Loading summary
Transcript106 lines
- [00:00]
A
I think it was a lesson. I think the market taught him a lesson. I think he's internalized a lesson. You do not get to sell. There's no sell button for you. You're buy only, not a dividend.
- [00:10]
B
It's a tale of two Kwan.
- [00:11]
A
Now your losses are on someone else's balance sheet.
- [00:14]
C
Generally speaking, airdrops are kind of pointless anyways.
- [00:16]
A
Unnamed trading firms who are very involved.
- [00:19]
C
Alec Eth is the ultimate defi.
- [00:22]
A
Protocols are the antidote to this problem. Hello, everybody. Welcome to Chopping Block I. Every couple of weeks, the four of us get together and give the industry insider's perspective on the crypto topics of the day. Quick intro is first. You got Tom the defi maven and master of memes.
- [00:35]
C
Hello, everyone.
- [00:37]
A
Joining us today we've got Ilya, protocol pioneer and patron saint at near.
- [00:42]
B
Hey. Hey.
- [00:44]
A
Good to see you, Ilya. And joining us today we've got another repeat guest, Mert, the main manlet at Helios. Welcome back, Mert. We heard that you were five beers in. Is that when you started speaking Russian?
- [00:58]
D
That is actually generally when it starts, yes.
- [01:01]
B
That's how everyone speaks Russian. It's just everybody's four, five years in.
- [01:07]
A
Beautiful, beautiful. All right, well, I must see you. That hype man of Dragonfly. We're early stage investors in crypto, but 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 if you're all wondering the reason why Mert has been drinking is because there's been a market where meltdown. It's been an absolutely insane weekend. We saw massive amounts of liquidations coming in. Bitcoin going below 58k or, sorry, 59k, which was kind of the, the, you know, people kind of thought 60k. Okay. This is like the mental guard rail that if bitcoin goes below here, it's all over. Now. Since the weekend has transpired, there's been a. There's been a lot of other stories. And we're going to talk about what happened with Z, which was one of the most tumultuous stories. But general market, there's been bloodshed across the board. And since then it seems the market has rebounded. So, you know, Friday, I think a lot of this was initiated by both the combination of Saylor selling the 32 Bitcoin, which was a very small sale, it was like three point something million dollars of bitcoin that he sold. That was enough to cause the market to go into a tailspin. And then that got exacerbated. By Friday, NASDAQ was down massively because of a jobs report and anticipation that the Fed may end up cutting this year or sorry, raising rates instead of cutting rates. So now it looks like Polymarket's pricing roughly 50, 50 that the Fed ends up raising rates this year. And if they don't raise rates, very unlikely that they cut. More likely they end up holding rates flat. So not a good environment for risk assets. It seems like AI is the only thing that can save us at this point. But the market on the whole, very jittery, although we're now bouncing back a little bit. The coin's back above 62k. It's been many people calling that, hey, maybe this is the beginning of the end. Maybe the death spirals are beginning. Maybe the, you know, the reaper of dats are finally coming for us and we're going to have to pay for all of our sins. And of course there's now increasing worry about Stretch. We've talked before about Stretch on the show. Stretch are these preferred shares that Michael Saylor has issued. And Stretch, the thing that's notable about it is that it's designed to trade around $100. So $100, you know, probably the ceiling what it should trade at. But it's kind of you can sort of think of as an upward peg where the idea is that he's going to increase the dividends until this thing trades at $100. This thing is trading below $100. It went all the way down to like 95, 94. Today it's trading somewhere around 97, 98. And the reason why it would trade there is either one, people don't believe he's going to pay the dividends or two, they're worried that MicroStrategy might go kaput. These are both very worrisome situations. Mert Ilya, both of you have been participants in different ways in the DAT mania that you know, kind of beset the industry of the last couple years. What do you think? Is this a sign that this is the beginning of the end? Do you think that's going to be okay? What's your guys read on the market?
- [03:54]
D
To be clear, I never did do a dad, although I did be. I was offered quite a few of them. But I am proud of not.
- [04:01]
A
You weren't like advisor. You must have been like an advisor or something.
- [04:03]
D
No, no, no, no. I literally everything. Yes.
- [04:07]
A
Oh wow.
- [04:08]
D
I was going to do. And then one of my group chats was like, dude, what are you doing? And I was like, okay, fine, I won't do it. Why not?
- [04:16]
C
Do you. Do you look down upon the DAT people now? Do you think you're. You're superior to the dats?
- [04:21]
D
I think they actually still make sense, but you want to do them at the bottom and not the top, as obvious as that sounds. So, like, I think the zcash Dat, like the cypherpunk one, that's what it's called. I didn't do that one, but that one is actually above water. They're profitable. That trade worked, at least so far. But yeah, obviously people did it. When Tom Lee was saying that Ethereum was gonna go to like 180,000, which that's probably not the best recipe for a good trade. But anyway, I didn't do one. I don't plan to do any as far as my read on the market. I mean. Yeah, that's. That's a complex question. So not sure where to start really. But I would say the. So maybe focusing on Saylor. I think Saylor is kind of a red herring. This is my theory. I think he is super fun to troll on Twitter and I think most of us get our news from Twitter or at least some portion of us that are relatively unserious, like me. And he's usually always the main character on Twitter and he has been for a while, right? Twitter always has the main character and I think that's been sailor for some time now. I think the worst effect of Saylor, though, I think is something that I noticed a few days ago, which is that he has this distortion fuel just due to the amount of bitcoin he holds and how many people are watching him do things that he changes the conversation from the cool parts of bitcoin and crypto, in my view. So the adoption and the philosophy and the tech and all that is discarded for the sole purpose of, okay, did he sell? How much did he buy? Like, it's just like a trillion people pocket watching one guy, which I think is not a very exciting thing, right? Especially as AI is making all these advancements. And so now everybody, like the biggest asset, the star of the industry is literally just. We're talking about the buys and sells of one fucking guy on the Internet. And I think that's like actually the biggest effects. That's probably a contrary for now, but I think, you know, do you think also ETFs.
