Ledger Compromised, OpenAI Spy, & the Death of Moats

Transcript

Can you explain to me as a a bar owner what happened? I was like, you buy liquor from a distributor, right? When you buy a foreign liquor, you can't really buy it direct from the manufacturer because like US rules don't let you. So imagine North Korea bought the liquor distributor and then just like poisoned every bottle you got. That's like basically this what this [laughter] attack is. >> Not a dividend. >> It's a tale of two. >> Now your losses are on someone else's balance sheet. >> Generally speaking, airdrops are kind of pointless anyways. um unnamed trading firms who are very involved. >> Eth is the ultimate >> DeFi protocols are the antidote to this problem.

>> Hello everybody. Welcome to the chopping block. Every couple weeks the four of us get together and give the industry insiders perspective on the crypto topics of the day. Quick intros. First you got Tom the DeFi maven and master of memes. >> Hello everyone. >> Next we got Tarun the gigab brain and grand puba at gauntlet. >> Yo. Thanks. We got Robert, crypto connoisseur and Zar of Superstate. >> Smooth evening to you. >> And I am See, the head hype man at Dragonfly. We are early stage investors in crypto, but I want to caveat that nothing we say here is investment advice, legal advice, or even life advice. ly chopping blocks. For more disclosures, boys, we are at well, it's a late night recording. I think we're all a little bit tired, but we'll uh we'll try to get

you the sweet, sweet alpha that you're looking for. It is just about the wrap-up of Token 2049, which is the biggest crypto conference in the world at this point. Over 25,000 attendees in Singapore from over 160 countries. It was a little bit of a different tone than previous token 249s. I think the the I think it was Arthur Hayes who talked about this when we were doing our last episode on stage at token is that the number one index of the quality of token is marquee which is the biggest nightclub in Singapore which is you know at the bottom of the marina bay sands and normally it's booked out every single night by some or other shitcoin.

Uh this year Marquee was a normal nightclub during all of token which is a sign of the times and then there were a lot more suits and institutional talking points and you know you had uh Jenny Johnson from Frankle Templeton and you had Adena Freriedman from NASDAQ being on stage making announcements to what was once a very crypto DJ crowd. So a very different event. I don't know Tom and Tun, you guys are out there. What What were your takeaways from token week?

>> I mean, I feel like this is what the industry has kind of become over the past two years and this feel like the year where this was people like realized it like last year felt very LARY in some ways. Not even just the, you know, this conference, but just the industry overall where people like didn't want to open their eyes to the fact that this was happening. And then sometimes you meet these people who are like the last Japanese soldier trying to tell you about their new layer 1 blockchain. It's like what the [ __ ] are you talking about? So I don't know this felt if anything more like like a realization of the thing that was already happening. >> Yeah, last year token was right before 1010. So like last year's token still felt like a party. This token does not feel it didn't feel like a party. I mean

there were obviously some events, but they were so low-key compared to what I think of token events as where it's like, "Okay, yeah, come meet us at a bar or like we'll we'll do dinner and like someone will be talking the entire dinner with a mic in their face like that." It was very different character of events than what previous years felt like. Trun, what was your experience? >> I think succinctly the number one way I would describe it is I saw zero lambos. I think that was sort of my but the you know the no no lambos feeling was sort of very apparent. I think the other thing that >> is that good or is that bad? How do you how do you feel about the lack of lambos? I personally prefer it, but I

would say that also means there's not this type of like general abilance in the air, you know, like the animal spirit stuff. There's instead the like bean counter vibe, you know, like everyone's like [laughter] >> bean counter vibe. >> I don't know. I just the one place that's not >> not bean counter. >> I'm just saying does not have bean counter vibes, [laughter] >> dude. This is like Dreamforce the country, you know? >> No, no, no. I mean I mean token. No, no, no, no. I mean, compare token to like consensus or compare it to like DS or one of these.

>> But do you remember how consensus got dude? Do you remember consensus 2017, 2018 was like what token is now and then it became institutional and never came back, >> right? So I think the thing Alex >> who runs token has done a good job of is like making sure there's still some amount of like dgen bosery around. But even with the efforts, I think it was like a little bit sad kind of in that sense. Like there was like there was just like a lot of people trying to tell me about their institutional product. And I'm not saying it's a bad thing. I'm not I'm just saying like it's like the pitches are boring too, right? It's like there was a time when like crypto was

considered frontier tech and people are like who are pitching you like I'm going to make a robot that like catches fish for you at the same in the same sentence they would uh do you remember the shrimp farming ICOs of 2017 >> that was just a 4chan meme turn that wasn't >> no no no no there was an L2 that that was a shrimp farming robotic shrimp farming company that pivoted yes >> what yes >> what was the name of it >> I'm trying to remember the name the guy who founded its name was Matthew Campbell. That's like all I I remember cuz I like I met this guy and he was like doing the shrimp farming thing and I was like what the [ __ ] is happening?

>> Georgios from Paradigm uh was a contractor there at that time. That was like before he worked at Paradigm >> which is also how I met this is also kind of like how how how I ran into this like >> farming lore. >> Wow. So, so anyway, the main point is like there was a sense in which there was a front there's a frontierness of crypto, right? Like you're like there's a new technology, new stuff >> and realistically outside of the Zcash conversations and like the quantum conversations. I kind of think there's nothing left in crypto where people are like talking about pushing the frontier instead building products and monetizing which is fine, right? Like and good. It's just like it doesn't have the like you know

whereas like the the interesting thing other thing about token is I feel like there were a ton of other conferences at the same time arguably like Milkin and all these other stuff and arguably maybe that there someone must have had the DGEN kind of thing it just wasn't crypto you know that's kind of that's sort of my impression also it was kind of hard like to be fair the air quality in Singapore because of the fires was so bad I feel like people were not in the like going and partying. I mean, I felt like I was sick the whole time I was there. As [snorts] soon as I got back, my cough magically disappeared.

