Crypto Is Shifting Gears + AI Router Wars - The Chopping Block

Transcript

We had lunch with the king of Bhutan. He's actually very excited about AI. Told me he uses cloud code and Obsidian, but he's been playing around with Kimmy and he was impressed with Kimmy 3. >> What? >> Not a dividend. [music] >> It's a tale of two fun. >> Now your losses are on someone else's balance sheet. >> Generally speaking, airdrops are kind of pointless. Anyways, unnamed trading firms who [music] are very involved. >> I like that 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. So, first up, we've got Tom, the DeFi maven and master of memes.

>> Hello everybody. >> Next, we've got Tarun, the gigab brain and grand puba at Gauntlet. >> Yo. >> Then we've got Robert the crypto connoisseur and Zar of Superstate. >> Hello everybody. >> And finally, I am Hed, the head hype man at Dragonfly. We are early stage investors in crypto. I want to caveat that nothing we say here is investment advice, legal advice, or even life advice. Please see chopping blocks at XYZ for more disclosures. So, for those wondering, Tun and Tom are not having a spa day. Actually, they're joining us from Bhutan. >> Getting congratulations. [laughter] >> Thank you.

>> Well, yeah, that's very very exciting. Uh tell us what you guys what are you guys doing in Bhutan and uh how are things over there? >> Things are lovely. Um, we're here for a variety of reasons. Um, there's there's a demo day here. We met with the king. We had lunch with the king of Bhutan. Learned a lot about Bitcoin mining. He's actually very excited about AI. Told me he uses cloud code and obsidian, but he's been playing around with Kimmy with Kimmy. Yes, he's very >> King of Bhutan. Honestly, one of the most amazing stories of how he decided to mine Bitcoin when Bitcoin was $4.

>> I mean, the guy is extraordinarily eloquent. I was definitely blown away. I was like I was wasn't expect I kind of my problem is like the British monarchy where everyone's kind of a like Prince Harry or whatever. Like I just kind of expect everyone to be a when when I hear and this king is like the opposite. He's kind of like clearly a genius. >> Wow. I I did I did meet the king of Bhutan once at an event that Binance was hosting and he he was I remember him being very suave. Just like super cool guy. He had these like he had these little he had these like little glasses that and he was just kind of uh seemed incredibly gentle and I don't know he seemed like a character from a movie. It was like okay this is I the first king

I've ever met and uh very good introduction to the world of kings I will say. >> Yeah. Yeah. Yeah. For sure. He made me you know I my impression was just like again reality TV shows with the British royals and I'm like wow they're all morons but clearly that's not true here. Are are you a monarchist now? >> No, I don't think so. Maybe. I I don't know. >> Maybe for a good one. Yeah, if you you can get the king of Bhutan, won't be so bad. >> Well, I mean, they also have a prime minister. Like, it is a democracy. It's constitutional monarchy. So, like they >> But England has the legacy of a monarch. >> But the people kind of Yeah. The the vibe the vibe around it feels kind of bad. I don't know. The vibe here around it is like everyone loves King. It seems

like kind of chill. >> And he seems much more involved. who's working on this GMC project which is kind of like their own SEZ kind of thing. So they want to get more entrepreneurs over to over to Bhutan and you know he's saying and he was right like I feel like there's this Central Asia focus like like the the orb is kind of shifting from Southeast Asia to Central Asia and uh well positioned. >> What is the King of Bhutan vibe coding? >> We didn't get that far. Sorry, it was a it was a 12 14 person lunch, so I couldn't really go deep. But he uses Obsidian, so I think that's that's another kind of, you know, deep cut.

>> All right. I feel like you got before you leave Bhutan, you got to get the word of what he's is he like vibe coding government services for the for the country of Bhutan or like what's going on national ID ID system on um AWS? >> We did we did make the the the the folly mistake of not asking him if he wanted to be a chopping block guest. I apologize. [laughter] Okay. >> Next week on the chopping block, [laughter] >> we do get some leads on other guests though for future episodes. So stay tuned. Yes. Very good. Very good leads on us. So >> Okay. Okay. Excited for that. Well, it's a it's a it's a good time to bring on different guests because uh crypto is finally cool again. So the big story of the week has been this big rebound. No,

>> I think it's Look, I think we're we're back in the headlines. People are excited. We're getting a lot more energy in the markets all of a sudden. And this has happened right as Bitcoin is just scraping against $80,000. It went over 80K yesterday and now it's kind of right around 79.5,000. There's been a huge amount of short liquidations uh over 2 billion in ETF inflows, which the ETFs have been very quiet up until very recently. All of a sudden, now we're seeing inflows again. And a lot of this has been accelerated by what's happening on the macro side. So big story in macro has been Secretary Bessant has started to try to pull in the long end of the curve buying back longdated bonds to issue short and all of that has been very stimulative of

risk assets. So it has brought back what's been called the debasement trade this idea that oh okay you know the central bank is starting to or not the central bank sorry uh treasury is starting to meddle in controlling currency markets and bond markets again. Besson basically thinks that the long end is too high which showing some lack of confidence perhaps uh or lack of demand in long issued government bonds. So all of that has gotten gold to rally. It's gotten Bitcoin to rally and it seems to have people thinking again of like hey maybe finally it's time to revisit that old Bitcoin thesis or crypto thesis. And all of this was was capped by a rise in Hyperlid. So, Hyperliquid recently hit close to an

