Why Erik Voorhees Says Crypto Was Really Built for AI Agents: Uneasy Money

Transcript

I I have zero faith in the political process. Like I I don't vote for presidential candidates. I think the the train has left the station. Like the the institutional momentum and inertia of of the state cannot be controlled by anyone sadly. The only solution is actually technological. And this was why Bitcoin was so amazing was because you didn't have to go anywhere and vote for anything. You just use it. You didn't require anyone's permission for it. It actually allowed you to opt out of the system just on your own. That was what was beautiful about it. Hey everyone, I'm Kane Work and welcome to Unneasy Money because what happens on chain never stays on chain. Before we begin, here is a word from the sponsors that make this show possible.

>> This episode is brought to you by 1 in Aqua, the shared liquidity layer from 1 in. Back multiple liquidity positions with one wallet balance and keep your tokens in your wallet until a swap fills. See how it works at 1in.com/aqua. >> All right. Hey guys, I'm here with my co-host Taylor Monahan, security expert, and we have a very special guest today, Eric Vorhees, the founder and CEO of Venice. Welcome. >> Hey guys, >> thank you for joining us. Um, all right.

Uh, it's been uh interesting week this week. We've had some ETH uh price action for the first time in a I even Bitcoin like at 72K, it's been feels like we've been ranging for like six months now. Uh going nowhere. >> I know. went up. Yeah. Yeah. Exactly. Yeah. I know. I know. What are the odds? >> Um so, uh so let's let's jump straight in. Um we're very lucky to have um your perspective here, Eric, on on some of these topics because we kind of mix like DeFi and and AI on the show often. Um and one of the probably biggest things that came out this week was Stripe buying open router. Um I think there's a couple of interesting things. one open

router uh one of the founders of um OpenC um clearly also yeah yeah yeah also uh um big fan of the word open um we'll see what he does with his next startup um but uh but I think you know um in terms of like crypto pivots to AI this is definitely one of the most successful ones right but pretty hard pivot out of NFTs into open open uh kind of uh API inference um for these guys.

>> Yeah. And uh open router of course wasn't um crypto at all. It was just pure >> just pure Yeah. Like full pivot out of crypto like goodbye. >> Full pivot out of crypto into AI inference. Um yeah, it's an incredible achievement. I mean the company is like a little over two years old, something like that. >> Yeah. um nailed the timing, nailed the product, uh everyone loves it. They built a marketplace with providers of inference and consumers of inference over the API um and just totally crushed it. So >> yeah. Yeah. Very impressive.

>> You did a good job out GG. >> I know, right? Yeah. And and the other one I think uh the other one is uh Hermes. um those guys were were crypto and they kind of pivoted uh pretty hard maybe a couple years ago, two and a half years ago um out of crypto. But I think the of the let's call it crypto plus AI projects, Venice is probably the leading example we have right now. I would say like incorporating crypto into an AI product and and kind of mashing the two up as opposed to a hard pivot. So um you know maybe maybe like give us a head like why why did you incorporate a token into

uh Venice versus you know just going full uh pivot out of out of crypto into into AI. >> Uh yeah I mean first I can't pivot out of crypto I already re it flows through my blood. I I care I care deeply about it because it is it is a new way for money to move around the world. Um and and because money is, you know, endemic to all business activity. Um I don't really see it as like it shouldn't be its own ecosystem like it started that way, >> right?