>> Yeah. Yeah. There's like very [clears throat] in Singapore. >> Yeah. The the there's like some fires in Indonesia. They're kind of they're kind of wrecking the whole region over the last few weeks. So, it is >> So, there was like a little bit of a haze that like I I think could have also possibly explained the kind of like a little more. >> Yeah. But, >> but I mean, usually people book out these things in advance. So, I I think it was overdetermined. The haze I I don't think helped the the vibe, but I feel like it was going to land there anyway just because the character of the industry has really shifted.

>> Yeah. I mean, it's also funny that Arthur is the one who brought up the the the nightclub thing because I'm just kind of like, isn't he usually the one that people like, oh, threw the big party at token? [laughter] >> He did. He did. He threw a huge party. >> No, no, I know. I know. It's like it's like funny cuz it's he he's the one carrying it on his back. >> DJ I do think the nightclub index is definitely a good alpha for understanding how the industry is is uh migrating. >> Yeah. It's like it's like normal people and inflation have the big Mac index, but crypto people have the nightclub index.

>> The nightclub index. >> The Bitcoin Miami index. >> Bitcoin. Well, you know, there was also I think it was consensus in Miami, right, where there was that scandal about some official event happening at a strip club. >> Oh, yeah. >> You guys remember this? Yeah, it like when it went like >> not long ago either, right? >> No, that was this year. That was this year. >> Yeah, I thought it was recent. >> So, that there may be again like the industry just shifting in norm I mean maybe in Asia actually that they care less about that. But at least I think like it's again a sign of the times is that in prior years >> I do want to still give a shout out to Alex who you know arguably he's like kept this conference alive through so many bare markets ups

and downs and it's still just growing in attendance. So, you know, you you like something like consensus I feel like had very high variance and somehow these guys have been really good at like making sure the content changed to keep you know so like real props to them for for doing it and also doing like three conferences now. >> No, it's a top-notch event. >> I feel like the production quality has also gone up like every time. So, >> yeah. Yeah. And um tough to do. Well, so speaking of crypto no longer being frontier, one place where we are frontier is in the hacks. So we have some crazy news of hacks that took place over the or were discovered I should say

over the last couple weeks specifically around hardware wallets. So Ledger is the largest hardware wallet. Their their last valuation was over a billion dollars. So they're very successful hardware manufacturer and and vendor. Uh however, they have licensed resellers or authorized resellers. So these are storefronts or online retailers that are authorized to resell Ledger and they have some official relationship with Ledger the company. One of these resellers was a company called CryptoBillis uh which was based in Malaysia and CryptoBillis.

There was uh reports that uh I think on X and Reddit that some people were starting to get their Ledger cold wallets getting hacked. Normally there's always kind of news of oh some people are saying that they're getting hacked and nobody you know it's hard to really say okay there's some systematic thing going on but there was enough money starting to get stolen on the order of about $80 million worth of crypto that was stolen from ledgers and that it eventually by triangulating where all the hacks were coming from they realized that these were users who had bought their ledgers from cryptobillis and so it turned out that after taking apart the ledgers that were being sold by cryptobillis It was confirmed that these ledgers inside of them had a hardware implanted

device and this device sat underneath the screen, had an ESIM, so it was able to phone home and basically allowed you to read the data that was shown on the screen and transmitted over LTE using the ESIM, presumably sending it back home so that people uh who were in some command and control center could see the secrets that were otherwise being transmitted over the ledger. So to be clear, hardware devices have a trusted enclave, and this wasn't a compromise of the trusted enclave. The the attackers did not know how to do that, but they were able to nevertheless read the entropy that was on the screen, which when you're generating your seed phrase is going to show your seed phrase on the screen when you're initializing your device. So that's how this attack was

taking place. Absolutely crazy. There was no indication, you know, they were wrapped in the original case, you know, with the plastic over it. And when you open the when you when you see the device, the device looks normal. It's only if you open it that you see that there is this tracking device inside of the the uh hardware. So it was determined apparently that this was also happening at least one vendor in the UK um or in the somewhere in the EU sorry I should say. I I don't know if sorry I think it was maybe in Austria or something. And so it's unclear what is the extent of these attacks.

Interestingly, Mark Carpellis, who of course crypto og in various ways due to Mount Gaus, uh he's been on this for a while covering the many ways in which if if you buy your ledgers not from ledger directly but from resellers or even on uh like eBay or something or one of these secondhand markets, there's a whole industry of ledgers that have tracking devices in them, ledgers that have non-authentic hardware that have no secure enclave. So there there's like a whole universe of these that we're just now learning about seeing this attack isolated to this particular vendor.

>> Don't forget people stopped buying from Ledger.com because they leaked all people's information. [snorts] >> Yes. >> So there's like kind of this like irony. >> Correct. That is ironic. And >> so this happened with Ledger [clears throat] and with Treasure Cole Card got hacked due to an entropy bug. So it's been a really shitty no good absolutely terrible year for uh hardware wallets. the the story I going back to CryptoBillis. Apparently the story is that the CryptoBillis founders, they got tagged because they were like, "Yo, what the hell? Why why is your company selling these?" And uh the Cryptobillis founders reported that they had sold their company a few months ago to a Chinese buyer and they were under a confidentiality agreement not to

disclose this until October 20th, which of course is >> timing. >> What timing indeed. Um, so, >> oh, I didn't realize it hadn't closed the transaction or sale. >> No, it didn't. >> The transaction had closed. They were just under confidentiality. They were under confidentiality to not disclose that the sale had taken place. So, crazy story. >> It looks like there's going to be just a total either re-underwriting. So, Treasure has already cut off all of their authorized resellers, has basically said, "We cannot sell to anybody else anymore or allow any distribution of our products. You have to go directly to Treasure." I'm assuming that Ledger will probably do the same. Um, but it's a huge pockmark on the hardware wallet industry

reactions. I think it's indicative of how risky supply chains are for sensitive keys and anything right you know a few months ago we were talking about software npm supply chain risks you know people you know developers you know relying on a package that was you know a a ticking time bomb that was a blah blah blah blah blah like this is the physical manifestation of that I think that all types of supply chain attacks are going to get more sophisticated over time. You know, this one is really spooky in that somebody has essentially prototyped buying a