all-time high. Uh, or actually, I think no, it did hit an all-time high. And this was accelerated by Trump at a White House summit talking about crypto where he said that Celig, the chairman of the CFTC, was quote working to bring Hyperlid into the United States in a fully compliant and legal fashion. This saw Hyperlid rally to crazy heights uh over $80. And um this seems to have been a culmination of a lot of the groundwork that the hyperlquid policy center has been playing in the US to try to figure out how can you bring hyperlid to the US in a way that is more compliant. And lastly, there's been a lot of petitioning from the hyperlquid policy center for the CFTC to exempt non-custodial DeFi like hyperlid uh from legacy exchange rules that would

normally cause them to have to register, do KYC, etc. and and all the things that uh normal exchanges has to do within the US. So there's also a push to see if clarity can pass. There was a lot of mention from Trump at this White House crypto summit about clarity, but clarity is still sitting now around 15% on poly market. It looks like clarity is not very likely to pass this year. But uh if you if you draw the line forward for another 2 years, clarity is probably about 50/50 it seems like to pass before the end of 2028. So stop there. thoughts around the horn and and what are you guys hearing sitting there in Bhutan about the sentiment towards crypto and Bitcoin?

>> Well, I I think the king is just never been more bullish. I mean, it's kind of crazy. I I I think the other thing that's interesting is um I think whenever you go to Asia, you always remember like there just people who just love tokens, you know, like it's just like sometimes in the US there's this like kind of it's like a on andoff relationship, right? It's like, oh, we love tokens, we hate tokens, we love In Asia, everyone's like, no, no, no, no. Just one more trick. There's going to be a new token. [laughter] There's going to be one token that makes make it through.

you know, like >> I think that uh that mentality you can kind of tell is like people, you know, last time I was in Asia and it was like kind of more bearish. Everyone's like, "Ah, yeah, there'll be one more token eventually." And now it's definitely like, "Yeah, wait, it could be my token, you [laughter] know, like that's how I would describe that." >> Weirdly, I think there are like you still see some of the US specific trends like FOMO obviously has been, you know, on on a tear the past few weeks. and has a big fundraising round as well. And there's just really not as much interest or excitement or even like awareness of of FOMO um in Asia from from some of the folks that we're talking to. But I mean overall people also obviously do pay attention to the US. We ran into CZ.

He's very much paying attention to what's happening in the US. And so people are curious about about this this crypto summit. I think there's one perception I think especially among retail that means that this means there's going to be full boore anyone in the US can just go use Hyperlid and it's going to be you know in in your Schwab account. I think that's like not going to happen. I think it's going to look more like what lighter did with with Robin Hood in Europe where there's a KYC gated version of it and maybe there's some some interoperability. Maybe you can use collateral across different deployments, but there's going to be, you know, some sort of KYC gated version or some version of it that that meets sort of the standards of compliance that you the US expects of new venues.

>> Rob, what's your take? >> My take is the devil's in the details. Definitely everyone is aligned with trying to figure out how do you wrap your arms around the big complex new thing that everybody wants. I think there's going to be a lot of figuring out the nuances here about how to make it work. I don't think it's going to be a quick thing. I think everyone is directionally aligned. But I think there's also going to be a lot of adversarial conflict with legacy incumbents as they watch something huge and new come in and they're afraid that they get treated to a easier standard. Right? We

talk about no KYC. Everybody on earth would prefer no KYC, right? It is a massive friction. Nobody wants to see, you know, something so permissive. Not everyone, but yeah, most most human most humans on Earth who who who have the fortune of never getting a legal degree [snorts] >> but let's just say it is an overwhelming advantage to not have to deal with those frictions, right? It it's operationally simpler. It's easier for your users.

It's blah blah blah blah blah. And so, everyone's directionally aligned. I'm just skeptical that it's going to be an easy journey. Yeah. I mean, to your point, the CME stock took a big whopping on the >> Oh, yeah. Like, do you think the CE wants to see this? >> The what's his name? Duffy. He was just like on CNBC getting angry. So, like I don't think it's like a hey, everything's going to just happen type thing, you know? >> Right. I mean, it it's interesting that this made it a Trump speech, you know?

That's a level of specificity that I would not have expected at this stage without more details. Um, >> so Baron is trading on Hyperlid, right? That's what we're >> somebody in the inner circle seems to really care. Hard to know who it is, but the given how like I I think Poly Market is probably the clearest blueprint for what that looks like, which is that okay, you know, there's an there's an overseas version which is no KYC, which is, you know, totally open. It's on the blockchain, blah blah blah. That one's decentralized. The US version is probably not going to operate that way.

The US version is going to have to be, you know, market surveillance. It's going to have to be KYCed. CFTC is going to have to oversee it. They're going to have to know everybody who's trading on it. They're going to have to be able to pull their names if there's some kind of market manipulation. You know, right now more than half of their volume. >> They might not approve ADL. If you look at sheet one, that a big point right? >> Actually, I I I wasn't aware of that. How on the per side, what's the alternative to ADL? >> Like normal futures clearing, like centralized clearing.