>> Much rather see the primitives the best primitives from the crypto or DeFi world start leaving the crypto world and actually get into normal stuff. um if it's all just like crypto for crypto people serving crypto products with crypto then we we've kind of missed it. So I >> I wanted to bring I wanted to bring some of the primitives into Venice um to just like show the world that these technologies are not just for like speculating on coins that go up and down that >> right are these are incentive mechanisms and there's a lot of ways to do them badly but there are some ways to do them well and we're trying to demonstrate that in a mass market consumer app. I

mean is it fair to say that uh like the existence of tokens right or even let's say like models generally like inference right like the these you know and I say tokens like not crypto tokens right but like uh you know inference tokens the the new the new hot >> is really unfortunate actually >> it is a bit unfortunate although maybe we'll get like some you know nice tailwinds from people going oh I love tokens they're they're uh really good so so is it fair to say though that um one of the kind of interesting um uh side effects of the fact that the world wants inference and wants these tokens so much is that you have a product which is a digital product that needs to be paid

for. And we have this like, you know, digital money that you can kind of combine the two things in a way that like if you're trying to sell, you know, uh, computer hardware or or, you know, some physical item, the the use case or value prop of like buying it with crypto just was never, you know, I mean, we've all been there standing at a a register trying to pay for something with Bitcoin and, you know, even even with like the Lightning Network or whatever, like it just never really was like uh you know and we might get there eventually but like tokens inference is something that is like globally distributed it you know it's streaming is that something that that you saw as like oh this is a

perfect use case for paying in in uh crypto >> um that's not really what we were excited about. Uh certainly crypto can be used to pay. That's kind of like the level one basic use case. And so certainly if someone wants to buy tokens from Venice, they can AI tokens, they can pay with crypto, >> right? So that that's sort of like table stakes in my opinion, you know, like okay, >> crypto crypto is a payment method like Visa, like you know, Stripe merchant processor, you know, like any of that.

Um, more interesting stuff I think is actually the ability to give people stake in a platform and a project. Um and so for example, uh the all the major AI labs are private, you know, with the exception of Google. Um and because they're private, like normal people can't really have any kind of stake in its success whatsoever. And this is causing a lot of consternation actually where people see AI taking over the world. they see the tremendous amount of wealth being generated from it and they just kind of have to sit there wondering when their jobs are going to be stolen, >> right?

>> That's that's that's the impression that a lot of people have um rightly or wrongly. And I think that's um that's avoidable. And so in Venice's case, uh we have a token and we haven't deployed all the methods in which we will do this yet, but um we want our users to be able to have a stake in the growth of Venice itself from from its early days, from its founding. And um we have to be careful about how we do that. But I think that's a much um much more equitable and interesting way to build a company is like if your users can be participants in it with you instead of just consumers

of the thing, >> right? >> Yeah. You said you said incentive mechanisms, right? Like to build you know tokens uh are you know a way of coordinating people and again like you know people anywhere right anywhere on the internet. So, so not just like as a a you know medium of exchange but like actually getting people to to kind of contribute to uh to you know the project and the ecosystem. >> Yeah. Um and like something we haven't we haven't announced or deployed yet, but I'll convey it. um we want to essentially like take a portion of uh a user's payments to us, their monthly subscription or the money they're spending on credits and uh like give a portion of it back to them, using it to

buy a VBV and giving it back, >> right? And then letting them have it so long as they stay on the platform for a while. So essentially, you're like buying loyalty, which gives you the chance to teach someone the product and if they like the product and if they stay, then they end up with a stake in the thing without having to pay. becomes becomes sticky, right? It >> becomes it becomes stickier, right? Um, so this is something we're going to be experimenting with. And I like that better than like I I don't want to go to the users and be like, "Hey, we have this token, go buy it, you know, like that feels a little dirty." But if they're already spending money on Venice, if we can take some of what Venice is earning and give it back to them in the coin of the of the project,

>> then um I think for some portion of users, they'll find that to be a really cool rewards program. like >> and I mean, yeah, this is this is somewhat of a novel situation that you find yourself in where you have a crypto product that people actually want to buy that isn't the token itself, right? So, you have all of these levers you can pull to do exciting things when you when you've got something like that. >> Yeah. And yeah, like people were people have been experimenting with crypto so so much that they didn't realize that the crypto ultimately needs to serve something else. like in in Bitcoin's case, Bitcoin is serving the purpose and the good of of base money.