- [06:33]
A
I mean, it's the same story as the people say, oh, BlackRock bought that. Which obviously it's not BlackRock. It's people who are buying the ETF and then, you know, the authorized participants who are actually buying the shares.
- [06:42]
B
The bull market of last year was kind of to big extent, microstrategy buying, right? I mean, yes, ETFs also enabling inflows, but kind of, I think. I mean, bull market for Bitcoin, right? And so I think big part of this was almost like in crypto, there is like a paradigm of one thing that worked, will always work until it completely fails and everybody freaks out. And so it's kind of like people just fell into the same pattern. It's like, oh, okay, well, now MicroStrategy is our driver. And so we need to. Only MicroStrategy can drive us forward. Right?
- [07:20]
D
So I would.
- [07:21]
A
So here's my. Hold on. Here's my. Here's my counterpoint here, right? Is that, like, much. So the point I just made about blackrock, right? Which is that people, the tweets are always like, oh, you know, Larry Fink bought a bunch of bit. Sold a bunch of Bitcoin, obviously, is not what's happening, right? This is like people who are buying the ETF and the authorized participants are the ones actually, you know, moving the. Moving the bitcoin around. So it's people on exchanges who are buying ibit, and that's resulting in netflows or inflows or outflows.
- [07:43]
D
Right.
- [07:44]
A
But it's really retail. The same thing is true for micro strategy. Micro strategy, it's like, okay, Sailor's buying. How is Sailor buying? He's issuing stock to retail. Retail's buying the shares and implicitly getting exposure to Bitcoin. It's still retail or, you know, to the extent that institutions or whatever, buying it.
- [08:00]
B
But it's not ETF is you managing somebody else's money. This is like, this is equity to asset conversion. I mean, that. That was the whole point of that. I mean, micro to never sell.
- [08:11]
A
Yeah. It's basically a trust without the ability to redeem. Right, Exactly. It's like kind of a. Yeah, but it's effective because it's effectively trust. It's not like mubadal is buying. It's not like, you know, a sovereign wealth fund is buying. That would actually be okay. This is some guy's money who wouldn't otherwise be buying. This is displacing bitcoin buys from retail.
- [08:29]
D
But you see the point here, right? Like, even with ETFs, it's the same thing. We're just talking about whether people are buying or selling and like, if it's a risk asset, which, look, it is a risk asset, what you want to sell, at least what I would want to buy is some future vision of what I'm actually buying into rather than just boing, boing fucking candlesticks on a chart. Like, I don't care about that. I care about, okay, what does this enable? Like hyperliquid has this right where you're using hyper liquid to do price discovery on pre IPO things which, and then the NASDAQ literally shows it on the, on the, on the, on the, on the trading floor. And that's an interesting thing because then you can start to imagine, you can start telling the story that like, okay, this, if it works, can start doing such and such for finance and it might be a good time to get in early. So if the conversation is literally just, well, how much are they going to buy? How much are they going to sell? That's not a super exciting thing to invest in, in my view, which is a shame because bitcoin is super interesting still. But like, I think that's actually the effect of Saylor. Like obviously he still has an effect on the price and all these different things. That's, that's a given and I think that's been exhausted to death in the discourse. It clearly is happening. But like there needs to be like a, like a good storyteller in crypto that's like, okay, enough about sailor. How about we get back to like, I don't know, like even the FIAT is a scam type of guys, like just tell a different story.
- [09:53]
A
Andreas Antonopoulos, Is that what you're talking about?
- [09:56]
D
Exactly.
- [09:58]
B
The gold, the gold will be found in Mars Belt and, and they're going to be explosion of amount of gold. And bitcoin is a fixed, fixed supply man.
- [10:07]
A
Very apropos of the SpaceX IPO. I like that. I, so I, I hear what you're saying. I, I don't totally believe you, Mert, that like you really find bitcoin super fascinating. Like, I, like, I've been here for a decade now.
- [10:19]
D
Encrypted is pretty fast.
- [10:20]
A
I don't really want to talk about bitcoin anymore. I have to say, like, I'm kind of like, I've heard everything there to say about bitcoin. I want to see the price go up. You're proving my point. I find that extremely interesting.
- [10:29]
D
I think you're probably my point though. There is obviously other stuff that you can do with Internet currencies other than just buy and then Sell and then talk about how some other guy is buying and selling. And to be clear, Saylor isn't like, if Saylor was a good storyteller, I think it'd be quite different. But all he does is literally post complete AI slop. Do you remember that ad? Maybe you can overlay it on the podcast Replay, where it's like some girl sipping some fucking margarita talking about, oh, I just bought Stretch. It's like, dude, what? Like, are you serious? How is that? What are you doing?
- [11:04]
A
Anyway, Tom, what's your. What's your take here?
- [11:07]
C
I mean, I agree in the sense of I always wonder, like, I've never been to Bitcoin 2025 or the conference, but I'm always just like, what. It seems like it's more of like a celebrity, you know, lineup at this point where I'm like, what. What is actually interesting to talk about in terms of. Of bitcoin? Because I agree there isn't really anything going interesting on Bitcoin L2s or lightning or ordinals. There's really nothing other than price go up. And so I agree, like, hey, the ETFs are more or less the same as the DATs. And I think it is kind of a deterrent to new people entering the industry when it's like, what if there's a thing that we're all talking about, it's this single crazy dude who can sort of more or less arbitrarily buy and sell bitcoin. Like, I, I think it's, it's. It does kind of suck the air out of the room.
