parameters, uh, but only 37 billion are active at once. Um, and, and so it's the, again, the further analogy, and I know like people, you can't reason through analogies, but it's like having a large team but only calling experts when needed. And that is another efficiency gain. When you talk about like that $100 million versus $6 or $5 point something million, it's really that like specialization and reading phrases, not words, that have allowed them, people think, this sort of efficiency gain. So anyways, not to sound like an expert, which I'm not, I'm just kind of regurgitating stuff that I'm hearing. I think there's a couple of like implications about all of this. The biggest one is you're wondering, um, are AI services inflated? And what about the CapEx spend that these companies have been doing and going forward will be doing? And so naturally, in my predictions coming into the year, I felt NVIDIA is going to be the single most important stock to watch. Um, and because that's a proxy for just everything in the market, I think just the, the amount of excitement for AI. And NVIDIA makes its money through a couple ways, but like you have to look at when analysts are looking at the earnings report, it's like how much GPU, how much H100s have they sold, how much, like what's the order, uh, and the backlog for that. So if all of a sudden you don't need a $100 million CapEx investment to run these models, you have like an order of magnitude less, then obviously that, that's a big adjustment to your model for NVIDIA. It's like they're no longer maybe going to sell as much, um, hardware as you thought they would. And they make some money on CUDA and like software, but nonetheless, I mean, it's, it's a big hit, I think, to why people were adjusting. And then there was like a— I don't know if that's like the right interpretation. I've saw some memes of like that being a very belt, like middle curve take. And the left and right was like, no, this actually just increases the demand for GPUs.

Yeah, Jevons paradox that everyone was talking about.

Yeah.

Which, by the way, is— I'm not going to pontificate on this. Actually, I will pontificate. It seems obvious to me that that is actually correct. It seems like you— I'm just thinking about this from a company's perspective and people allocating capital. Like, let's say you're— let's say things are getting cheaper to build. You're not going to stop building the thing. You're, you're not going to allocate less resources. You're basically just going to try to go faster. Like for like, let's use Blockworks Podcast, for example. It's gotten way cheaper to produce these things with AI, right? And just like there's way— there's all these efficiency gains. We're not going to create the same amount of podcasts for cheaper. We're going to 10x the number of podcasts that we're making, right? So we're going to, we're going to actually spend the same amount of money and just 10x the output. You don't decrease. So it's—

look, I think it's— we'll talk about this a little bit later, but it's hard to piece together how much like the deflationary aspect increases demand. Like, and the point I will make in this podcast is one, and it's just how I reacted to it was it's going to take the market some time to process all of this. There's earnings report coming up. There's a Fed. Meeting that happened yesterday. We can talk about all of that, but my initial gut reaction was, um, this deep-seek moment was one where it compressed the timeline for the market's patience to start asking, when are— when is the— when is the P&L going to show the benefits of all this investment in AI? And, and, you know, I was on a podcast yesterday, I said Someone asked me what's the most overhyped sector right now. I said AI. And it's not because I don't believe the impact is going to be massive. Don't get me wrong. I think it's hugely transformational. But like everything, and we see it in crypto in extremes, is there's always— it's never a question of how impactful a piece of technology is going to be. On the margin, you might argue it's more so the spread between the timing of that impact and when it's actually realized. Relative to baked-in expectations. And I think this is— it comes out and it's a huge shock to the system, which is like, you know, NVIDIA is selling all this specialized hardware with 90% profit margins. OpenAI and Anthropic are selling their services at a pretty hefty— they're pretty expensive. And this competitor just comes in and says, no guys, it can be done far, vastly more— vastly cheaper. So there's a natural disruption to that. Um, so there's two things. One, like, um, I think if you're an analyst looking at a company that spent hundreds of millions of dollars on, on, on GPUs, you're going to ask the question in the next earning call saying, hey guys, what have you done and when is this impact going to show in the bottom line? Like, so my point is, if we had 2, 3 years to prove to the market that this massive investment in AI is going to prove out, I think the market might just lose its patience a little bit more because you have this thing called Deepseek, and now you're, you're going to put more pressure and be less patient. Um, and, and so naturally I think that, that might just break the hype cycle, you know what I mean, uh, of AI. Um, is it gonna— I'm not dismissive of like the transformation the companies will be able to realize. It's just more so the, the timing of it is compressed. Instead of having maybe 2, 3 years, maybe have a year now. This will be the year where companies will have to prove that they're, they're spending 10, 20, 30, 100 $200 million in AI and CapEx, they're going to have to show that result in the, in the bottom line, you know.