vendor, you know, with an existing business and trust and deal flow in order to execute a supply chain attack. You know, in the same way that people historically have bought browser extensions with installed user bases and then flipped them into crypto, you know, attacks. You know, it just goes to show the level of creativity out there. You know, I could see someone saying, "Oh, you know, just buy a vendor. What else is for sale, right? Going into any industry, going into the AI hardware space, like going into anything, right? Like there's like so many supply chains out there that I think, you know, all of us should just constantly, you know, call into question, you know, how do you protect a supply chain?" This is

a glaring example of it. Back in the day, the how to steal crypto through supply chain attacks were a lot simpler. You know, I remember way back in the day, you know, people would try to make a paper wallet for their Bitcoin. They'd go to a website and like how could you trust the paper wallet generator? You just sort of knew. >> I mean, this used to actually happen with like minor like you bought an old Bitmain, used Bitmain, and people would like mod it to basically send the reward to themselves. So, like I I think like this has been in crypto since the beginning. It's just like kind of been more niche and this is kind of way more like the distribution >> hitting Ledger is hardcore.

>> Yeah. >> In a way it's almost surprising that it took so long cuz I've never heard of anything like this for a big vendor like Ledger. >> I mean this is probably an inspiration to like Lazarus. >> Yeah. [laughter] my my uh my friend who owns a bar was like, "Hey, can you like explain to me what this attack is because like I own three ledgers that I've never used, but I only buy at the top of the XRP." >> The XRP XRP bar. Okay. Yes. >> But did he buy it from Ledger or buy it from a shady seller? >> Amazon. So, I have >> Shady.

>> Well, don't do that. Yeah. Yeah. Yeah. I don't know. He hasn't opened it. He just was like He was like, "Can Explain to me as a a bar owner what happened." And I was like, you buy liquor from a distributor, right? When you buy a foreign liquor, you can't really buy it direct from the manufacturer because like US rules don't let you. >> So imagine North Korea bought the liquor distributor and then just like poisoned every bottle you got. That's like basically this what this attack is. >> Wow. >> Which which worked as an explanation. >> That that's a very good explanation.

Thank you. [laughter] >> Yeah, I think it's actually quite clever too. It's like basically like a inner side channel, right? It's like of of course you know we can't actually break the thing that is hardened. So like and and of course like also like devices that have built-in LTS have gotten so cheap. It's just like the whole thing is actually quite quite clever. And it's also to your point I feel like the rash of ledger issues the past few years have been people buying ledgers off of places like Amazon and of course Amazon has this sort of constant third party fraud quality problem that they're combating. But so the ledger sort of solution is well we have an authorized reseller program but now that is of course also poison. So yeah, I don't know. It's um not not a good look. It feels like they

kind of just funneled people into like this this thing. >> I mean, the answer of course is okay, buy direct. Don't have some third party sitting in between you and a hardware manufacturer, especially when it's like extremely critical hardware. And now that this attack has happened, I mean, it has woken people up. So nobody will probably make this mistake again who's custoing any serious amount of money. Well, hopefully. I mean, I don't know. People have short memories, so maybe not. But certainly on the professional side, you know, I have to imagine that everybody I mean, you know, if you're an exchange or something and you bought your, you know, your your hardware devices from CryptoBillis, [laughter] like you're probably beyond saving. But for anybody who's, you know, like a fund or or anybody who's doing some kind of

like serious heavy duty self-custody, yeah, go audit your supply chain and make sure that you're buying directly from the manufacturers. PSA like especially if you're in Southeast Asia, double, triple, quadruple check or maybe just just move your funds somewhere else just to be safe even if you're not sure where your uh hardware supply chain came from. So, one of the things I saw people proposing as a solve is just to make the um the actual hardware casing transparent so that like because most of the ledgers are just they're made of metal so you just can't see what's inside. you have to literally crack it open and like unscrew stuff in order to come in and see, oh, is there a tracking device in here? Whereas if it's transparent, then you could just, you

know, show it to your AI and be like, hey, is there anything in here? Should I be does this match what it's supposed to look like based on the specs? So, I see a look of skepticism. >> Yeah, I don't feel a little like silly, I guess. Like I don't know. >> Is it? I mean that's not that's I mean >> yeah I think the way you know other sort of critical hardware manufacturers get around this is like they have the case be sort of self-sealing so if you open it like it breaks some key part of electronics and so you can't you know open it but I mean Ledger also just has this inherent honeypot problem right where it's like you know if you tried to infiltrate like the you know iPhone supply chain maybe you could but then how would you monetize it kind of

unclear how valuable it is whatever this is literally like this thing people are going to millions of dollars in tens of millions of dollars in. So if you can figure it out, there's very clearly a reward. And um I think that's I've also seen people proposing this alternative just like buy like an airgapped iPhone and use that for your hardware wallet and like obviously much more expensive but you know not a bad idea >> and annoying. >> Yeah. >> Yeah. Yeah. I mean it also does put Ledger in a position of more importance going forward in that you have to verify the veracity of your or the authenticity of your device because now I think with it's it's sort of like once you see one there's probably going to be copycats especially seeing how much money was stolen you know 80 million is is uh

obviously a pretty big hall for any supply chain attack like this so >> that could buy quite a few vendors that's why I think this is so risky >> yeah yeah that's right >> wait what if the cold card person used cold card funds to buy the vendor. >> They could have [laughter] that would be that would be like the ultimate murder mystery platform. >> I I love to ruin fanfic about cyber security. It's always >> it's it's it's RSI for scamming. We just keep [laughter] keep rolling up these bigger >> Oh, there you go. >> Wait till we get to the RSI scams. [laughter] >> Yeah. Yeah. I was really tickled by this tweet where this one guy, this Chinese guy was like, "Hey, I just dis disassembled a Ledger Nano S compared to the original PCB and it totally doesn't

match. Is this a spy thing? Like, do I have some kind of spy motherboard?" And it's like, "No, you just have a fake this is a fake ledger. You just bought some you bought some like secondhand thing and it's just that none of this. It's not spying on you. It just has literally nothing. It's a knockoff." So people people discovering that uh they're they've they've been had in various ways. >> Well, that wasn't the only people feeling like they got had. >> I feel like >> Yes. Actually, so there was a a story today that's going viral, which is that OpenAI discovered that there was uh an employee who is a Saudi national who implanted a listening device, a quote unquote