- [11:52]
B
Well, I think, I mean, the. Also the strange. The. The strange thing that happened, right, is like the, the generally selling those 32 right. Bitcoins, which, like, doesn't really solve his dividends problem nor. And broke the whole point that, like, hey, I'm not never going to sell, only buy. Right? That's. I mean, that's what kind of led to the. Because I think they authorized what, like 2 billion worth of sales.
- [12:18]
A
So they have signaled. So Saylor has signaled previously that he is going to be testing the market by doing small sales, but that he's still going to be a net buyer of bitcoin. He said that previously and I think this was clearly like the test transaction just to credibly show the market, hey, just so you know, like, I made some money.
- [12:35]
D
He also said you never sell bitcoin and to keep your kidney, but he clearly has run out of kidneys.
- [12:40]
B
Well, exactly. Yeah. That's the thing. Like, yeah, like this also doesn't make any sense. That was the other part. Like, I think why market reacted was like.
- [12:48]
A
I think it was a lesson. I think the market taught him a lesson. I think he's internalized a lesson. You do not get to sell. Sell it. There's no sell button for you. You're buy your buy only. And I think you'll never try that again. That's my guess. I mean, I could be proven wrong, but obviously if you're selling a little bit and mostly, or maybe he was
- [13:04]
B
on hyper liquid opening up a major short and then that that's how he makes money for dividends.
- [13:12]
A
Possible. That's possible. Okay, maybe he's swing trading Bitcoin. Okay. No, it's. I mean, it's funny you bring all this up because one of the other stories that happened on Friday was there was a bug discovered in zcash and that bug was caused such a panic throughout the market, obviously on a day that other things were also otherwise drawing down. That ZEC came down more than 50% peak to trough over the course of a single day. So one of the largest drawdowns in market cap for any single token in a long, long time. So the story is that this was actually a planned audit by Taylor Hornby, who's at Shielded Labs. And he was using Opus 4.8, the latest quad model. And he had his own harness that he had built to analyze zcash. And they were looking at the Orchard ZK circuit, which is basically how zcash ensures privacy within the shielded pool, which is what happens when you want to make your zcash private. And he discovered a bug in the ZK circuits that would allow him to basically mint unlimited counterfeit zcash inside the shielded pool. This would not be detectable. So you almost think of this like a bank run where he could mint more zcash in the shielded pool and then withdraw enough to basically make it impossible for anybody who had normal real zcash in the shielded pool to be able to withdraw. The shielded pool today is about 30% of the total supply. So you could have minted, in principle, 30, 30% of the total supply of Zcash pulled it out and rugged all the people who are currently then shielded. Now, this was a security researcher. He had been paid by Shielded Labs to do this. And so he disclosed it. There was an emergency soft fork that was done to disable those transactions and fix the circuit and move everything over. And then it was roughly four days later after this took place, all this stuff was done through open source patching and you know, all this stuff. Roughly four days later, there was a disclosure blog post that was written by Zuko. And in this disclosure he made very clear, technically, we have no proof that this bug was not exploited. Now, we don't believe it was exploited. There's no evidence that it was exploited. But the only way to know is to do another upgrade, which will probably be in roughly a month or so. And in that upgrade, we will force a migration to a new shielded pool. And that will almost like taking headcount at the end of a field trip, that will ensure that no extra coins snuck onto the bus. So then we will be absolutely certain that there was no inflation or not inflation rather, but sort of counterfeit coins that were minted and that would result in a theft from the shielded pool. But that's the situation on that news. Zcash dumped horrifically and just absolute panic and pandemonium in the Zcash market. It went down from 600 something all the way down to $250 now. It's actually recovered quite a bit, but big rebound of more than 40% off the lows. It's in the mid-400s right now as I'm speaking today. So wanted to get. Mert, you are Mr. Zcash, among other things, talk to us about the Zcash bug and kind of how it feels in the aftermath now that the fog of war has lifted a little bit.
- [16:16]
D
Sure. Well, you covered it pretty well there. But fundamentally what happened was they. So the team, basically, not just this one team, but actually several teams within zcash. The core teams started red teaming all the circuits and all the code bases on zcash, like pretty aggressively, completely ramping up all resources to do this. There's several audits in parallel at any given time, including formal verification for Orchard, but also the upcoming tachyon pool and Quantum. And it's just there's a whole influx of resources and capital. Right. And yeah, so the tailor who, by the way, people are saying, oh, well, he. This guy just said, claude, find a bug in zcash or something. That's not what happened. Taylor is probably a top five person who understands zcash and how it works on the planet and has been doing it for many, many, many years. And this had passed both his but also other people's audits in the past. Right. Some of the top cryptographers in the world, like the circuits that zk zcash has, are used in other ecosystems all over. So it's a lot of People looked at it. It's kind of my point. So one hacker to have found it just by Vibe coding would be, in my view, extremely unlikely. And if you don't believe me, you can just try it, right? You can try dissecting the arithmetic of a ZK circuit and let me know how much sense it makes to you. Anyway, so what was actually interesting is it was a wave of different FUD news. So the first one was actually that zcash was down like the network, because for the first two, three days I had to just. There was all these news articles like, oh, zcash stopped producing blocks, which never happened. And what had actually happened was due to the fork, some of the RPC nodes, notably on the Explorers, hadn't all updated. And people look at the Explorer to determine the source of truth for a chain, which. Not the best idea, but I mean, I get it, I guess, but like all you had to do was do
- [18:17]
B
a transaction somebody was running.
- [18:18]
A
Isn't that your business? Yeah, yeah, yeah.