>> Oh, I see. >> Like you have like a a broker who takes the the hit, like they put up capital, risk capital. So I I I think that that stuff is actually going to be the real devils in the detail. >> Uh is like how the the kind of clearing settlement stuff works because like that that's sort of where governments spend a lot of time writing law also. So like I don't think they're just going to be yeah we're gonna rip that all up and have it be fully algorithmic. That's my guess. Again just from the vibe I get from CFTC people who've talked to it's it seems like oh to your point I mean there's precedent now around you know offering you know onshore per to limited extent through through ki. I think the big question is can you offer a compliant DAX a separate you know

deployment or a separate version of it to people in in the US and and still sort of meet like you said you know for example surveillance requirements and I think that was part of the discussion was like hey actually a dex that's well built has a lot of things that that you would normally have to sort of do do manually for compliance built in um like the code actually does it for you and so that would be a huge step forward versus hey this is just another you know centralized per venue which I kind of imagine there's a huge pipeline of of uh you applicants that are about to be approved. >> Yeah. Well, I mean, one of the distinctions though is that well, I mean, it's it's like Poly Market US in that you probably do have to have a separate pool of liquidity, right? You

can't have domestic people trading against liquidity on the offshore uh or on the overseas onchain exchange in large part because of course you don't know who the other person is. You have to be able to unmask the person you're trading against. So if somebody is filling you or you're being fil or you know you're filling somebody else and that person is in some unknown country well then half of the trade cannot be surveiled right and that's a big part of what the CFTC does is to make sure markets are orderly no one's manipulating anything blah blah blah and given that half of their volume is RWAS more than half of their volume now is RWAS like obviously that exists under very different regulatory regime that has to continue to be respected if you're going to do that domestically. So

how that's going to look obviously TBD we got zero details nobody seems to be super clear exactly on how this is going to be structured but it is very clear that pers like we're going to see this what poly market did is going to happen on the per side as well probably you're going to see there's been a lot of chatter about lighter in which we are we are all investors we're investors in hype as well there was a lot of chatter about lighter doing the same thing of course they have this partnership with with uh uh Robin Hood and lighter has also been rallying on the news that hey you know it looks like Pips are going to be coming to the US and there's going to be some room for these products that are decentralized to have some kind of gated US version. So big story but on the

other side we've also seen a lot of regulatory developments now that clarity seems call it unlikely. uh both the CFTC and the SEC have been signaling that they are going to be starting to pass some kind of rulem to basically get the same thing that's in clarity but via rules instead of by law. Right? So we don't know how long clarity is going to take. There's an outside shot that it passes this year but most likely it's going to be either 2027 or 2028 story under a very different Congress. So in the meantime the SEC the the CFTC has signaled, yo, we're going to do this.

We're going to move forward with something or other. We're going to pass some rule making on both prediction markets as well as for you know the they have a compute markets RFC uh as well as a bunch of stuff that they're going to do on on pers and uh crypto exchanges. And then the SEC came out with a 42page release for a set of rules called regulation crypto assets. Their first actual crypto rulemaking that they're going to be doing. They they've uh elicited a 60-day common window. And in this so I haven't read this entire release to be honest. I can't imagine most people have. I threw it into my AI and started poking around a little bit.

The main stories behind this rule is that there are going to be fundraising exemptions. So, one of them is a one-time 5-year fundraising exemption over four years. So, you can issue some tokens, sell up to $5 million, or if you're issuing audited financials and some ongoing reporting, you can raise 75 million over 12 months per over a 12-month period. So very substantial fundraising opportunity for for uh token projects that they can do under this new rule. There's also a conditional safe harbor. So you maybe uh if you're a qualifying token uh you can you can basically elect that look there's not an investment contract and you can exit from security laws entirely from some

kind of SEC approval. they can go and look and say, "Hey, you're no longer fulfilling this kind of managerial role and therefore this thing, you know, the investment contract has now expired and all of this is still subject to comments. So, I think a lot of crypto funds and crypto projects are going to be weighing in on this set of rules that are that are coming out. So, uh, Robert, given that you are the one probably closest with your ear to the ground on the regulatory side, what's been your perception of the SEC rules as well as what the CFTC has been signaling here?

>> Yeah. So, I'll just focus on the SEC and rank crypto because this is very tangible proposed rulem. It's not nebulous in any way. This is rulemaking that in a lot of ways dates back to the Dow report and the process of raising capital using a an ICO, right? Token fundraising. We've had this regulatory chasm going all the way back to when Ethereum started conducting ICOs. There's been a lot of concernation about it and with the stalling of the clarity act the SEC has really stepped in and said hey we are going to define something that needs to be defined when is it appropriate to raise capital using

a token to like sell token for money to build your project it's something that has been talked about for a really long time and how do you create a framework to do it correctly and properly right this is something frankly that I think they could have or should have released 8 years ago. This is not something that required any legislation. You know, I think one of the reasons why it took so long is, you know, especially in this administration, they were waiting for the legislation to come first, right? If the Clarity Act passed, I don't even think they would necessarily need to come out with reg crypto. What reg crypto is is it looks a lot like other paths, alternate paths to raise capital. Looks like reggae, it

looks like reggae plus. It looks like all of these things have been developed in the past to say how do you raise capital with a lighter burden than going public, right? And so I think reg crypto does that very well, right? It says how do you raise a little bit of capital? How do you raise a lot of capital? And what are different requirements? Raising a lot of capital, it looks like rag a stuff. It looks like mini going public. the burden on a project is not insignificant, right? Like this requires you to be a pretty buttoned up project if you want to raise $75 million, but if you want to raise $5 million, it's actually very permissive. And I think a lot of projects could fall into this path, [clears throat]

right? There's still a lot of projects that are going to say, "Oh, it's easier just do it offshore and not offer it to us people and blah blah blah versus do a little bit of paperwork." But I actually think the $5 million path is going to be potentially exciting. If we go back to a market where a lot of founders are coming into the space, it isn't very burdensome, frankly, to raise $5 million using this legalized framework, right? I could see projects doing token sales in a small way, you know, to get off the ground. I could see people testing this.