100%. So yeah, so what you're saying is like the OpenAIs maybe don't need to show the profit because they're a startup, but Google, Facebook, uh, Microsoft, the analysts on, and you know, the street is going to start saying, hey guys, when, when do we actually see these margin improvements?

Hey, I'm paying— I'm going to be paying a lot of attention to earnings calls. Microsoft, I think, just reported So stock was down quite a bit, but, you know, listen to the commentary. Like, I like to— actually, you could put the earnings call transcript into Deepseek and say, you know, give me all the commentary from Satya or from, you know, all these executives around AI. And I think they're going to be put on the spot of trying to answer this question, which is, hey guys, like, great, you've spent all this money, you're an investor in OpenAI, how does this kind of change your your model, right, a bit. Um, but it's incredibly exciting. Also, the other thing that we, we've discussed it in some AI pods before with Sam from Kusco and Casey and Dana from Modulus, it's this idea— and this is Casey's observation, I spoke with her quite a bit over the weekend— was it really just reinforces, um, that open source models are beating closed source models. And I think that's both quite positive for crypto, um, And I think that was one observation that Casey had, which I also agree with. And so I think that will have a spillover effect into crypto, into crypto AI. Not to sound overly jargony and hypey, but I do think that Zuckerberg has been teasing at this also in investor calls and Facebook events, which is he's very much now like Llama open sourcing these models. And so that's, I think, just bodes very well for crypto as an open source movement.

Geodnet is a first-of-its-kind DePIN protocol that's revolutionizing satellite navigation services. They're actively generating on-chain revenue that's growing 500% year over year by providing precision location services that are 100 times more accurate than standard GPS. On Empire, we always talk about finding how crypto can power completely new business models, and they're a perfect example of this. They're already trusted by over 30 enterprise customers across industries including automotive, agriculture, construction, and even robotics. The network is powered by its native token, G-E-O-D, and easily mined with their satellite miner. By setting up a base station, you can get involved with this, uh, groundbreaking Web3 ecosystem. Learn more by heading over to geodnet.com or just by hitting the link in the description of today's episode. Today's episode is brought to you by Unichain, an upcoming Layer 2 solution that's changing DeFi forever. It's built with a vision to be the home for cross-chain liquidity and DeFi innovation. As part of the Optimism Superchain, Unichain brings you up to 95% cheaper transactions than Ethereum's base layer. As well as near-instant finality with 250-millisecond block times coming shortly after mainnet launch. To make things even better, seamless cross-chain transactions. If you're a developer, this means you can build DeFi applications that truly push the envelope of what's possible on-chain. Visit unichain.org or follow Unichain on X to stay updated on the launch, upcoming features, and getting started on building your dream applications.

So Santi, can we pivot a little bit into talking about what did you do with your personal portfolio?

Okay.

Yeah, I think that's what I want to know.

That's what—

because I know, I mean, you did message me over the weekend, say, you know, you're trimming a lot of your positions and this was kind of spooked you and you're moving out of the— not out of crypto, but out of a lot of the risky stuff. So what did you actually do? Like what? What? Yeah.

I sold all my, uh, I have like a wallet that is just more hot in the sense of like more like tactical, I'd say. So I sold, um, all of it. It's not, it's not a math, it's like less than 5% of my portfolio.

Were those all your meme coins?

Yeah, those are, those were the— that's a meme coin bag, like everything when I punt meme coins and stuff like that. So like I, I'd rotated back into SOL largely to just be a bit more defensive. And then I went to the stables over the weekend.

Hmm.