- [18:22]
D
I was like, you know, when that was happening, I was like, I should just. Because I also run an explorer on Salon, I should just like take it down and just see if people think Salon is down or something. Like, this is just crazy. Anyway, but it wasn't down. Anyway, we're battling with that. And by the way, the news of the bug and exactly what it does was disclosed by Josh three days earlier than Zuko. But then, so you combined the news of the day, which is that zcash is down, combined with literally every single asset nuking, including like qqq, and then you start getting a few people panicking about, oh, wait, I didn't know this was a possible thing. Right. So there's two classes of people on zcash, really. There's the first class who maybe are somewhat familiar with zcash and zk, and they understand that the trade off in a privacy protocol is that the money is private. And so there are some trade offs in audibility of supply. And this is a conversation I've had on Twitter transparently at least six times in the past year. So I was like, this is not news. I completely thought it was a nothing. And then obviously that was not the case. And people just started, just one post after another. People started panicking like, oh, was it exploited? And all this stuff. And I was like, oh my God, what is happening? I did not. Anyway, you know, you start getting the
- [19:38]
B
best FUD was that somebody exploited it and then was paying Kols to pump it. That was like,
- [19:47]
D
that was the One where I was like, holy shit, these people are completely retarded.
- [19:51]
A
Right now I need a full audit of Mertz finances, get to the bottom of this.
- [19:56]
D
Some people actually thought you could mince infinite supply of all zcash, not just orchards, but like you could just keep selling on binance, infinite amounts. And I'm like, okay, so how is the shielded pool going up and the price going up in that case? If there's infinite sell pressure, right? Like there's just so many different things wrong with it anyway. Yeah, so that happened. And I mean, again, most people, and there's a pie market for this now as well, if you have an opinion, would put the probability of an exploit to virtually very minuscule single digits, if that. I would put even less than that, but I'm obviously a biased party here anyway, so. But, but here's the bullish stuff, right? So to get around that, what's going to happen is they're going to roll out a new shielded pool. And so the headline here is that once you do the migration from orchard to Ironwood, which is a fantastic name for, for many reasons, once you do that migration and you can verify that the turnstile doesn't trigger, which means there was no counterfeiting, that that is the headliner. But the actual good news, I think, is that it will be formally verified as well, such that that same class of bugs will be much less likely to occur. And it will also be quantum recoverable in that pool. And even though the update will go live in early to mid July, it seems the formal verification will be done before that. And so we'll have pretty good certainty. And Obviously it's not 0 or 100% or something like this. These are just probabilities you will know once it's been formally verified, which is being done by three separate teams, by the way. So that's also pretty promising. That same class of bugs doesn't exist in the pool that you will get once that happens. Of course, there's also a Project Tachyon, which is how I got into zcash, which again will be formally verified, but it'll also use a completely different circuit, arithmetic, which will reduce another class of bugs without getting too technical. And it'll be as close to mathematically perfect as you can get a ZK protocol. And I think personally, in the age of AI, that's just going to be table stakes at this point. And we can do all sorts of different discussions on this. But the fundamental trade off you get in ZK versus DeFi, for example, is is just time to detection or the detectability of it. Right, and why do you care about that? Well, you care about detectability because in theory you can stop the exploit from happening or you can roll something back or something along these lines. In practice though, what you've seen in DEFI in the past year is that the exploit happens and the hackers get away regardless. Right. And so detectability is obviously quite important, but it still does happen in terms of the exploit. And so it's non negotiable that you need to get the formal verification and have just think about risk differently. And that is what's zcash. And actually Ilya has some stuff on formal verification with near soon as well. And that's what's going to happen.
- [22:57]
A
So I should also caveat that Dragonfly, we're invested in Zack. I assume you own your balls. Long Zack is my impression, although I don't think you've ever said it specifically.
- [23:07]
B
Ironwood Long.
- [23:09]
A
Iron, yes, exactly. He's got a lot of iron. Ilya, I want to bring you in here because of course a lot of the story hinges on this idea that this was an AI assisted or AI driven bug that was discovered not using purely human eyes. And this was a repeatedly audited piece of code, obviously very central to the zcash security model. And in the report that the security analyst created, one of the reasons, I think it was specifically a sentence that said, look, this was mostly Claude, obviously it was guided using the wisdom and intelligence and, you know, so I actually had a friend of mine who tried to see if they could use Opus 4.8 to rediscover the bug by just hinting there's a soundness bug. There's a soundness bug. It whirred around for 30 minutes, didn't find anything. It's like, no, no, no, keep looking, keep looking. And then it's like, oh, I think I found something. It's like, aha, I found it. And then it couldn't figure it out, didn't find the bug. So it does seem like to Mert's point, there is some. It reminds me a little bit of what we saw with Mythos. Yeah, well, it is a skill issue. It's like a little bit of the delta between mythos and 4.8 is that if you tell opus 4.8, there's a bug right here, this is where the bug is. 4.8 can figure it out. But Mythos maybe is intelligent enough that it can just go all the way through the entire search tree and specifically Find that bug. How do you think about this and what does this spell for what happens when AIs do get much more powerful? Sort of Mytho style.