I could see people using this. I could see victories coming out of this. Frankly, >> in these these token sales, you can do full crowd sales. These are not like only to accredit investors as far as >> correct. Correct. C >> can I ask a very stupid question because I haven't read this. Is it only in USD funding? Like I I can't do like an ETH denominated ICO sold. >> No, you should be able to. >> You should be able to. Okay. I wasn't sure if inkind mattered versus not. >> No, it's just about like the value of what you're raising, right? It's like how much capital are you taking in by selling tokens. I just feel like there's lots of ways of getting around the $5 million limit if you're taking in kind.

That's why. >> Oh, right, right, right. But like, yeah, you're able to raise $5 million of capital by selling tokens. And I think this is a good thing, right? This is objectively positive to allow this framework. I don't think it's going to be that controversial. This is not something that completely disrupts how traditional equity markets operate. This is not something that necessarily like runs into the complex friction of incumbents. This is, you know, just like no one was opposed to Frank A. Everyone's like, "Oh, making capital formation a little bit cleaner, a little bit better. High five. This is not that controversial." This is not a controversial set of rules that are being proposed, right? If you wait, so

Robert, so okay, the $5 million threshold, I agree. It sounds pretty light touch. You know, there's some self there's some self-certification you have to do. There's some like basic information that you have to disclose, but it's pretty it's pretty easy. It's not super expensive. The $75 million threshold, right? To get up past 5 million, you have to get PCAOB audits. >> Yeah. No, it's it it's I don't think that many people are going to adopt it. Frankly, this is my contrary take. I don't think it's going to get used. I think they designed a rule >> and this like to get audits that are PCOB compliant, it that's probably going to cost you in the like hundreds of thousands of dollars, I would assume.

and then all the lawyer filings and whatever else you have to do to to get a >> you know whatever that threshold is. Now to your point, Robert, if you want to sell to non-Americans or you don't want to do a crowd sale, you're good. You don't have to do this. Which means that presumably to sell to VCs, you know, private investors, just do the normal thing you've been doing in the past, right? Exactly. You don't need this. You don't have to do this. >> Yeah. It's a form D. It's private sales. It's exact. Right. Right. So, I I I am also skeptical that people are going to use this $75 million tier of public fundraising in the same way that we've seen very few. I don't know if there's been any. I mean, if there have, it's gone under the radar. I don't think anyone's really using the

reggae reggae plus $75 million threshold. Nobody wants to do 100 financials, right? Like, it's just if you're doing that, you literally may as well go actually public. you may as well file an S1, right? Because like that's the same requirement. And so no one's using reggga A plus tier 2. I don't think people are going to use this relatively burdensome tier 2 of REGG crypto fundraising. I think it's probably going to be untouched, but because the standards are no different that already exists, there's really nothing for anyone to complain about. It just really says instead of selling equity and raising $75 million, you can sell tokens and raise $75 million. I

don't think it's going to be commonly used because very few people sell equity under reggae plus tier 2. I do think people are going to take advantage of this lower tier. I think it's going to be popular potentially. I think it's going to be exciting. You know, if you're a smaller project, like it used to be $5 million was a lot of money. doesn't mean you have to sell all your tokens. You could sell a small sliver for $5 million to get started. Whatever. I'm excited to see people test this. I think it's a great channel. I'm actually proud of a lot of the staff and the commissioners and the hard work that went into like thinking through this.

Um, and so there will be teams that come along and use this and I think that's a good thing. >> Drew and Tom, what are you guys thoughts? >> I mean, I think uh I agree with Robert and that you should have come out eight years ago, but maybe more in a more pessimistic sense. I you know I think the really the crux of of you know the industry's beef with the SEC right now I think it's more around asset categorization what is a security what's not a security and what how's that path sort of you unfold and obviously that plays into fundraising but I don't see a lot of new teams that are you know in the spirit of a 2017 ICO trying to raise some money to build a new network that gets decentralized and what does that threshold meet etc etc building a new D5

protocol it's much more about there's some sort of you know app or some sort of centralized company that has some value and they also want to issue a token. They want to send some value back to the token. How these two interact it's it's a sort of this open question we talk about all the time. But that's really I think what people look for and obviously that you know allows there to be you know regulatory clarity within the existing uh u um uh you know reggga or regggd uh um security sales regime but short of that you don't actually know what the token is and so therefore you know hey you can have safe harbor but you know to what end?

M so your your point is almost that th this set of regulations is really designed for an older generation of crypto where a lot of those founders were building decentralized stuff >> today you know how many things do you get pitches for that are like yeah I I intend to walk away from this project and no longer be involved because it's all self-sustaining on chain. Yeah, exactly. And that was like, you know, the spirit of those those ICOs. And that's why, hey, there was this all these these lawsuits around, well, what were these tokens um sold as? What did they become? How did they become what they what they became? What was that that path? And so there's both of those sort of components. And obviously the two are intertwined. But today, you just don't see a lot of teams that are from

day one or even for successful teams talking about, well, we want this to be, you know, totally decentralized. We're going to do our first raise as a crowdfund, you know, build a decentralized network. maybe it'll come back and and maybe hey this is providing the path to do that but it's not really what what we see in the market. >> Yeah. I would also just say like the the word decentralized has been watered down and changed so much that like I don't even think there's the threat what what the SEC considers decentralized what Hyperlid considers decentralized what Salana considers decentralized what Ethereum considers decentralized are all kind of different notions. So, I'm not even sure like this is Yeah, to to point >> there's going to be a little standard

now. >> Yeah, it's it's just a little feels like Yeah, a little too little too late. But I do I think the 5 million threshold is interesting. I'm kind of curious if there's going to be someone who like tries to make like a holding company that launches like 20 different tokens to get to the 100 million without having to do the audited financials. Like I'm sure there's going to be some some kind of like clever way of farming the 5 million limit to to emulate the 75 and someone will do it. It's crypto. There's always someone whose financial engineering interests will lie in doing this. I think to also Robert's earlier point, I think STAX did a reggae offering 10 years ago and Mun was always complaining about how much of a pain in the ass it was. So hopefully this is,

you know, less of a pain in the ass. >> That's what I'm saying. look like it will be that much less of a pain in the ass. It looks quite pain in the assy. >> No, the $75 million one sucks. >> Yeah. But my point is like imagine I I find a way to make like a holding like especially in a world where everyone there's 100,000 token, you know, there's hundreds of thousands of tokens now, right? This is not the world eight years ago where it's like there's very few. It took a while to launch one, whatever, right? There's going to be someone who like memecoin style dodoses this legislation. Like I I I I I'm willing to bet that that will happen.