I'm like, I'm just going to sit this out. There's going to probably be an overreaction from the market at some point. I just want to be, you know, and was paying attention. It's like it's going to be a big week. It's earnings season, you know, and also there is the Fed meeting that happened yesterday. So I was like, I just— there are moments where I'm like, I don't get paid enough risk, like the amount of risk units that I need to cash in for the potential upside. I just felt like in the conversation with Pranav, and yeah, he, um, him and I chat a lot during these moments, it was just like, it just didn't feel like you were gonna get paid enough to be long versus the current scenario where the market— you talk about all the dips being bought, but the market was looking for reasons to sell. It was already very jittery around the Mag 7, accounting for all this it was just looking for reasons to sell. And when I, when I, I feel that, I was like, yeah, I, I think there's just more downside here. Um, it sort of might contradict a lot of what I've been, you know, my style of investing, but I do think that I, I take this tactical wallet as, as an opportunity to develop my more agile, like, investing trader. Um, so it's, it's like, um, a thing that I'm trying to develop, right? Reacting to markets, interpreting that. So anyways, I sold that into stables. Over the weekend and then and rebought yesterday after the Fed meeting. There was a lot of really good comments. That's one thing I did on the on this particular one. Again, it's like 5% of my book. It's all very liquid. It was meme coin gains that I, you know, was largely lucky. So I, you know, I wanted to lock in that profit. And my thinking there was, hey, could we go down 20%, 30%? Maybe likely could, could be possible. Nvidia definitely did. I was actually surprised that crypto didn't go down as much. Like, like Nvidia going down.

Well, I mean, some things did, like Bitcoin, like, you know, Bitcoin has performed remarkably well. Something like Fartcoin is down from $2.50 to— I think it fell to sub-a-dollar, fell to $0.80, right? So, yeah, you know, like Virtuells is down 70% or something like that in 30 days.

Like, in 30 days, right? I mean, it had been coming down. It had been a lot of it dragged down, like it sucked the air no pun intended here, or Filecoin. The Trump meme coin phenomenon just sucked the air out of most other alts. But so the alt market was already in a downward trajectory and losing momentum. And so I felt that this was just another reason to get out. And so I rotated to Solana. So that was a good trade, meme to Solana during the whole Trump launch. And then I just felt that I want to stay bullish now.

But then you said, but then you said you rebought. So did you rebuy just Solana or you bought back the meme coin?

My entire position. I rebought it slightly lower. Like, I sold at like $234 or something, $236. It dipped, uh, and then I bought it like at $232.

Did you buy anything other than Solana?

No, no, for that wallet. Like, again, the long-term stuff I'm not, not touching. Um, this was just fun to just adjust. But I will say I did other stuff in my, in my traditional portfolio outside of crypto Um, I— people may remember, like, I've been doing this trade for now the better part of like 4 years.

Phoenix—

they call them Phoenix notes or reverse autocall converts. And so I— the idea is basically this. So like, you like a particular company, say you like Nvidia or you like Coinbase, um, and but you know, in your portfolio you want to optimize for income generation. Because you're so exposed to crypto, you want to have like equity-like participation. You like a particular name like NVIDIA, but you feel that it might be expensive. And I felt that the market was like— hasn't been cheap for a long time. So what do I do? Basically, these products are ones where you set a barrier, which is like, say NVIDIA is trading at $100 right now and you say, I'm going to— it's a derivative and you basically say, I'm betting that Nvidia doesn't go down below 30% to the current spot, or you basically can choose 20%, 10%, 30%, up to 40%, 50% if you wanted to. And what you're toggling is downside protection with the amount of coupon that you're being paid on an annual basis. To give you some perspective, like for instance Nvidia— maybe not Nvidia, but like, uh, if you wanted to do like a Tesla, for instance, reverse auto call convert Phoenix Note, it's the same thing. You say, I like Tesla, I love it at down 20%. I would actually be a buyer down 20% or 30%, uh, but I'm not ready to buy it now. I want to get income generation in my portfolio. So you'll structure this product with a 30% barrier. So Tesla— so long as Tesla is not down more than 30%, you're going to be paid a coupon, and that coupon could be anywhere from 10%, 15%, 20%, 30%. Depends on the volatility of the underlying. And so that's pretty good from my standpoint, which is if you look at the historical returns of the market it's like 8% a year, 12%, 14% a year, right? And in a good year. But we've been on this run for quite a bit of time. So I'm thinking, well, if I can, if I can clock in on above, above like historical average return of the market, which is 10, 12%, more so than like credit funds. And I know it's a different risk profile, but nonetheless, I'm like, I'm getting paid a really interesting coupon, which is reviewed quarterly. At best, Tesla doesn't go down more than 30%. I get paid this pretty high coupon, uh, and I kind of outperform the S&P, and I'm not taking as much risk. Worst case scenario is Tesla runs up a lot, so you miss out on the upside participation, but you— and you get called. So you might get paid a coupon one quarter or so, but you get called, um, and so you lose on the upside participation. And the wor— all the other scenarios, Tesla goes down more than 30% and then you have to buy that at whatever the prevailing price is. And so that gets reviewed quarterly.