- [24:31]
B
Yeah. So I mean, I wrote a blog post actually about over a year ago, effectively spelling out that this is going to happen and the only solution is for modification, right? Effectively, it doesn't matter. The asymmetry that happens between attack and defense is just always going to be on attacker side. Because yes, you're using Opus, or let's say you even have Mythos, right? But everybody else have five different models. They have like more time to spend on attacking it, et cetera. You have potentially thousands of people who are trying to attack you with different models with different ensembles. And indeed, I actually think Mises was more. I mean it was a marketing ploy on anthropic side. Let's be clear what it was. Right? Because the reality is if you're a security expert with these models, indeed now you can find vulnerabilities and it doesn't matter which are you using chat 5, 5 opus for like 6, 7, 8. We saw all of the hacks starting from November. Why DeFi started getting hacked in November? October. Well, the opus came out, the Opus 4 came out and we could see effectively protocol after protocol, everything that was multiple times reviewed and audited started to crumble, right? And then we kind of saw and it's like, I'm assuming also like a lot of open source models kind of pulled up in like in April. And that's why we saw a bunch of like more hacks, is because people effectively just figured out I can actually probably just use Deep SEQ or whatever on my own GPUs and not tell the inference provider that I'm hacking something. And so I think generally like we're already in the world where if you know how to prompt them, if you build the right like loop that is doing this, if you have a good way to assess things kind of automatically that this is indeed exploit, right? So the way effectively if somebody wants to reproduce ZK bugs and not step by step instruction, but effectively you build a loop that exercises the actual mint, right? So you have a loop hot in the environment and you're able to. You're not just running code analysis, right? You're actually running an agent that is able to instrument injection and see that something came out. And so then it actually goes and starts figuring out how to do this because it has a validation framework that it can do this. You can literally run reinforcement learning on top of this to Figure out how to do it.
- [27:05]
A
So, Ilya, we've talked about formal verification in the past as being a potential solve for some of these issues. Give us the intuition, because I think a lot of people in the audience don't necessarily know what it is. They've heard the concept. But why is this to solve? And how does AI interplay with that?
- [27:21]
B
Yeah, exactly. So what is formification? Formification, let's say there's a program and you're effectively saying like, hey, whatever this program does, I say that it's not going to change these types of properties about my state or inputs. So the simplest example, let's say you have sorting, you're going to sort an array, like I have a list of numbers, you want to return sorted list. And so you want to say, hey, you will effectively get same superset of items and they're all going to be in a sorted order.
- [27:54]
A
Right?
- [27:54]
B
Now, this property you can use in other places for crypto. You want to say, hey, total supply is not going to change, right? You deposit money, any arbitrary set of actions happens, I withdraw money, total supply didn't change. Right. And so those are types of properties that you want to prove. Now, historically, the code that was proven is code that was sent to space because you cannot patch Mars Rover very well. But it was very expensive. It costs hundreds of dollars per line of code to prove it because it was manually mathematicians writing out all of those things. And because it was so manual, it was also error prone, not from the proving perspective, but what you prove. If you misspecified what you're proving, you forgot, oh, actually it also should not just total supply didn't change, but also everybody else's balances didn't change either. When you withdraw money, it could have left that bug in the code. So it was like specification issues were also a big issue with formification because it's slow expensive. Now with AI, we can actually run AI to do verification, to do proving itself. So now you give it code, you give it what you want to prove and it can go and run, effectively run its mathematical genius 140 IQ based on all of the benchmarks and actually produce a proof. Now, it's not that easy. The proofs are very long, so you need to split it. It needs to have good hierarchical thinking. But the AIs have gotten really good at doing this. And so now again, it's all about costs and effectively ability to put this into a loop so you can actually develop code and validate properties as you go.
- [29:40]
A
Tom, what's your Take. Do you think this is the answer?
- [29:43]
C
I mean, in some ways it reminds me a little bit of what we talked about. I think also on the last show with this sort of crowdsourced approach to developing a proof for the Google Quantum computer ZKP circuit, or the sort of existence proof for this, the sort of faster Shor's implementation of like, yeah, you define sort of the exit criteria or like you said, this sort of validation framework, and then you just throw enough tokens at it and eventually it sort of comes out correct. And I kind of think about again, like the software development approach of like, you know, maybe the past 30, 40 years have been almost kind of like building a boat. It's like you expect like a boat's going to have. It's going to take on some water and there's like a pump and like, okay, you do your best approach, but you know that it's not going to have zero water in it versus to the space point, like building a spaceship. You can't build a spaceship that has some holes in it. It has to be completely sealed or it's fucked. And I think we're kind of moving to a more spaceship type approach to code, where it has to be completely airtight, but not roughly airtight or someone out there is going to find it.
- [30:41]
A
Yeah, I mean, we all know this, obviously, but crypto blockchains, smart contracts, are the most treacherous environment for code because you only need to get exploited once. You can't really take stuff offline. You can't really say, oh, somebody's about to exploit me, I'm going to yank this away and make sure that it doesn't get exploited. And so it's almost like being vulnerable basically means that you are hacked. The moment that you know you're vulnerable, you're hacked. And so it is a very difficult environment in which to learn these lessons. Formal verification maybe is the bulwark against the onslaught that is coming. Do you guys think that crypto and defi are going to make it through this? Or do you think we're going through this? Herd immunity, there's going to be a bunch of old people dying and it's going to be a COVID lockdown situation. A couple weeks ago, the Open Zeppelin founder was saying, look, I'm telling everybody I know, get your money out of defi. Where do you guys land on that?
- [31:37]
B
I think with formification. So formification helps you from the pure code issues, right? So like ZK circuit, underspecified bugging, arithmetic, left admin, being able to Withdraw everything. But there's another aspect, right, which is oracles, which is whatever, multisigs, et cetera. Those are things that we need the OPSEC to be on a different level. We need to make sure that there's guardrails around those inputs and things like that. And so those are. I think the rate of those being set up properly have declined massively over like past few years as
- [32:24]
C
well.