>> I think that the the DOS protection is the fact that you have to get like it's still the legal costs I think to do the filings. I was asking JPT to estimate the cost of like the just like the administrative burden of even the $5 million threshold. It's probably like between 10 and $20,000 of just like the legal work you have to submit to the >> 15 15 subsidiaries of one company and I can now I can now share those costs and it's not going to be 15 times 20k it's going to be more like 100k right so like [laughter] >> yeah fair enough >> I'm just saying I like there's there's there's going to be someone who figures out the engineering to like basically do many of the small offerings I I I don't

doubt that that will happen >> yeah yeah actually that's that's interesting because just yesterday somewhat unrelated but actually in a weird way related. I was I was reading this um this economics paper about this concept of like AI flooding of government services and the idea is that you know there's a lot of government processes that have a high fixed cost because of the the costs of basically navigating administrative complexity or just sort of bureaucratic And the costs have effectively decreased for a lot of government services of like you know for example applying for refunds or contesting parking tickets or sending comment letters like a lot of stuff like this that's actually designed to have

this proof of work that makes it costly and now a lot of those government services are getting overwhelmed because of LLMs and basically people sort of didn't realize that there was this fixed cost. They were sort of pretending that there wasn't a fixed cost, but that was kind of their flow control was like, you know, it because it like it's hard and you have to hire a lawyer, but now it's not hard and you don't have to hire a lawyer and therefore it's basically like a one-click thing that you can get your trust to do it overnight. And it was this paper was looking through a lot of these different mechanisms in different government services and being like look you have to kind of revisit what was effectively a speed bump that has now been removed from society and like a lot of these things are now getting

effectively by just genuine people trying to consume government services that we didn't realize we were rationing through bureaucratic And I kind of feel like this is the same thing potentially for capital markets, right? there's like a sort of rationing of capital markets through fixed costs that might go down because you can get your LM to like prepare the all the administrative statements or whatever and it's like okay now the SEC is getting overwhelmed by these. >> Yeah, I I don't I don't doubt I don't doubt there's going to be some civil attacks like it just like seems people are too clever and like this will be easier than the 75 million thing, right?

So >> yeah, totally. So this is ultimately good though. I I mean I'm very curious like at the end of the day this is still mostly focused on crowdfunding right if you're going to raise money from VCs probably you don't need to do this and any of the VCs who are worth their salt they have some kind of overseas structure and they have the ability to you know whatever kind of be non- US doiciled for for investing into some of this stuff. Um, so the interesting question here is like who is actually going to avail themselves of the $5 million crowdfunding limit, right? If you're a memecoin, you don't need this because you're not a security, right? We already established that memecoins are fine. So what are the projects that are going to be doing $5 million crowdfunds

to US investors and have to do some amount of disclosure, right? It's not nothing. You you still have to disclo, you know, you know, you know, you know, you know, you know, you know, have to disclose the team, you have to disclose what your managerial efforts are going to be, etc., et, etc. Like, you got to you got to put out some stuff here that does give you exposure. >> I I have a a maybe crazy hairbrain theory, but I think like to me this is the next token type that will be the L1 potential. Like I don't know if it'll be as big. Like I don't think it'll suck up as much capital as L1 tokens did. But but I really think that the Venice token like inference play of like I'm giving you this real service and like the state maybe if I stake I get a discount or if

I stake I get kind of like I I kind of get zero data retention from the inference provider, right? There's like different levels of quality of service you can you can do with the token and the token can have revenue on day one from the inference providers. In fact, we were talking about that with the king who actually is already thinking about this. He's very excited about tokenization. >> Very excited. Yeah. Yeah. Yeah. The king was like super excited about this. But the these types of tokens I think like you're seeing a lot of crypto companies becoming inference providers because it's like very similar to running a decentralized network, right? You're managing a bunch of nodes. Maybe they're more coll-located. You don't need them super distributed. But you're seeing this like for instance with IGEN layer becoming inference provider with with Venice of course. Um and you're starting

to see a lot of other teams kind of go this direction. And I think it's actually a very natural tokenization because you're delivering a digital good, right? It's not like I'm doing some off-chain RWA type thing gets complicated. I can kind of value the good. Now, there's some trust assumption out there for sure, but again, like I said, no one's trying to say we're tokens equals full decentralization anymore in 2026. And I think there's going to be a lot of these types of like small cap AI type of raises because you know the minimum AI raise like now is like feels like it's like 20 million, right? But there might be a lot of these smaller teams that are going to raise less to like do a small inference Friday to kind of like localize certain