Quarterly.

I structure it bespoke, which is I reviewed semi-annually, so at least I lock in half of that coupon if it stays within that range. But there is callability risk and there is just a missed upside participation. Like with Coinbase, I've done a lot of these, you know, arguably it was better to go long when it was $12, $13. Like I was, I was doing all of these in size. I'm like, I'm getting paid 40% coupon. Like, I'll do this any day.

But you missed out on Coinbase going from $50 to $250.

But I missed out on Coinbase running from $50 to $200, like a 4x. But, you know, it's just a matter of like, in this part of my portfolio, I'm not optimizing for the 5x, 10x. I'm optimizing for cash generation, you know what I mean? And so what I did over this week was I bought a lot of these, and I like to buy these when vol spikes quite a bit. Because then these products pay a higher coupon. So I bought like a basket of like Tesla and Meta— sorry, it was Tesla and Coin. Why? Because I think this administration, like, you have—

so like, say what you did with Coin. So Coin's trading at what, and how much do you get paid on it? So Coin trading at $291.

Coin was $291, and Tesla was like at $336 or something like that.

Wow, CoinDrop has done 25% this year. Yeah, so, all right, 25% in a month and a half.

I'll tell you, um, what the, uh, and coupon ended up being. So I did a CoinTesla Phoenix note reverse hard call convert. I paid— I was paid a 35% coupon on that.

30 at what price?

Whatever the spot was, uh, closed yesterday. Um, and it has some features. It's like 30% barrier, so down 30%, um, and conditional memory coupon, which means that it— if, if it gets reviewed in 6 months and it dips below 30 but then it goes above but it stays within that range, I catch up. Like, the entire coupon is paid. So I like to have this conditional because if markets are more volatile, I want the conditional feature, which— or memory coupon, which like has this phenomenon where like it catches up. If at the end it goes up within the range, you get paid the full coupon. So I did a basket, not just one name, of Tesla and Coinbase, and I was paid a 35% coupon. And then I did a basket of 3 names, Meta, Tesla, and Google, and I was paid a 22% coupon. Well, so like it's above what I think the S&P will give you. Now I think last year, what was the return of the NASDAQ like? It was like 20%, 18%. Yeah. Um, could I have done better just sitting and chilling? Maybe, but I'm not— the income generation is like pretty compelling in my mind.

Yeah. Um, anything else on this topic?

Uh, just a reminder, like None of this is financial advice, but I like to do it for, for the reason that I think you and I have talked about extensively here a number of times, which is there are parts of your— it's good to like compartmentalize parts of your portfolio and have income generation strategies, have things that are, you know, less correlated to crypto and will give you peace of mind. And peace of mind might be a wrong word, but just will give you the ability to be aggressive, tactical when, you know, the market's kind of puking, freaking out kind of thing. So yeah, not on my standpoint. I think maybe looking ahead, I don't know if you want to add like what you did, or it sounds like what you did was you—

I did the retail move. I mean, I did sell, I moved out of meme coins and I just bought Solana. I moved nothing into spot because I mean into stables because why would you know Why would you ever do that? So I sold a bunch of meme coins, moved it into SOL, and then when Nvidia dropped 17%, I just bought the dip at the open.

Bought the dip. Yeah, you're probably— I'd be very intrigued to see the comments here of listeners like, oh, Jan was a better investor. Probably, ladies and gentlemen.

I can assure you that is not the— I think the numbers don't lie. I can assure you that's not the case. I just don't— I'm just not sophisticated enough for that.

Earlier in today's episode, I mentioned Ledger's latest products, the Stax and Flex, and I wanted to take a few seconds to tell you more about them. These wallets feature the world's first secure touchscreens, which don't just simplify your digital transactions but also keep them extremely safe. These wallets are virtually impenetrable to online attacks because they store your private key in a secure offline chip. You can manage more than 5,500 tokens and install up to 100 apps. Choose the most trusted name in hardware wallets and take control of your digital security today at Ledger.com or hit the link in the description of today's episode to check out their latest wallets.

Should we talk about Powell's statements? There was a couple that I really liked that I just want to emphasize.

Yeah, I've got a statement that he said that I think you won't say that I think was my favorite statement.

But should I say what I'm trying to say? And if I say it like I get—

you get some points, I'll wink at you or something. Yeah.