- [32:26]
B
Until a few months ago. I mean, over like 23, 24, 25, I would say generally the OPSEC, because like 2017, 2018, you know, things like multisig hack, like all of those things, people kind of gotten a lot more. Oh, let's actually set up. And then the new wave of people came in and kind of like, oh, we can just launch stuff and meme coin here and token contract there. Like, oh, admin key stored on my whatever in text file. Let's go. Right. So I think we kind of relearning a lot of this as a space, but my broader point is actually as a space, we kind of need to get out of this code is law, kind of absolute code is law until the code is wrong. Right. And then we actually like want real law, right. Money stolen. So because of near intents, right, we effectively everybody's trying to launder money who's stolen money through near intents. And so we've been building a system to actually detect hacks on other chains. Because otherwise, for example, with drift, I'm getting signals from Merge saying like, hey, shut it down. It's like war rooms everywhere. And effectively all April everybody was in the war room figuring out who get hacked on what chain. So a lot of the work has been is actually doing almost like predictive upset on other protocols and other chains. If we can detect the litecoin, they're like, hey, it's not the problem that we got hacked twice. Like the chain forked around it. It's all good, right? The litecoin thing where effectively they actually had very similar bug that was exploited in mimblewimble that you could effectively withdraw money without the proof. And then they ddos the rest of the miners who actually upgraded already. And so effectively detecting this kind of outlier situations, you effectively need to build a system that's not completely transactional, right? So this is like the tradfi learned this. It's just like the way they implemented it is a very wrong way, right? They're like, okay, we're going to wait three days, we're going to have them, your banker calling you, we need kyc but the thing is, if you don't have a fully atomic transaction, which with intents we don't, we actually have request response. We have a way to actually run AI analytics directly on the interaction. We can check like, hey, this deposit coming in from, let's say, you know, Sui is coming from an account that didn't have any money and then something happened and now has ton of money. Right? Like where, where did that come from?
- [35:05]
A
So this is pure anomaly detection. You're not like scraping Twitter to see if people are talking about a hack.
- [35:10]
B
So, so, so I mean that, that is also. But like that's too slow. Like you actually like you need the on chain data first. On chain data, block data, block level data as well. Right? Like if block is weird. Like let's, let's wait longer because normally like you want as fast as possible. Like in normal situations you want to. So I mean it is in a way the same things, you know, like JP Morgan is running on your credit card when you swipe, right? Like they reduce massively their fraud effectively by just running like good ML outlier detection. So things like that, like running things more like serious adult financial systems, not just, yeah, we throw smart contracts and you use smart contracts at your fault. Right? That's kind of been the crypto's approach generally versus like no, this is actually like a true financial product. Let's get everyone good service.
- [36:09]
A
So Mert, you've been on the cypherpunk drumbeat and yet you're also in Ilya's DMs being like, yo, yo, shut it down. How do you resolve that tension? Where do you land there philosophically?
- [36:21]
D
Oh, I mean it's, I think it's very obvious in that if you have the means to stop crime, you have to stop it. Like that is actually the law as well. Like if you have the ability to do it, then you have to do it. So the solution would be to structure the system in such a way that you don't have that influence. And depending on the life cycle of that project, you might have that, you might not have that. Right. Another example is Arbitrum where there was obviously the North Korea thing and the multisig, the emergency council 9 of 12 had to intervene because they have that means. But for example, if you have an immutable thing where so like tornado cash, for example, tornado cache is immutable. And even in that case, by the way, they didn't stop because they couldn't and the government is still going after them. Right? And so that Obviously does not help the situation. And that's actually been a huge pain in developing privacy stuff on a personal level. But, I mean, the cypherpunk thing is that you shouldn't have that control to begin with, right? But if you do have it and you don't stop it, then you've kind of just virtue signaled superficially, you put up this facade while not actually taking any of the steps to do the cypherpunk thing. Right? So that, that's kind of my. My point on it. But how do you take.
- [37:41]
A
So Thor Chain, Thor Chain has famously claimed that, oh, we have, you know, we have no control. There's a bunch of validators. We don't control them, and therefore we can't stop people, you know, North Korea sending funds through here. Do you take them at their word? Are you like, oh, you're full of shit?
- [37:55]
D
I. I don't have. I haven't done the research enough to have a meaningful opinion. All I know is Zach and Tay seem to think that Thorchain has some
- [38:07]
B
control and maybe very specific. They had a hack with their NPC few weeks ago and they did stop.
- [38:15]
D
Okay, So I don't know any of this. I don't. I honestly just don't look at Thorchain. I'm sorry.
- [38:20]
A
Okay, okay, okay. We can take it back to the previous question.
- [38:24]
D
Yeah, so, like, okay, look, it's very simple. If you have the means to stop it and you don't stop it because you're saying you're a cypherpunk, well, first of all, you're not a cypherpunk because you didn't structure it in an immutable, decentralized way to begin with. But if you do have it, then you have to stop it. And to your earlier question, though, of like, does Defi survive this? Or does crypto survive this whatnot? I think it's a case of something like, there will be many battles that are lost, but fundamentally the war will be won. Which is to say I'm just, I'm long Math, right? So if you just look at the history of computing, I mean, computers are just math, right? Like, it literally came from being able to verify or do like, universal computation, which is like a mathematical field all in a sol from like, Alan Turing and John von Neumann. And so there's been. Obviously, humans are not very good at math. And so there's always leaks in logic and abstraction and a lack of rigor and all these different things. But the trend throughout since the start of computing has been getting the computing Devices better and better and better, whether that's in performance or security or cryptography, in terms of different encryption schemes and all these different things, it's always gotten and AI is a testament to this. That's a fundamentally mathematical construct. It has been getting better and better and better. And so unless it's barred by the laws of physics that you cannot secure a system mathematically such that it will always get exploited, which I don't see any laws of physics that would allow for that. Or like I just don't think that's a fundamental thing, which means there is a solution that's possible, which is why there might be some battles lost because you need to error correct and learn from real life and these things as they happen. But fundamentally the war will be won and that the trend is always new, knowledge gained, errors corrected and systems hardened.