technologies. So I I kind of could see that market actually being like they're going to be like many Venice tokens and like there's kind of going to be a boom there and that's like the non-mecoin token boom of the next cycle. >> Yeah. you you I feel like you saw a little bit taste of that this also this past two week like all the um non-crypto AI normmy people were really excited about uh was it dark bloom which was like the like Apple silicon you know sort of distributed uh >> which is from >> yeah yeah like like backyard compute kind of thing and it's like almost rediscovering the like decentralized inference you know idea from from kind of first principles and like running into the same issues and you know this obviously being kind of like the next step in okay well once we actually have

this like you know network quote unquote of of computers that we can we can you know do do inference on you know how do we actually go about you know distributing some of the value and bootstrapping like the rest of it I think it's going to look like mining pools actually it's like it feels a lot more like mining pools except like you don't have cryptographic guarantees you have these like statistical guarantees the cost of cheating is high versus like cryptography making sure it's like basically infinite but I think like if that we start to see more of those tokens take off like I could totally see those people raising on this like that makes total sense to Well, speaking of AI inference, one of the big stories this week has been that Stripe purchased Open Router for $7 billion. Now, Open Router was founded by

Alex Atala, who is the co-founder of OpenC, long time, you know, kind of in the crypto world ended up in 2023, I believe, founding Opening Router. Uh, founding Open Router. Open router for those of you who don't know, it is basically an aggregator of inference providers. Um, so it's kind of like 1 in if you if you know any of the DeFi aggregators, but they do that over LLMs and particular models for you know finding you the cheapest or fastest or uh most of you know uh low latency or highest caching rate LM provider for a given model. So they've been growing like crazy. Um this has been one of the big stories is just how fast AI inference has been growing and of course how fast open models have been growing relative to the centralized players. In

a world where, you know, it's just open AI and anthropic, maybe you don't need something like this. But in a world now where we're seeing so much more competition from all these other open models, especially from the Chinese labs, uh, something like open router suddenly becomes very compelling. Now, Tun, you recently published an article called caching cheaters on open router. A bit of a pun at like catching cheaters through caching cheaters. >> You you gave me the idea for the title. >> Wait, I I think the What was the original title? The original title was something very dry. you wrote catching and I was like actually why don't we just make it a pun by removing one letter >> I that was it was very very good very well done uh do you want to summarize you want to summarize the paper of what it was about because a very it's a very

like defy me inspired paper looking at what's happening in inference markets >> yeah so maybe I'll give a little highle view of this and and also why I think like there's a huge genius and what Alex did like whether he explicitly realized he was going to end up making like a very crypto-like thing out of AI stuff or not I'm not sure. He clearly obviously has amazing timing and taste historic like from two having two large kind of exits like partial exit and one and you know exit the other like it's kind of like you know there's real skill like he he really understands he he sees the future you know but there's something interesting about how open router started. So let me just give a

little bit of history on the open source model world. So like chat GPT comes out there's this huge uh you know burst of interest and demand this pre-clawed and that was at the time where like all the big tech companies were getting very scared of losing their modes right like Google feels like they're going to lose search or Facebook feels like people are going to use social media less because AI will make the content worse etc and so Facebook of course famously made Llama right the the one of the first large open source models there were a lot of smaller open source models but they were just not very good they were like not competitive. But once Llama launched, there was this whole cottage industry in 2023 of making like fine-tuned versions of Llama, like Llama for coding because at that time the

coding models weren't very good or llama for images to compete with uh like like you know people would make open source versions of midjourney like stuff like that right so there were starting to be these like flora and fauna of like special purpose models built off llama and it was really hard for you to know I want to make an image of se with hair right like that's my prompt and it's like okay which one do I go to and I think in the closed source world right you just get one prompt right it's like open AI gives you one thing clog gives you in the open source world it's very unbundled you have to figure out okay like this llama thing made by this academic lab fine-tuned apparently does really well on this benchmark for generating hair correctly realistic hair okay great that's the model I use right

so that was the original pitch for open router is like the flora and fauna of these open source models there's too many of very similar to how in DeFi there were tons of forks, right? There was tons of, you know, at once AMMs were popular, everyone started forking unis swap. Same with a same with compound, right? So, so you you kind of actually see this very similar thing and there was this kind of unbundling that occurred in DeFi of the centralized service, right? Like the centralized services Binance and Coinbase offered, DeFi unbundled them and but then it made it on the user to figure out how to use them and there was services that aggregated and that was like the natural thing.

So where we kind of are right now with the open source model world is something that looks really really similar to DeFi. So there's there's sort of the front end, the harness, so like cloud code, cloud desktop, whatever your your app that you're using. There's the router which decides which model to use. Then there's the model itself. And so if it's open weights, you know, you can go download, you know, the model weights and and run it. And then there's the inference router, the people who provide the GPU that actually runs the device. This is very similar to to onchain finance where the harness is like the wallet. It's the front end. It holds the data users private data. It might hold secrets like private keys, API keys, things like that. Then there's the

router like a DEX router. So if you've ever used a swap function in Phantom or one or unis swap under the cover there's a router that's routing you to different pools and you don't even know that you're using it. Then there's a model like the protocol you know they they collect some fees usually uh less than the other providers and then the inference are like liquidity providers or or validators in a decentralized network. And [snorts] if you look at the lessons of crypto, it kind of tells you that the the value flows to the edges like it goes to the the liquidity providers or insurance providers or it goes to like the wallets.

And so a natural question is if you see this kind of analogy, do you a natural phenomena in crypto is is me, right? Where validators might reorder your transactions, might put certain preference, certain users, might frontr run you. And the question is like do you see this in opener? So that was something I spent like a couple months just scraping data from open router. And the nice thing about open router is it's on a blockchain, right? It's like a private service, but they publish and post prices continuously. So these inference writers like a Venice, a base 10 together, they you know they can update their price all the time for like how much they charge you for the different tokens.