Okay. All right. Blink twice. He said Powell stated, quote, banks are perfectly able to serve crypto.

I thought you were going to say some macro take or something.

No, no.

He said banks are perfectly able to serve crypto customers, says Fed is not against crypto innovation.

Yeah, let me read that expanded. Let me read that again because I actually— it's very important in context of if you guys remember last week, SAB 121 got rescinded, which means that banks can now custody crypto. And so Jerome Powell said banks are perfectly able to serve crypto customers as long as they understand and can manage the risks. We're not against innovation and we certainly don't want to take actions that would cause banks to terminate customers who are perfectly legal just because of excess risk aversion, maybe related to regulation and supervision. And I think Nick Carter, who's obviously been either leading Operation Chokepoint 2.0 or created that term— I kind of forget at this point— he said this is particularly notable because to Nick's understanding, the Fed was the nexus of Operation Chokepoint 2.0. So yeah, very important statement by Powell. Especially in context of SAB 121 getting rescinded.

Absolutely.

Yeah.

He also, he also talked a little bit about like regulation, crypto regulation, saying like we're kind of ready to move in this direction and chapter. So that was very, you know, that was very interesting too. He said, and I'm quoting Walter Bloomberg, all caps, would be helpful to have greater regulatory apparatus around crypto. So yeah, yeah, it was good. Yeah, very positive. Makes you wonder, right? Like technically the Fed should be independent from like an independent unit, but like you have a different administration.

I think we're— I think we're long gone from that.

Different tune overnight. Actually, speaking of this, I saw a very, very strong take from Gemini saying that they will not hire anyone from MIT now that Gary Gensler is back there again and I don't know what you feel about that. But that was a bold statement.

I think you're just seeing this age. I think founders feel very empowered right now, after 4 years of not feeling empowered, and 4 years of kind of operating on their back heels. And I think we have entered into an era where founders feel incredibly empowered, and maybe like maybe a little too excited to just like, I don't know, take down anybody. But I mean, I, I love it. Like, let the free market go. And if you say that you— I mean, Gensler terrorized these companies and specifically really came after Gemini pretty hard. And so, yeah, I mean, I actually love the move.

Yeah, absolutely.

Anything else by Powell? I have two other topics I want to run by you. Go for it. One is the— so last week we talked about the Vine founder launched a token called Vine.

Oh, yeah.

And I don't think he was actually building anything. I think he was just like, let me launch a token. The Venmo founder launched— actually launched an app this week called Jelly. Jelly is basically— it's like FaceTime, but you can record conversations and share clips on social. So, you know, I could FaceTime you and we could talk about like what we're, you know, Deepseek or here's a better example. You could have called Casey or it sounds like you did call Casey and you're talking about Deepseek. You could have recorded the FaceTime and then, you know, instant edited that and then shared the clip on social. He launched a token to go along with the app and it's called the Jelly Jelly token. So it's the Jelly app, Jelly Jelly token, and you need the token to actually access the app right now. And I think this is just exactly what I was talking about the last couple of weeks where we're moving from tokens as a technology that's used to secure a network to tokens to drive interest and tokens for fundraising. Right. So the token's trading at over $100 million. I don't know exactly the price, $100 or $200 million. So, A, that's a big capital raise. But, but B, it's driven a lot of interest in his app that you normally wouldn't have had. I think if the— I think if he had initially launched this thing, you know, you try to get covered by TechCrunch or whatever it is. And yeah, I think this is just— you're starting to see tech founders, like Web2 tech founders using crypto as their go-to-market strategy. So I think it's very interesting. Actually, I loved— I loved Raj had— Raj, where is— where is— let me pull up Raj's Twitter because he had a meme that summarized it perfectly.

Yeah, like internet capital markets.

Now, this is— this perfectly encapsulates what I think is happening here, which is dog coins lead to cat coins lead to celebrity coins lead to AI agent coins lead to Trump coin lead to actual companies launching tokens. Yeah. And which leads me into the second conversation, which is I can't share any names here, but had a very, very, very interesting conversation with the founder of a private equity firm, one of the better middle market private equity firms in New York. These— can you hear the heater just blow? Can you hear this?

I can. It's like the tinkering.

The tinkering.

All right. Well, sorry, folks, for the—

every time I hear that, I'm like, are those rats running around the—

just Welcome to New York, baby. Uh, so yeah, it's just the radiator just blasting, turning on or something.