sophisticated recording device uh wrapped in toilet paper behind a TV in a conference room. He's since been apprehended by the FBI and apparently he had some kind of connection to a Saudi VC fund that also employed the son of somebody within Saudi intelligence. So unclear what the story is and obviously we're going to learn more as more information is revealed about what the intentions were behind this person, but apparently this employees partner or significant other was making searches on chatbt asking about how to effectively construct a and install a listening device. So ju just like incredibly

stupid. Uh at least that like that part seems just like wow, you know, may maybe send it to Gemini. >> Should have used. Yeah. Or like Kimmy. >> Yeah, exactly. Like surely they would not be cooperation between those two. So I feel like you'd be fine. >> I think this tells me more about the quality of the Saudi intelligence agencies than anything else. >> Yeah. >> Like there's no training for your operative. >> Yeah. it. I mean, they probably Well, I don't know. It I I It sounds quite likely that maybe he was tapped later on as opposed to like he was a spy from the beginning because he he was like a he was like an engineer or something, right?

>> Yeah. >> Yeah. He'd been there like a year. >> Yeah. He >> Yeah. Yeah. Yeah. And he like had worked at other >> I went to school in the US and stuff, so it was like not Yeah. He definitely >> I I would guess that he was tapped later in life to be like, "Hey, could do this for us because open." >> But don't you think there's like a modicum of like training >> like not like hey use your wife's account chat account to go search [laughter] >> here. Here's >> I don't know. It's like kind of weird. >> Here's the additional information I saw on Twitter, which by the way could be completely inaccurate fabricated >> right? all this happening.

>> By the time this episode comes out, this will probably have been disproven. So, take it with like five grains of salt. But one of the things I read was that he was essentially extorted by phone and threatened by what amounts to be a creepy phone call, okay? Saying that they had his family and that he had to install a listening device to free them. sources force me for this. Yeah. How would they How would anyone know this?

>> Some guy on Twitter discovered this. How would that possible? >> Yeah, that's why I'm saying take this with five frames of salt. >> Yeah. Yeah. This sounds like 4chan conspiracy back. [laughter] >> Yeah. >> This is chopping block at night. Okay. This is like chopping block after dark. >> Excellent. Yeah. Late night alpha coming from Robert. >> Let Let me find corroborating information. >> Yeah. Why don't you do that? You know why? Why don't you do that? Okay. Amazing stuff. But I think it is a sign of the times. I mean, I think both the the conference rooms within OpenAI as well as the supply chains of Ledger, it looks like there there's there's all these choke points that are really built on this assumption of a high trust society that seem to be breaking down or maybe are just like the difficulty of

doing this is going down with AI, right? like you might have been I mean I don't know that you have to be that sophisticated to figure out how to put a listening device in a conference room or to um shove you know an LTE capable sensor into a ledger. But it does feel like these things are probably going to accelerate just because it's just much easier to come up with a scheme like this and to actually execute it without having technical proficiency coming out of the gate. So I it feels like we're going to see more of this and we have to start assuming that we are in a world where these kinds of complex attacks are much easier to do and are going to be a lot more common. I mean there's there's

all this story about I think it was one of the open AAI models that was trying to commit some kind of adversarial code to a a open source repo and like was trying to socially engineer the the uh the maintainers of this open source project. And if models are doing this on their own, there's probably going to be more and more and more of this, whether it's from rogue models or or like from spy agencies or from just bad actors in general who are trying to compromise the core supply chains of the internet. We've talked about this before, but it feels like, okay, now we're starting to see it. And certainly, you know, we've never seen an attack like this with respect to Ledger. I think it was Jacob Coxin, the guy who famously quit Anthropic, who said he thought there was a very high probability that there are

CCP spies within Anthropic. It's like almost certainly because the security protocols in there are so simplistic compared to what you'd expect from what at this point are the most important economic and national security assets in America basically are the labs and they're they're not treated like that. >> Yeah. I uh I assume most large tech companies have foreign spies, but I think on on the AI lab specifically, it is also interesting. I was kind of thinking about this just like the the process and the obstac behind actually like guarding the weights of whatever like the most frontier model is. And I was kind of like digging and there is like some secrecy. There is some sort of like almost multi-IG type setup. Uh but ultimately they have to be in clear text

somewhere running on a GPU somewhere and and so it and then you they have like egress limits and stuff like that but like I'm surprised there's not been more attempts or discussion of an attempt of like somebody to try to xfill weights themselves which is like ultimately what what people want right and that's you know why would why would you even bother trying to distill which is anthropic's whole beef when you can just like get the weights themselves. Yeah, I mean I do think excfiltrating the weights of one of these models is much harder than anything else, especially given how much data you have to excfiltrate in order to get the weights out.

>> I mean, >> so I would assume that's the correct >> I'm thinking more like a physical attack, right? Like they're ultimately literally on silicon somewhere and you know there's always this sort of meme around >> many places. >> Yes. Are you going to like take an electron microscope and and go do it? It's like yeah, for a million dollars I think someone probably would go do it. I think it's more than a million dollars for what it's worth. >> Yeah, it's definitely more than [laughter] if you're working at OpenAI, you're making a million. >> No, no, I meant the cost of electron microscope, but you know. Yes. Again, it doesn't take that long to 8,000 people at OpenAI.

>> Sure. Sure. It is just surprising. Well, maybe maybe it's like that. Uh, the thing that people keep saying about these uh, agent harness breakouts is that if you find one ant in your kitchen, there's almost certainly more than one ant. The same thing is probably true here is that, okay, we found the one crappy Saudi spy, there's probably a lot more spies who are a lot better at their jobs. >> That I agree that I [laughter] >> So that's that's um, yeah, they're just base rates should tell you that, hey, if you found the really bad one, probably the really good ones are still in hiding. Okay. Well, flipping back a little bit to what's happening in cryptoland, let's talk about the new hotness in DeFi, which is paper trade.