>> Um, every other cryptocurrency needs to basically be doing something else, something a different level of the stack. Um, you know, Ethereum, ETH is is powering this smart contract platform. That's that's its purpose and its product. Um, each thing needs to needs to have a product attached that it is serving. And we got to the point where like coin was becoming the product. um which was never supposed to be how it works like >> wasn't supposed to be the case. Yeah, >> that's not how it's supposed to be like incentivizing and coordinating something other than itself, right? It becomes very self-reerential if it's >> exactly the thing. Yeah.

>> Also like a race to the bottom, which is what we've seen for the last like couple years, right? Is just the meme coins and meme coins for memecoin's sake. And >> yes, and >> you don't get any you're not you're not growing the pie at that point. You're just >> I don't know. moving it around. >> If people want to do like um memecoin speculating, you know, God bless them. They should have the right to do that. >> Uh people have the right to go to casinos and gamble and that's perfectly fine. Um but it's not accreative to society. It is a form of entertainment that is consumptive. Um and it's tragic that uh projects started just becoming memecoins in part because it was risky from a compliance and regulation

perspective. I mean, that was Yeah, exactly. Like, you're sitting there >> trying to build something and you're like, "What the heck?" Like, this these guys are uh just launching nonsense and not getting in trouble, right? Meanwhile so >> if if you had a project that that had utility to it and you had a token, it had a higher chance of being deemed a security than if you just released a memecoin. >> Yeah. >> Um really tragic actually and deserves it, I think. Like there's there's a reason that like post ICO boom in 2017 that we ended up with several years of where meme coins just became the thing and actual projects with legit tokens attempting like DeFi summer was really

the the last time that that >> that was >> Yeah. >> Yeah. This is I have a question for you Eric and it's slightly adjacent to this to this topic but have you so I've been observing sort of let's call it like the cryptos crypto bro approach to regulation and law versus the AI approach and we've seen it a lot lately with specifically like open AI and anthropic and their policy stuff how they're approaching DC uh I would say the AI guys are much more keen to uh bend the knee the government than the crypto bros have

been. Um, I wanted to hear your take on on sort of the dynamic or the dichconomy between those and >> anything any other thoughts just because I find it absolutely fascinating. Right? They're two tech stacks that are moving really really fast that are both sort of antagonistic inherently antagonistic to global power but maybe should play ball. Maybe. >> Yeah. >> Right. Yeah. Any thoughts? lots of thoughts on that. Um, so this is actually kind of why I started Venice, which was that um, like in the crypto world, we have we have principles. We we say we do. Some of the people in crypto

attempt to live up to them. Many of them just talk about it, but don't really care. But at least we like have some principles that we can all maintain some degree of fidelity. And the principles are generally shared. These tend to be things around decentralization, privacy, sovereignty of the individual, decentralization of power. Like these concepts are common and we debate the details of them, but we agree generally on the direction. >> Um the AI world is and I'm making a generalization was completely uh not that >> same.

Yeah. Um, and so like again, and I I'm overgeneralizing a bit, but the AI world is is a much longer history and tradition than crypto. People have been studying this largely in academia for 60, 80 years. Um, and basically the mentality or the philosophy, the ideology within that world was very smart people in academia that believed things should be controlled. And it's just a what matters is who is in control and what the rules are.

>> Um very very top down um very centralization of power and make sure that the person in charge is the right person kind of mentality. >> I mean is it fair to say like academia is that right like it's very you know you're in >> Harvard or something and you're tenured and you're the you know so like that's the ecosystem that they came out of to some extent. >> It's the ecosystem they came out of. It's the incentive structure of how much of that was funded over over decades. Um, and I think it's also just the general demeanor often of people that are extremely smart like science types.