Um, are you— are you— how do you— how do you know it's not rats?

It's not rats. These things is the— it's the New York radiator.

I mean, New York rats, New York radiator.

Dude, tell me you never lived in New York. That— tell me you never lived in New York.

I, I've lived in New York, sir. I've seen the rats too.

Eric Adams took care of the rats. He hired the rat exterminator. He took about the rats but not everything He took care of that. Yeah, exactly. So anyways, this private equity firm, you know, been around for, I don't know, 30 years or something like very successful private equity firm. I got— I've gotten to know the founder of this firm pretty well over the years and I get on and he had emailed me saying, I want to learn more about meme coins. And so we get on and he knows very little about crypto, but, you know, we just— by us being close, he's kind of follows the industry and he said, look, like, We are kind of getting smoked by some of the BlackRocks and KKR and Apollos of the world. They all have these products that are just eating up like all the middle market because they're so big and they can offer so many products. I need to figure out a way to differentiate. Should we launch tokens? Should we be launching tokens for our companies? And it reminded me honestly, a lot of Inversion actually, and like what you've been thinking about, which is, hey, there's all this capital on-chain and liquidity on-chain. There's tech improvements that you get by moving things on-chain. And he said, look, I really know nothing about this stuff, but it seems like there's new technology that people clearly like, right? Like BlackRock's launching Bitcoin ETF and maybe more stuff. And you get that combined with like a huge valuation mismatch, right? You have a data company in tradfi, it might be valued at $100 million. You put a data company on-chain with a token, it's now worth $1 billion. And He said, look, we've got all these companies that probably— I don't know why they need a token, but should we be launching tokens? And this is all— it's all gearing up towards every company is eventually going to get a token. I don't know if it's going to happen in 2 years. I don't know if it's happening in 5 years. I don't know if it's 20 years. But you'll see— you'll eventually see the Nasdaq and New York Stock Exchange be able to trade tokens on those things. Or if they don't, you'll see crypto-native exchanges like the Coinbases of the world overtake these guys. And you'll see crypto companies launch tokens like they are today. Like the next phase of this is crypto companies that don't have a blockchain launching tokens like Kaito. I keep talking about this. I think you'll see a lot of those kind of folks do it. And then you'll see the Netflixes and the, you know, Ubers and the maybe the private companies who aren't big enough to get public, but they need liquidity. They'll launch tokens. And it's just like, it's just crazy seeing it in real time. So.

Yeah, everyone I've talked to about my vision and the strategy— and I— it was an iterative process, if you remember. Everyone has been very receptive, and they understand the mission, the vision, and the structure. And that the structure piece was something that took me a little bit of time to iterate on. But as soon as I repackaged this from a traditional private equity fund to a permanent capital vehicle on-chain, it was like the reception to that structure was far greater. And some people had some concerns on the margin, but for the most part it was that. And from my standpoint, it was really informed by looking and talking to most of the foundations and understanding that their I feel that there is far more opportunity to be aggressive as an industry to acquire, acquire users in a more tactical way and to solve the biggest problem that crypto has, which is go-to-market. And, and I think that's— there's a big, big opportunity from both sides. You're seeing traditional companies trying to— like the tokenization of things and real-world assets. Like, why would BlackRock, KKR, be interested in, in, in that? Well, because they can tap into these massive pools of liquidity. Like, you have global internet capital markets. That's profound. Like the amount of liquidity moving money at the speed of light, 24/7, 365 in real time, like settling has huge cost efficiencies and allows you to monetize more like the amount of velocity of money, the ability to monetize that user in many ways, like trading fees, all that stuff. Like, yeah, it's happening. And by the way, people are now— the nice thing about all of this is that perhaps people were thinking about and doing stuff internally. Now they feel more open to— they feel more comfortable being open about it. And that has, you know, like Vlad, there's only been a few folks that have been in like talking about this before, kind of this regulatory like 180, like Vlad from Robinhood. We've had that clip that I always think about, which is comparing his, his business of traditional brokerage, right? And he's like, the crypto side of my business is an order of magnitude cheaper, more efficient than the— and lucrative than the normal one. Yeah. I'm like, by the way, you saw Kraken, like, like there's a couple of like very tectonic shifts. Like you have the—

we've all—

we've talked about this at length, like Coinbase, Krakens of the world, no reason why they can't become large banks and ex-JPMorgans or Binance, right? We have Richard. I asked him that. He's like, I think he just gave the politically correct answer, which is like, we're not looking to be anything other than the best exchange. I'm like, I think in the internal strategy meetings they got, they got to believe like if they're going into markets like Indonesia or whatever, they're like, you could serve that customer things much more so than crypto, right? Yeah. So yeah, I think there's going to maybe as an interesting exercise is like, who do you think is the largest company? Not that announces like buying Bitcoin, I think the more interesting question is, what is the largest— what will be the largest company or the biggest splash of a company doing something on-chain other than Bitcoin?