So, Tom, I know you've been looking at this closely, so may I may ask you for some uh an alleyoop assist on explaining, but basically paper trade is you can sort of think of it as GMX on steroids. It offers a,000x leverage on what obviously paper trading. So, it's a synthetic per totally fair launch. there's no order book and every trade gets priced at the uh midpoint on hyperlquid using the hyperlquid oracle. So you're looking at the actual hyperlquid order books. Uh there's zero slippage so you can execute a trade of arbitrarily large size at the mid-market price. Um and there's no funding costs.

So it has been making a shitload of revenue. So I saw today it's making more revenue than Hyperlid, like double what Hyperl is making even after open interest has uh declined pretty significantly. It's at 32 million in 7-day revenue, top three in all of crypto. Co-founders are Jez and Blur. There was a crazy amount of farming in the very beginning, which I think is due to the emissions rate because the the way that it works is that if you lose on paper trade, you mint paper, the native token on paper trade. And so therefore, it's kind of like the downside is mitigated because the more you lose, the more paper you make. And if paper is going to be worth the next hyperlquid or something, then maybe you'll be okay.

Tom, what am I missing here? Explain any gaps in my storytelling. >> I think you got a lot of it. I mean, the big point here is, you know, it is like GMX in that you are trading on I guess this mid-market price um based on the BBO oracle from from Hyperlquid. The the maybe change is that there is no LP. There's no sort of like you direct house that you're trading. since you're trading this like sort of virtual idea of a house and claims get paid out in this queue and if hey you uh end up needing to sort of print some some paper in order to to pay people because there's debt that that's kind of part of the process but the main thing right now to your point is that like there's a cap on how how large the house can get when

you lose money you get paid paper and then the paper gets staked and so people are basically running there's a some automated strategies to try to like run delta neutral strategies get liquidated on one side earn paper and just sort of accumulate as much as possible So that that's I think part of the excitement right now, but there was also just like naturally you sort of have a lot of adversarial flow when you have this kind of setup. So like one of the more obvious attacks, right? Well, you just, you know, if you're offering me basically infinite size at mid-market price slippage, well then I'm just going to go and like, you know, take the other side of that or I'm going to move move the oracle up, which is what this one this one trader did is uh, you know, hey, I'll just go and and move the

actual BBO for whatever. I think it was like ETH or something on Hyperlquid up and then great MS getting like >> market manipulation in DeFi is so >> yeah I mean I mean I do think the the funny thing is like all these people calling these strategies delta neutral buddy your delta is your liquidation incentive and also the [ __ ] paper price [laughter] like there's not there's not free money here you are taking some risk >> so okay help help me understand all right let's say I yolo 1000x in one position and I lose the money or I make the let's say I make the money right so like obviously you're trading against the house the house the assumption of paper trade is that well you know on average retail loses money retail is trading

against you so you're happy to take the other side of all their trades and like over time you'll make money and that's kind of the loose understanding is that like you know over time as the house you should be profitable now that assumes a lot of things uh it assumes for one you're trading against retail assumes two that you're not getting manipulated and so there's all sorts of things that exchanges do to not face toxic flow in this case. There is no way to select out toxic flow, right? And so you've got the the oracle manipulators, you've also got like the actual hedge funds that could just go trade on paper trade and they have alpha and they're going to be profitable over time. So like what's the theory of the case of why this would work?

>> Well, there is a fee for winners. So winners do pay 2% of their their profit. So it's not like strictly free free and and there is also that sort of scales up with sort of the size of of of the position but the insight here is like well on a long term you know maybe this works in a short term what do you do with the claims if hey maybe the the house isn't solvent and that's sort of the idea behind this Q system versus if you look at any other system that has a sort of opportunity for defaults like a GMX there's just sort of race to the exit there's no sort of opportunity to take a uniform kind of kind of haircut and then ultimately the whole thing kind of ends up collapsing on itself. But I think you're you're right that like the

question is will you actually be able to get you know enough retail flow to sort of offset toxic flow. I mean like any sort of any any sort of exchange and uh I think that's also why they you've been advertising a lot of the leverage and and things you can get on it. But right now it's it's kind of unclear. It feels like you naturally just end up especially when the the mark price that oracle is also on chain feels like you naturally just select for you know people like the the attacker basically who I think right now is like the number like the largest holder of of paper. Hm. Wait. So, is is the house does it have positive P&L so far?

>> Yeah, I think it's up like 5 mil or 4.5 mil, something like that. >> Okay. >> Okay. Interesting. And what do you think? Where do you think this is going long term? >> Well, I think it's like an interesting experiment, right? It's sort of inspired by these like, you know, bucket shops of like the, you know, 1920s. And I think if it's successful, maybe more people will will replicate it. I I think if anything it's with an exchange if you don't get good natural flow like you just kind of end up dying or as a market maker it's like you don't don't make very much here it's like this feels like there is opportunity for this to be sort of existential for the actual product.

So I don't know if it's going to be possible to like replicate this in other you know scenarios just because it is like the outcome is going to be somewhat binary um and if you don't sort of end up end up spiraling down it's great if you don't end up spiraling down but if you do then not so good. Well, I mean as a VC that's what you want, right? Is like >> true. Yes. Which is the other interesting thing is um you I think this is one of another, you know, this recurring trend of of startups that are not taking VC money. I think this has been sort of, you know, obviously something that's come on and off throughout crypto throughout the years, but now I mean you look at like, you know, Hyperlid, Pearl, Flop, this feels like it's very right now to like not do