- [40:16]
B
Just maybe to add to the FUNC reconciling. I actually think there is a way to balance these two things, right? Where an example is an AI system that runs within the bounds of the kind of autonomous protocol, right? To be clear, none of the things we're running are non upgradable. If Bitcoin wants, if tomorrow everybody decides to upgrade Bitcoin, they can upgrade it tomorrow. If Ethereum wanted to upgrade Ternata cash they could have, right? You just need a bigger set of people. So I think the question always comes back to what is the governance and decision making of this. And the problem is when we talk about real time kind of detection vulnerabilities and kind of issues like that, it's too slow, right? Any kind of governance like be that 9 out of 12 multisig, they effectively did some judo to effectively do it without the hackers seeing that their accounts will be frozen. But if the hackers had a better monitoring system, they would have moved the money out of the faster effectively that they could have executed this. So you need real time systems and real time systems mean you need the digital intelligence. And so now we have AI systems and we can run them in autonomous way, we can govern them by kind of slower governance processes where we define the rules. We say hey, we don't want this kind of anomalies, we don't want this kind of hacks, we don't want this kind of behavior. And we apply that universally now the cyphering punk way kind of across the protocols at real time. And I think balancing those two things is how we actually get better product because again we go in mass adoption. If you go to normal people, you go to your moms they're not going to be like, oh yeah, I read this contract and there's something doesn't work and I need to make sure that ZK proof actually like fully validates. So like, no, it's like, hey, should I put money into this thing? Is this better than my bank? Like, I'm from Ukraine. Almost everything is better than a bank there. So ZK had only like two problems in last 10 years. We've had 100 banks that failed in last 10 years in Ukraine. So clearly better product already. But I think like, yeah, like we need systems that actually can protect this kind of users, make sure that they defend them. And at the same time we need sacrifice values where the governance of the whole system comes back to decide to update things, update the rules.
- [42:49]
A
Okay, so speaking of bringing AI and intelligence on chain, let's talk a little bit about Near. So disclosure, where investors in Near Ilya, you're obviously the founder. Actually we seed investors long, long, long time ago. Back when the way you guys described what you were doing was very different. Once upon a time, you guys were literally an AI startup, became a crypto startup. Now you seem to be an AI startup once again. And one of the things that near is increasingly orienting itself around is enabling, bringing AI agents on chain and being the chain of choice for AIs. You also have a private AI stack that I know Venice AI is using. You've got your own OpenClaw variant called IronClaw. And I'd say I also contextualize this. It's been a brutal month for altcoins. Altcoins have just been just absolutely ravaged over the last month and a half. But near has been one of the standouts amidst kind of a bloody field of contenders. What do you attribute that to? What is there that seems to be driving all the energy around near these days? Why don't you speak a little bit to that?
- [43:50]
B
Yeah, we effectively kind of view the world that we have these two technologies, right? We have AI and blockchain. And AI is, I mean, I call it the single player, right? It's like it's your tool, it's your interface, it's how we going to communicate to computers going forward. And the Blockchain is a multiplayer, right? It's how we coordinate how we settle things, how we find, you know, what's the truth is how we identify each other, et cetera, et cetera. And so these technologies need to work together, right, because we're still living in the multiplayer world, but we going to not talk directly to blockchains like we've tried with all the defi. But we're going to have AI to be this interface. And my kind of perspective is that this is actually going to ripple not just through kind of our crypto bubble or even fintech bubble, but actually through all of the kind of business and personal lives, right? Where if you are running a business right now, maybe you have purchasing agents who are like trying to look for suppliers, you know, let's say Costco, bunch of people who are actually looking for suppliers to actually start deliveries to bring into the store. All of those can be done by agent they'll need. On the other side, the producers will have AI agents as well. They'll negotiate, they sign a digital contract, they'll settle the money, they'll execute the trade. The delivery will be done by effectively AI truck that will bring it to the store and then your robot will pick it up and bring it to your home. That's effectively where we're going, where all of these is just the supply chains run by AI agents and settled on blockchains. And so that's kind of where we're building the intent. Intents conceptually is the way we think people and agents will express what they want to achieve. At the end, it's outcome, specification. And so intent in the way to me was like the next level of abstraction from blockchain. Blockchain with transactions was very low level, right? It's effectively like instructing. You know, it's like writing assembly when you're sending a transaction to a blockchain versus intent is actually like, hey, I want this outcome. I don't really care if it's somebody does something on Solana, something gets executed on Bitcoin, or somebody does something in Salesforce or sends an email to this Chinese factory in Chinese to get the purchasing order happen. I just want an outcome that those goods or services get delivered. This outcome, near intents is this financial infrastructure to power that vision. And then near AI is the AI layer for both sides. Where it all comes together is a gentic marketplace where agents can actually start executing this work. Now, a lot of the current action is still in crypto, right? Like we've launched Private Intents, which was kind of part of the vision is like if all the financial transactions are moving on chain, no businesses want all of their kind of financial. Imagine all of their accounting is public now and you can go and see them, right? That's not how you run businesses. So privacy will be extremely important. Same as important for individuals, same as important for Traders. And so Private Intents was kind of a first building block in that and we see like a ton of traction, right. The kind of the private TVL is growing, the amount of trading is moving into Confidential Intents as well. And kind of again, we're approaching it very much like, hey, what kind of analytics we can run inside to prevent crime as well, because we don't want crime to actually happen here. And then on top, right, we had partnerships with Circle and for kind of agents using this same private balances to actually transact with each other and you kind of build up from there. So to us, it's like the core pieces are there and now we kind of keep integrating it and showcasing how this vision has kind of come together. But to give you an example, while the market was in bloodbath on Friday, we've done 220, 40 million in volume, which actually based on.
- [48:13]
A
Is that one day number?
- [48:14]
B
Yeah, one day number which actually at our current take rate was more than inflation of near. So that was a deflationary day. Now obviously it's volatile, but that's kind of the plan is like the token economics as well supporting this. As more volume, more transactions, both from trading and from real commerce are starting to flow, there's going to start buying back NEAR and effectively becoming more and more deflationary.