You know what I mean? I mean, I mean, the, the very obvious one is Stripe, right? I mean, Stripe is going to move all of their merchants onto, onto stablecoins. And, uh, I don't know, you're, you're frozen again, so I can't actually see if you, you can see this or not, but I'm going to share this, uh let me share this thing that I posted on Twitter this week, which is I went to pay a stablecoin or I went to— I went to pay a Stripe invoice this week. And, you know, usually you go to pay your Stripe invoice and it was a bank transfer or a US bank account. It's always bank transfer, US bank account. And now it says crypto stablecoin. This is the first option ahead of bank accounts, right? So it says choose how you'd like to pay. It's usually just these two options, right? I've paid hundreds of Stripe invoices in my life. Bank transfer, US bank account. Then now it said crypto. So I click crypto and I go to pay and it says USDC on Ethereum. So I click USDC on Ethereum and it says select a currency, right? USDC on Ethereum, USDC on Polygon, USDC on Solana, USDP, which is Paxos Ethereum, USDP on Solana. So I mean, this is a boring answer, but like the obvious one is just like, I think everybody is still sleeping on how much Stripe is going to do in crypto this year.

Absolutely. Yeah, it will go down as, uh, it might take a few years, but it's, it's hugely transformational. Yeah. And we keep— folks should go listen to that episode that we recorded with Zach from Bridge. The acquisition hasn't closed, is that right?

Yeah, but I heard it's going to close.

Yeah, I'm pretty sure the regulatory, like, the approval process will happen after the election. It'll probably pass. But no, huge, huge. I would agree with you. Hugely beneficial.

We're kicking off DAS on the institutional track. We put Zach first.

Oh, whoa.

It's Zach, Rob Haddock from Dragonfly, and then Guy from Athena. Like, we— and we do a lot.

I love the title of the presentation.

Financial room temperature superconductors. Yes, exactly.

That is a quote from the Carlson Brothers when they announced the acquisition.

Exactly. Exactly. So that's great. Anyways, good chat. Anything else? Anything else you're thinking about or that you want to— you want to cover here? Oh, Kraken.

I mean, Kraken Pay was pretty interesting that they launched recently.

What did they launch?

Kraken Pay, which allowed you to— only if you're a Kraken customer, I think it allowed like instant payment across like their user base, which is pretty cool. Feature. I don't know if you could pull it up and show people, but Kraken's been doing some really good stuff. I would be— that's one of those companies I feel like it's going to go public this year, you know, soon. I think they're, they're doing a lot of interesting stuff.

Kraken Pay. I didn't see this. That's cool.

Yeah, it's pretty cool. Yeah, it's only for Kraken. So send any asset anytime to anyone anywhere, um, but, but it's only Kraken users, I believe.

Oh, that's cool.

Yeah, yeah.

If you have a Twitter account over 20,000 followers, you're saving your crack tag for you for a limited time. Oh, so they basically just built Venmo into Kraken. Yeah, that's a cool product. Shout out to, uh, Kraken folks and Lou and the whole team over there.

Yeah, yeah, pretty cool stuff.

That's cool.

Uh, no, that's really it. I mean, I, I think, uh, it's an important week, uh, looking ahead, uh, on just earnings. Just of companies. I'm going to be paying a lot of attention to that. And, um, yeah, but, uh, otherwise business as usual. Well, nothing else from my standpoint.

Cool. All right, folks, appreciate you guys listening. Have a good rest of the Friday, good Saturday, good Sunday, whenever you're listening to this, and we'll see you next week.

Hey everyone, just wanted to give a quick thanks to today's title sponsor, GeoNet. By setting up a base station, you can get involved with this groundbreaking web3 ecosystem. Learn more by heading over to geodnet.com or just by hitting the link in the description of today's episode.