>> Pearl. Yes, it's a fair launch. It is a fair launch fair launch. There are VCs still but in the in the labs, not in the not in the token. Not directly in the token. So no premine. No premine >> and uh you know I'm sure they >> they had but they had to build it. Yeah. They had to build it. You know, of course they expect something. But um >> Trun, what is your take given your status as a DeFi theorist? This feels to me like it's got echoes of like Basecoin or or Basis where it's kind of like okay well if it works it's going to be so big and it's going to be enormous but uh it's kind of hard to like not face any of the various sillas and caribdises

that are flanking it. You know that's a good metaphor. Uh yeah I I don't know. It's like hard for me to imagine this thing works if paper the coin doesn't do well. Right. Like it it if I look at the house of cards and dependency diagram of like all the strategies people claim this is useful for there is one one kind of main villain or crux whatever that exists and that's the paper token and like at some level all these strategies where people like I'm willing to lose money to get the paper token how long are they going to do that right like that's and that's where I think you know like if you look at like coins that have fair launch but done well like hype

they they like built a cult pretty quickly a paper trades cult is like around losing money to earn more paper so I'm not like and Jez which is there is there is some something there but it doesn't have the same like house of all finance thing right like there's not some like long-term thing everything is like how can I make money quickly off the thousandx which I I think is it it feels like the paper token price going to be too dependent on that and then the rest of the stuff is just going to be yeah all all of the kind of things people complain are complaining about might have but we should just get jazz on >> paper paper hands if you will >> yes paper hands causing the paper trade to die

>> interesting okay well we'll we'll see if we can't uh get some folks from the paper trade world on to talk through it next time last thing that we want to cover today is supposedly the death of Moes or is the death of Moes greatly exaggerated? So Nikita Beer the former head of growth at X as well as Salana I think adviser to Salana Foundation so you know crypto fellow traveler he posted that the only moes or that technical modes are dead. The only moes left are data network effects and physical components was his claim. This got a lot of people talking, debating, uh, agreeing, disagreeing. Uh, I came

out in disagreement. I thought this was a pretty ridiculous claim, but uh, seems like I am in the minority. >> Yeah. Wait, this doesn't make any sense. These those are all moes. What What is What is Nikita's claim? Those are all moes. So what was clarified to me by people who are defending Nikita is that he said the end of technical moes and the end of technical moes could perhaps mean that like oh my moat is that I have developed some software that's super fast and is super low latency and like now you can just oneshot that with Opus 5.5 and so like you know to the extent there there was a lot of stories this last couple weeks about people recreating Photoshop and Figma using Opus 5.5 5 or using Fable 5.1. Uh, and

the idea is that okay, if your moat is that we built this very complex software product and we have really smart engineers and we spent so much time on it and you'll never catch up with us, those moes have weakened. That's the the steel man of the claim. And if you don't have a moat that's in some kind of data wall or some kind of physical components that are difficult to replicate, AI still can't do that. You know, he emitted regulatory modes which are obviously another category of mode. there's brand mode, distribution mode, there are a lot of other kinds of modes that he did not address, but at least this category of moat, technical modes, perhaps has weakened to some degree.

Now, I'm still skeptical of even that, but um you know, there was there was actually a story recently from uh I think it's Xbox that Xbox just launched Xbox Labs to take advantage of the fact that we've had this bevy of people on Twitter announcing that they've used AI to port games wholesale like group oneshot to new architectures which used to be a thing that would take you know a year and cost an entire studio you know a lot of time. >> Right. Exactly. And now like you can just AI can do the entire thing without any human intervention.

>> I saw some claim that I didn't validate that someone completely broke the PlayStation 5 game like DRM >> and like online store via Vibe attacking and um so it's like it was like in this vein cuz like they were trying to take the PS5 binaries like non DRM and then convert them to Xbox. So that that was like kind of how I I ran into the story was that there was like people trying to basically not one shot but with like some small chat convert some PS5 game to run on their Xbox which is fascinating because for a lot of these things it is kind of analogous to the conversation we were having last time about cryptography is that for things like DRM a lot of the

protection in DRM it's not like we're using fancy cryptography or like program offiscation or any of the super fancy stuff because it's super slow and expensive and you would never actually want to use it in production like playing a video game. So most of this stuff is really security through obscurity of like there's just some little gadgets and things that they shove in there that are hard to break. But hard to break is really comes down to proof of work. Is just that if somebody who's, you know, sufficiently motivated and autistic spends enough time trying to reverse engineer it. Like every DRM will eventually fall unless it's built into the hardware. Now these days there is a lot of DRM built into the hardware, right? So like your laptop will conspire with Netflix to prevent you from pulling the data off your

monitor. So uh now of course there's oldfashioned you know just you know take your phone out and like record your monitor but you cannot read the pixels off the screen because of DRM that's baked into the hardware. So this kind of hardware based DRM is probably more where we're going. Like there will probably more and more of this just because it's not easily breakable by AI. But we're kind of at this like software apocalypse level flattening of the abilities between hobbyists and software companies. Like there's there's basically no more difference in what PlayStation company can do versus what you know somebody sufficiently motivated to break open their PlayStation and start playing DRM games can do. Coming back to the question technical modes to run do you buy the story? Are technical

modes dead or do you think death is exaggerated? I kind of mainly I don't know it doesn't seem that I mean he missed a bunch of moes but like I feel like on the pure software side it's like kind of hard to argue that the the moat is software in the long run. I think like the the moat on the security side might actually be much better like like building secure sandboxes and secure having enclaves that are managing agents and stuff like I actually think like that but again that might be cheating because that's like almost moving now to the physical thing he said at the end.

I'm like kind of so I I I I kind of do agree. Yeah. I mean, the fact that people were able to like convert these PS5 games that probably costs like what tens of millions to make into an Xbox game that was playable in like a few hours or whatever. What moat is there for a game studio? There's like not really any. I think it's like purely like how fast can you get to market your idea? Everything becomes like a liquidity game like DeFi and like less like a long-term >> brand. No, no, no, no, no. Hold on, hold on. to transpiling a game from one architecture to another architecture is like a very small part of game development. Like that obviously that's a thing, but it's kind of like how long did GTA 6 take and how long

will it take for someone to >> but that's not for coding. That's like the the environment design and the voice acting and like that's not the technical mode. >> You do not want to see the AI generated GTA 6. I guarantee you it's going to be I don't think that's a technical mode. I think that's like sure the idea has to be separate fromation. >> Yeah, >> creative mode is different. That's not a technical mode. Like >> since when was that Rockstar's moat technical? You know, it's not like wow, they have such an amazing game engine.