And basically uh the the type of thing that the paper focused on is there's a difference between cache tokens. So tokens that don't have to do the full compute. You you you get to to to reuse compute you did earlier in your your your chat versus input tokens. And but there's no cryptographic verification of that or there's no real verification. The the inference just says oh yeah I I used this many cash tokens or not. And so one way of manipulating that is saying, "Hey, I'm going to give a lower price on the expensive tokens, like the input and output tokens, like the fresh ones, but uh but I'm actually just going to say I never use cash tokens, so I can kind of charge more." And so there was there's evidence of of of that that already happening. The interesting thing

is a lot of the people who are doing it are used to be in crypto or we're Ethereum miners or or or have kind of connections to to to crypto. So I think the the lesson here is that this market structure which Alex kind of invented in some way like you you can really argue that open router opened up this market. It was not kind of you you need the router to glue together this market without the router. This market is a bunch of independent things you have to stitch together. It's like the UX in in a way a DEX aggregator. It's a little bit easier. But the the main thing I think that's interesting right now and and I'm kind of curious about your guys takes on this is how I kind of think we're about to have the revenge of the SAS company like enterprise company

because the last two years SAS company stocks have just been getting bludgeoned right like Figma you know obviously Figma had the the worst hit in the sense of like Anthropic literally made a you know initially it was a partner and then just like made exact a better a clone better version whatever and so there's been a lot kind of this conration of all these software companies are they going to survive because like they're they're going to lose uh their their seat everyone's just going to write their own CRM software using cloud they're not going to pay Salesforce but what you're finding right now is that these companies are finding a new way to monetize SAS users with routers which is sort of the revenge against anthropic so say you have data sitting in so so for instance like ramp uh the

credit card company launch a router Stripe obviously just bought open router, they have a router. Data bricks, Palunteer are both both have their own routers. And so you're starting to see this kind of like every SAS company at who owns a bunch of user data like user data that maybe the user wants more private or they they it's too expensive for them to like move that to Anthropic or in a data warehouse where Anthropic could access it as easily. You're starting to see them offer routers and charge a percentage fee on tokens and the routers won't route all the tokens to Anthropic. Of course, they'll route them to cheaper open source models. And so you're finding out that the monetization of this again looks like crypto. There's almost this like payment for order flow type of thing. Like I generate order flow from my users who

are using my SAS app. Some of the tokens I route to anthropic, some I route to to Chinese models and I can take a little bit of a spread on that. >> Well, there's a very interesting idea actually by Brett Harrison. and I think he posted this uh yesterday that the payment for order flow in token routing probably actually looks like hedge funds that will pay you for your LM queries just to get real-time information into what's happening in the economy. >> Yeah, for sure for sure. For sure. I I I I think like that is already happening a little bit but not not at like the systematic level of hedge funds but more in the level of the the inference providers are off offer deals to people who send them tokens like if you guarantee me two billions tokens I'll give you a cheaper price which is like

already moving in this direction but I think the uh the interesting thing about this market is it really really looks like crypto except decentralization isn't the most important thing but like comp open competition is right. So it's like you still have the centralized entity in open router like deciding which providers are added. It's not like anyone can join and leave but the open competition is the thing that's actually been driving the price down a lot. And I think this is like you know if this works it sort of says like every SAS company is going to be a router because like it's the easiest way for them to charge a margin on top of like owning user data and user flow. And I think to me this is going to be the the most

exciting thing in AI that that kind of like looks like crypto. It's like it to me it's like the one thing that's actually kind of a real place where they they might intersect. So anyway, I know I went on a rant for 10 months. >> Tom, what's your take? >> Yeah, I we've been jamming on that this a lot this week. I mean, also just like you talking about, hey, like offering, you know, discounts for ZDR. I think um it is feel like you kind of keep relearning the same lessons around like you know statefulness and stickiness being like the mo and thing that is like monetizable and you know I could see a world you know down the line where even the frontier labs end up you know maybe using thirdparty models and they just end up having so much stickiness from their existing user base and you already

see that a little bit right people will like yeah I love cloud code as a harness but like I want to swap out the back end and use Kimmy or something. I think the question is almost more on the models where you actually see again some some crypto analoges um Kimmy K3 when they when they when they publish the new license it's very bslike where it's open weights you can use it for personal use but if you're a company with over you know 20 million revenue you have to pay us a licensing fee and here's the fixed price um and so you can't you don't just get this get to use this for free which was also some of the original concerns about um you know open source DeFi which is you know why would someone not just pick up unis swap and and and copy paste

it and and make their own version, you know, how does this not just get absorbed by sort of both ends end ends of the stack? And like this is the answer so far, but it seems kind of brittle. It relies on, you know, the legal system to kind of adjudicate this. And we're al also discussing like you open RER also, you know, charges a listing fee for infant providers to get added. So like there's all sort of like weird interesting kind of kind of analoges um you know across both. >> So much like much like Binance and the token market what's interesting Yeah. Yeah. Yeah. I mean the the market structures are very similar in it being you know this being kind of this open worldwide competition for this service. I do think what we are going to see more so you know Darun what

you pointed out in your paper was that there's incentives to cheat by nominally lowering your price but not actually lowering your price like you sort of cheat on the back end to have like a lower sticker price but actually higher all-in price or the same all-in price or what you know whatever. uh there's a one can imagine so you know there was these stories about these data brokers that would give you illegal access to claude in China or you know overseas in countries that that claude is not allowed and these data brokers you know usually what they're doing >> arrested recently for like running one of these services actually >> oh interesting >> yeah students >> right so what a lot of them do is they are they're either one they're stealing accounts so they're uh aggregating like a bunch of account takeovers and just