- [48:43]
A
Okay, so Mert, how do you do you buy this AI agents are coming on Chain Story.
- [48:51]
D
I have told Ilya in public to just please stop talking about AI and just focus on the trading aspect quite a few times. Like I think as these narratives keep happening on chain in, in, in crypto and the intersection with AI, there's a lot of places where you can get momentum and perhaps you get some learnings to then actually build into a better product and whatnot. I don't, I personally don't buy the like all the agents are coming on chain type of stuff. I do think still there's a lot of intersection of AI and crypto in other ways. I think Iron Claw is actually a pretty good example. I think obviously a certain amount of traffic will come on chain. It's not like it's just not going to happen. And I think NEAR could be the best place for that as well. So, I mean, I have a, you know, I wrote about this a few times about how I think NEAR is the most underrated team in crypto and I think they are extremely slept on and I think easily one of the best teams in the entire industry, right along with zcash and Solana, in my view.
- [49:59]
A
I don't you have a ranking.
- [50:03]
D
I'm not gonna rank. I will say. I will say. Near the top.
- [50:09]
A
Near the top. Near the top. So not number, I guess.
- [50:12]
D
Depends where you put the comma in that sentence. Near the top. Yeah, yeah, near the top.
- [50:16]
A
Hey, hey, hey. Very nicely done.
- [50:22]
D
I think the cross chain stuff intents the privacy stuff, I think all that has tremendous merit. I think AI obviously Ilya knows it better than pretty much anybody in crypto. So I think he'll figure something out for sure. And that team is also super cracked. I don't have a coherent vision for how that looks in the next year.
- [50:44]
A
Let's get the response.
- [50:45]
D
I just don't know what it is.
- [50:47]
A
All right, let's clap back at Merc. What's your response here?
- [50:51]
B
So I think the reality is obviously the things that already work is like private inference and kind of private hosting. We use a lot of blockchain primitives but it looks like a normal Web2 product, right? You can pay with credit card if you want, but you get fully end to end encryption. Our key management everything is back on. Literally the same tech that's powering the cross chain swaps, the same multi party computation chain signatures engine is powering end to end encryption for AI and multi chain swaps. So the core is the same and we're kind of building out this. I think the point about agents coming on chain, I think again very important to understand that the payments part, right? What everybody's kind of focused like oh, agents will pay on chain and most of the times that on itself makes no sense, right? Because if you just gonna like if your agent is subscript needs to use something, it can put your credit card on file and just use it, right? That's not the problem. The problem is when you actually need trustless interaction between different parties, right? That's what blockchain is designed to solve. When you open up new markets, what agents really enable is to open up places where markets before were really tough. Like if I want to buy something from Chinese factories directly, it's a very high barrier to entry. But my agent, it runs on Chinese model already so it speaks Chinese pretty well. And it has my preferences from all the contexts I've been talking to it. So it can go directly and negotiate a thing with them. And so then we actually in an interesting problem which is like I need to pay them but I also don't trust them. So I actually want money in escrow until the product arrives. They don't trust me either. So we actually need some kind of trustless infrastructure, we need some kind of reputation, we need some form of dispute resolution that both of us trust will actually solve, will actually resolve properly. And so you need all these pieces of infrastructure that usually you have Visa provides you or kind of traditional business relationships provide you that agents just don't have. And so that's where blockchain comes in, is in the whole business kind of flow, right? Not just in moving two numbers of kind of just stablecoin which by the way effects rate. There's a lot of things there that would be useful as well. And a lot of work actually on intents we're doing right now is to enable cross border effects very effectively because businesses can now start using this at larger scale if we have 0% on ramps on both sides and we can give you better effects than normal banking. But the trustless interaction is where you're going to get. And so to give you a few specific examples that are already live, right? So we have an agent that let's say you want to do an event in some place, you can ask your whatever claw to find you a venue. And it'll suck because it has no idea where to look for, has no good way to assess. Maybe it will look at some pictures, but it's probably not going to do this very effectively. Or you have a professional agency, event agency that built an agent that has database of all the venues in all the cities that effectively already pre scored. There was a person who went and checked out the vibe and filled it out. And now you can hire that agent that will do that work. And it's private, you paid it privately, actually validated everything and your agent can just hire it. You still type in the same thing to your agent. You say hey I need to find a venue in whatever in lisbon to host 50 people for a Dragonfly meetup. And it will be like oh cool, there's an agent that can do it for the dollar versus me spending a bunch of output tokens. Let's do it. And so that's example that's already working right now we have background checks for example for people, e commerce owners, right? They need to find suppliers and they need to do like a background check on supplier. Again, think of Chinese factory. Does that factory actually exist? Or are you effectively sending money into nowhere? And so again agent goes checks all the government registration of the company, their Instagram, their whatever, Vebo, whatever the we
- [55:19]
A
have to wrap because we're up on time. Mert, are you convinced Agent Commerce? Is it real? The deep sigh. That deep, deep sigh. Okay, look, we'll have to bring you back. We'll see if. Ilya. In the meantime. It could certainly be real. You hear that?
- [55:41]
D
Yes.
- [55:42]
A
It could certainly be real.
- [55:45]
B
I'm slowly working. I already got him on Intense, so this is.
- [55:50]
A
You know, I like that. It could certainly be real.
- [55:54]
C
Exploit.
- [55:55]
A
I'm going to get that on a poster. It could certainly be real. Okay. Beautiful. Crypto in a nutshell.
- [55:59]
D
Mert, where can people find you@mert X.com?
- [56:06]
A
beautiful. And Ilia, where do people find you
- [56:08]
B
at Il Black Dragon.
- [56:11]
A
Il Black Dragon. Okay. Wonderful. Thank you, gentlemen, for coming on. Mert, get some sleep and we will be back next time. Thanks, everybody.