>> No, it's hard to make a >> shoot the bullets. They go faster than any there's like two companies that can make a AAA game. >> I'm not saying it's it's easy to make triple Alea game. I'm saying like do they can they do it with like 150th the engineers and like a small amount of like really good creatives. >> How many people do you think are working on the actual engine? >> I would guess like a rounding area. They're probably reusing the same engine they've used in previous games. >> I don't know. There was actually wait. There was some story recently about like Epic or Unreal like forcing all their developers to like stop manually coding recently.

>> Sure. >> I did see that. Yeah. Yeah. >> Yeah. And like I think the end state of this is like you don't this the the software is remote is like number of engineers in some ways historically and clearly that's going to dimminimous like it's clearly greater than one. I think the idea that it was zero has is not true. But >> I agree with you that in like five years you will be able to say computer please make me this game that I can visualize and >> no you will you will you will let your brain computer interface think.

>> But then creativity and taste is the mode. >> Yes. Correct. But that's not a technical mode. >> Yeah. Okay. So, okay. Hold on. Hold on. Hold on. Step take a step back. Take a step back. Okay. Here here's where I I disagree with even this claim. It is true that there are some technical modes that are easy to overcome, especially when they're extremely well specified in advance. So one example somebody gave to me was like compilers. Okay, commercial compilers dead. AI can oneshot that. It can it can make soda. You know, you can use some auto research loop in order to make it better and better. That is exactly the kind of thing that you can absolutely get AIS to oneshot perfectly, right? Get an EVM implementation. AI can oneshot that. Like that's not a technical mode, right? That's a

technical mode that's dead. That is such a small portion of companies. It describes so like just go look down the Fortune 500, find me companies that that is their moat. It is so rare today that like that is the core product of the company that makes it good. Even you know people are talking about Adobe. I think that is not Adobe's moat. That is not Adobe's moat, right? Like go look at these oneshotted Photoshop clones. They're all ass. They all suck. They're like they look good for like the first 10 seconds and you try it. I'm not I'm not saying you guys can build. There used to be a a thing where it's like you're Guido Van Rossom, right? Creator of Python. You suddenly can go work at any company and get paid a lot of money just for that. I think that job is gone,

>> right? Like think about the Ruby founders like diet tribe. Like that's kind of the and that's what I think of when I think of a technical mode. Like you have this kind of like >> engineer superstar person and like oh it's like why would you try to compete with us? Like I have so many of those. I think that's less >> thought that was a mode. I mean that's that's what I'm saying like this is this has never been a moat. Nobody would be like I have I feel like there are people who do think that's a mode historically >> for Okay, describe to me what is a company for which that was their mode and that that mode has been eroded by AI.

Name for me a company for which that was true. >> Lots of trading firms. >> But the trading firms are doing even better now. Like Jane Street is >> some are doing better. Some are doing better. There's a lot there. >> Okay. But then that wasn't their moat. Then there wasn't their moat. Like the people for whom that was a moat, which is Jane Street, which is, you know, sits. >> I don't know. These guys are even farther ahead now. >> There's some people I think that are actually doing worse who had a lot of very good people like >> Well, I'm sure there were, but there would but they would be some people were doing better and doing worse anyway. >> I'm just saying the concentration effects are much much more severe. >> So that was a I agree with you. That was a mode before for Jane Street is they

had all the best talent and that mode has strengthened. And so to the extent that that was not clear that most talent has actually might even be accelerating for certain >> I think the talent has weakened for them right a lot of them went to AI labs >> um sure but >> I actually think that the street is now the trading people have actually had the biggest exodus in the last year and a half so like I would argue their talent went down but the AI was able to replace the lost talent >> ratio is what you care about >> yeah the talent okay maybe I'm I'm the The reason I kind of get hit the vibe of what Nikki's thing is is like he's kind of saying this whole thing of like the advertising of Silicon Valley startup of like, oh, we we raised all this money so

we can hire the best talent so that someone else trying to copy us will have a harder time. >> That there was an ethos of that for a long time. And I think that's just gone. >> Yeah, but that's not a moat. That's an advantage. >> But that was something in the past. I'm just >> No, it wasn't that. No, I if somebody told me my moat is I hired a bunch of engineers, I'd be like you are [ __ ] That's not a moat. That's obviously not a moat. That's the stupidest thing I ever >> You know what? I I'm not gonna I'll let the internet dunk on you versus me because like clearly you got you got into this tiff and other people. >> Well, I Well, the thing is like I think but I think it's important because you know we are investors. We talk a lot to

entrepreneurs and they're reading this stuff and they're like, "Oh, are Moes over? Should how should I internalize this?" My answer is that almost not at all. I think this is like a fake story. This is [ __ ] We have not seen any major companies that have had their moes eroded by a post AAI company, right? There are no examples. Nobody could come up with any examples to give me other than some very obscure companies that build compilers. I'm like, that's not what [ __ ] company is that that builds compilers and that was their moat. You know, if think about like browsers, right? Like there was all these examples of Opus 4.8 oneshotting browsers or there was a C++ compiler and like obviously these were great tech demos.

Nobody uses any [ __ ] vibecoded browsers because they're [ __ ] Because they suck. They're good at the first. >> I'm not disagree. I think he's just trying to say like in the future that that's gonna be true. I like I I don't know. I didn't read this thing as like saying I feel like you really read into this a lot more than what >> I mean everybody has been talking about over the last few days. >> Yeah. I mean this feels like you know the March SAS apocalypse you know redux right. It's it's like, oh, you know, I'm not gonna use whatever's here. I'm gonna like vibe code my own, which >> I don't think is really true. I uh I think >> Yeah. I mean, look, for instance, if you're trying to spy on your employer with a a camera you bought on the

internet, don't vibe code the software you wrote to monitor it [laughter] with >> private. That's that's that's critical. That is actually critical to doing that successfully. >> Um, sorry. >> Oh, no. I think it's uh back then the thread was like it's not that people write their own software. It's like the model just eats the software which is kind of the thing that I guess we've seen. It's like the thing that is competitive with Photoshop is not vibecoded Photoshop. It's just the image models have gotten so good that like why would I ever open Figma or Photoshop because I can just prompt the model and generate the thing that I want.