jumping around between subscriptions and selling those to third parties. But another thing that a lot of them do actually besides lie, you know, some of them are just like that's not actually clawed. But what a lot of them will do is that they will give you highly subsidized prices because they are scraping the contents and they're selling them. And usually they're selling them in large part because there's a lot of secrets in them. So there are API keys, there's PII, there's, you know, all sorts of stuff that literally they will just come in, grab a bunch of the secrets that you have and then like drain your accounts. Um, so the one can imagine that the more of this kind of thing you do, we talked about hedge funds doing it to just have information, but you can imagine, you know, more nefarious ways of monetizing

the data that you're getting and that creates a lot of the competition of, you know, having lower and lower prices because you're making it up on the back end doing wier and wier things. So you're talking about a malicious version of this, but there's already kind of a non-malicious version of this in that thinking machines on open router and like just generally if you use their models and you don't use zero data retention like they delete all your chats and and prompts uh they give you a cheaper price because they're basically trying to generate Meror you know like these companies like Merkor Surge Scale.ai AI who like do all this human annotation very expensive right because you're you it's hard to to to to grow off that and you're starting to see

people basically do the 20 do you remember the 2017 token dream of like okay get paid for your data somehow use a token to do that whatever this is actually the first time I actually think that stuff will it might actually work in this trace model where you basically give rebates or cheaper inference which could be done via token model like a Venice type of thing or or not, but like some type of rebate to a user if they give you their data for their chats for your post training or for later usage. And like again, this is like these crypto dreams that were just like too early. They were like it just like didn't make sense at the granularity because like your data was sitting on this platform like on Facebook or in Google and like you couldn't like pull

it out. But the AI now makes you gen makes can kind of streams out your data for you. And so now you can actually resell it and like actually get a cheaper cost of of of using AI. And the other thing that's really worth noting here is that there's this huge rush of like people making data companies, right? Like the mers and surges and whatever because you know all the labs are talking about how they're spent anthropic said they're going to spend 10 billion on data like payments this year. And so there's this kind of interesting thing in the AI world which is very much like crypto where there's the cost side like training data collection and then there's the huge margin side inference right like inference is like 90 plus% margins like like in a lot of cases

people are basically running away in terms of like the the the financing cost right like all of the people doing the openweight models like the Chinese companies they're all of their revenue comes from just this in being one of the many inference providers in this marketplace and So, I think the the the the the crypto stuff like there's a sense in which we should be revisiting a lot of ideas from 2017 right now. Like like there there are a lot of them that didn't that just sounded sci-fi, didn't make sense, were too hard to do that actually might like the token models for the crypto tokens for making it easy to monetize these inference tokens. I think there's a there's actually going to be stuff there and that that might be our next real bull market.

So, speaking of bull markets, we've got we've got a couple minutes left before we wrap. >> I'm only my bully pull pit today. >> No, it's good. It's good. It's good. >> Given given that you guys are in a very different environment out there in Asia right now in in in Bhutan, what's the vibe that you guys are getting? Are we in a bull market? Has it truly begun? Is this just a bit of a rebound? Like, what's what's the what's the vibe check just quickly from from the the the from from us here at the table? Weirdly, I say most of the the the teams I I've been talking to are still very much in this kind of button up tradi RWA meta that we've been in for maybe the past two years of we're going to tokenize some you know we're invoices we're going

to do invoice factoring FX and so it's almost like oh like the gear shifted again and you're still like you know doing like STOs or something and you know I think there's those are good markets and I think there probably plenty of good good projects but it feels like a lot of people were caught off guard with the shift. It reminded me a little bit of uh you know when like you know Larry think went on TV and said Bitcoin is good and you know the ETFs were getting approved and everyone's like oh like where did that come from? It feels like we're kind of in in a little bit that gear again too. >> True. Same same read. I think the interesting thing to me is like I think of like you know Asia as the the true source of like K gambling punter you

know like trading token 199 in market cap as if it with their entire life savings right but the funny thing is everyone seems like less den than the US like memecoin stuff the US memecoin stuff is actually very crazy like like there's something weird about that that like it hasn't translated here and I I That's the thing I'm I don't know if that means there's like another bull cycle from or like mini cycle for that or not, but like you know, you were saying that earlier, right, too, right? Like we we they like they haven't had the like FOMO pumped up fund stuff as much and like >> right >> there's a lot of people trying to figure out how to like Asianize that >> Asianize u [laughter] volumes are more volatile.

>> Make local more local friendly, right? The social training stuff is like very western >> word is localized. >> I know I know I was just >> anyway localiz localized FOMO like that I think there's a lot of >> like people want to bring and like that maybe that is like a >> Binance life is doing really well. Binance life is uh is is rallying. That's like the one the one Chinese characters on page one of coin market capab now is is Binance life and memecoin. So, um, yeah, the there's it's the trenches are back to life. Uh, we are up on time, so we got a wrap, but thanks everybody. Welcome back to crypto being cool again, or at least starting to get the green shoots of being cool.

Um, hot again. Hot again. Hot again. I think the cool the coolness and the hotness are very correlated. I will say that. >> Okay. And the longer that Tun and Tom wear those robes, the cooler and hotter it's all going to get. >> The [snorts] cooler and hotter it's getting. I will say that we're wearing our normal laundry layers. Yeah. [laughter] Okay. All right. Nice. All right. Well, enjoy Bhutan you guys and we will be back next week. >> Yeah. >> Everyone [music] [music]