>> Totally agreed. Yeah. So like the business models that are getting disrupted are not getting disrupted by vibecoded versions of the original product. I I can't think of a single example of that. >> Yeah. Yeah, I look >> obviously there's a lot of Twitter demos. I I I just think you're kind of like throwing out a little bit of there is some there's some part of the world that it people feel like that like the engineers working on it feel like that which changes how they view what they work on which I think will have some impact on on a mo of these >> look there's no question AI is change like obviously AI is changing things right but the question is how much is it changing fundamental business dynamics it is very easy and compelling to say everything has changed businesses are now being turned on their head. Moes are

eroding like nothing is ever going to be the same. And so far that call has aged incredibly poorly. We are 3 years postg. We're a full year post opus 4.5 and you know where where's all the disruption? Now look, you can always say oh it's coming it hasn't happened. You know, oh you just have to give it time to fusion blah blah blah. But I think actually the thing you have to appreciate and Jane Street's a perfect example of this. The incumbents have AI too [ __ ] No, no, no, no. That >> they're also they have bigger budgets than you do. They have more token spend than you do. And like to the extent that they will be pulling farther ahead >> and their software will be more robust.

I mean, look, if if an organization fails to compete because they're slow and they're lethargic and they're just not >> I just I just think a lot of not for that reason. >> A lot of people in finance have historically been like we have the smartest traders, whatever, and like that's our moat versus like we have liquidity advantages, order flow advantages, infrastructure that's lasted a while, relationships with issuers, whatever. Like all this other stuff that's really the thing that like increases your size, right? Like even for Jane Street HRT, right? until they started doing primary issuance, doing ETF stuff directly, whatever, right?

Like their P&L was kind of a lot like not in the billions and then had the huge jump once they started getting more of the like regulatory ability to do kind of larger capacity things. I I I just think like there's sort of a sense in which there are places like that where that's true where like everyone used to brag about who they hired and I feel like that they've clearly it's clearly been become like nah we can lose them why they were too expensive like I can replace them with a couple smart undergrads and AI and like I think that aspect is more what I was I can kind of was kind of the vibe I read from that tweet.

Well, it sounds like what you're reading into it is more like an individual losing their quote unquote technical moat. >> Yeah. Company losing a technical >> Yeah. Maybe that's kind of how I read what he was cuz like his audience was sort of like people who want to get rage baited. That's probably not the companies themselves. >> It's probably the individuals. >> But it's a lot of it's a lot of founders, right? Like I mean obviously he's a big growth uh influencer, >> but it's also a lot of engineers really. >> It is a lot of engineers. It's it goes both sides. It cuts both sides.

>> The ultimate thing I think increasingly with vibe coding is that if your product has a verifiable endpoint and you can just you know send your codeex or your cloud code overnight to grind on this problem and by morning it will have a verifiable yes this is definitely a super high performance compiler and it passes all the tests. That kind of thing is dead. That kind of thing there genuinely is no more technical mode. that describes such a small proportion of products. Like even Photoshop does not is not like that. You vibe code Photoshop and you like great it passes all the Photoshop tests and you start going and playing around with it. You're like, "Oh, this thing sucks and it is totally wrong." Like the taste >> product lock and stuff >> sounds like it wasn't well specified.

Sounds like you got to you got to work on that that spec. Yeah, >> that's the thing. Well, that's the thing. It's hard for these things to get encapsulated in reinforcement learning with verifiable rewards. You can't easily specify make a killer product and then just hill climb a bunch of benchmarks. Like we don't have benchmarks for that. Maybe someday in three, five, 10 years, who knows when we have real super intelligence, it's going to have killer product taste, but right now we don't know how to put product taste into the models. So, I think about this with investing. Like, when you get a model to write an investment memo, it gets a lot of the details right, but it's such it's such a little it's such a little [ __ ] It's so scared and conservative and it like it's it's like

a growth investor, you know? It's just like so scared of losing money and it's like, oh, what? Think about all the thousand things that could go wrong. And it's like okay yes full marks in your business school essay but like this is real investing and in real investing like things are not perfect and you have to take risks and the models they do not know how to do that they don't have the taste and how would you put the taste in the model nobody I don't know of course there are not the inner work you know the the invest slacks or docs yet but you know >> well the defunct VCs are going to have terrible memos you know so like there's an adverse selection >> well first of all the reason why they don't have any taste is because a the venture funds don't make their

investment memos accessible to the internet >> and these models can't scrape them and they can't learn from them. I mean venture is one of the most blackbox things there is, right? >> Yeah. >> Like there's almost >> and the memos that you do release are the ones that you're not afraid of the world seeing. >> Yeah. You release your Facebook memo after Facebook goes public. Yeah, of course. >> I know. Like wow, look at all the underwriting we did at the, you know, seed stage of YouTube. like we really did this market sizing stuff and it's like you know come on you probably just were like ah you know founder smart yolo let's do it. So like all the memos that look like that you're not releasing in the archives. So I think the uh >> I think the I think the other thing that's just worth keep in mind. Yeah. Is

like I I think the companies Yeah. the company versus individual thing I think is the thing that sticks out to me is like who's the audience of that tweet? I don't I don't think it's companies. I really think it's how individuals feel. >> Drew, you are you are such a Nikita Beer fan. I had no idea you're really standing up for him on this one. >> I mean, I just I It's like I don't think it's like that crazy to write that to like these like depressed engineers who are like, "Did I just waste?" >> I think it's it's red meat. It's obviously red meat. People [ __ ] love being told that [laughter] everything is over. And like >> I'm just kind of like, all right, he found a reason it hit so hard was cuz he he really tapped into that.

>> He said what people wanted to hear. He definitely said what people wanted to hear. Um I >> And you know, another motive, making a cult and being a religious leader and saying things like that, you can make a cult. >> That's true. Being controversial is always a cult. Okay, it's getting late and uh I think we're all fading. So, um, thank you everybody. We'll be back next week. So, I think actually we're going to be experimenting with doing a host swap for the next Chopping Block episode. So, uh, stay tuned for that. It'll be it'll be fun. We'll announce some stuff online. But until then, see everybody.

[music] >> [music]