There's a certain degree of like I am a master of my domain and confidently so and thus I project out that there must be a master of every domain including society itself. And of course they what the logical fallacy is is that they mis construe the narrow domain which can be understood for a complex system which cannot >> right. >> Um so anyway that's like sort of the milloo out of which the AI world emerged and when I like you know two and a half years ago when I started getting into this stuff um nobody in AI cared about privacy nobody cared about asation nobody cared about user sovereignty. In fact, it was all all the opposite like

uh it was all like this technology is extremely dangerous. Government needs to regulate it uh immediately. Everyone needs to be watched. You like the prompts that you send need to be censored. The answers need to be censored. Um everything needs to be tightly controlled. And our sin is that we're not controlling it tightly enough. >> Yeah. >> So that was a completely different cult. That was like the doomer the doomer perspective. >> It's weird. It's weird. It's and it's it's been weird to watch >> and it's I mean it's fascinating.

>> It's not weird for normal people. It's actually the more normal. >> We're just weirdos. That's the problem. >> We're the weirdos. >> Yeah. Right. Fair. >> Been here for too long. >> Yeah. We're the weirdos. Uh so, so I looked out and I saw, you know, like um AI was obviously going to start taking over the world. You had open AI and you had entropic with the frontier models and they are very top down monolithic control everything. We will enforce safety with our rule sets. The government will govern us and we will you know do what they tell us to do and that's how society should be ordered.

And to me that was going to lead to like a very dystopian situation where machine intelligence itself was coming essentially through the official mouthpiece of the state. Uh to me that was the dangerous outcome and um I like to half joke that Venice was set up as a as an AI safety company basically prevent that kind of thing from happening. But yeah, Taylor to your point there's a there's a reason that that those types are much more likely to talk with the government and invite them in.

>> Yeah. >> I mean I think there's there's like a philosophical thing to this as well though, right? like because you know AI um like crypto is you know you said this T like it's destabilizing to power structure right like you know it is it is kind of inherently a thing that can route around and you know we've seen this um with DeFi like DeFi versus uh you know um call it like you know the trady approach to innovation right um you know sandboxes and things like that that never really went anywhere like fintex you know neo banks like it's all just like rappers around the same thing.

Um and DeFi was like, "H, hold my beer. We'll reinvent finance from principles," which has been good and bad, right? Um so, you know, crypto is inherently destabilizing, uh to existing power structures, but AI is arguably like destabilizing to homo sapiens, like >> possibly, right? And and so therefore, like the stakes are even higher um you know, potentially, right, if you if you get this wrong. Um and so you know you look at like Bostonramm and and some of the people that you know kind of created this like uh sense of the stakes are really high here. To your point though, when there's high stakes there, there's

like a that that academia um you know solution is like well let's create some rules and regulations around it and you know that'll reduce the the stakes from killing. It is the way to frame it is like this thing is obviously powerful and may be dangerous thus let's centralize it or this thing is powerful and may be dangerous thus let's decentralize it. >> Yeah. >> Right. So crypto people tend to fall in the latter category and most of the AI people fell in the form.

>> Yeah. It's also it's been dare I say some of the way that the the the AI companies have been navigating especially like policy in DC. I feel like perhaps Eric you and your buddies would be screaming at me on X like this is what we this is what whatever. Okay. So for history, me and Eric sometimes fall on different sides of the fence with the details on how like crypto should do certain things regarding like say hacks and moneyaundering.

Basically the argument, right? The argument is that if you um if you do too much or if you take the wrong approach, you're basically shooting yourself in the face. And it's not necessarily a good thing because a lot of times you're just moving where the power is, where the centralization is. And that's what AI has done essentially, right? Is it's it's a battle right now. We're not it's not like done, but it has fully sort of the the question is not um whether anyone is going to control it or not, whether it's going to like serve people first or not. The question is who is going to be the person who like sort of wins that control and what role the

government's going to play, right? >> Yeah. >> And I think it's interesting that I think it's just really interesting that one they've been so eager to do this and two just how quickly it's sort of I don't know uh I mean they can't release models at this point, right? Like literally OpenAI and Anthropic are are not releasing models. thing >> there. They are all releasing models. What's not known is what is is getting not released.

>> So, we see we see what we see. We see what's released, >> but you don't actually know how far ahead of that the model the labs are um internally. Nor do you know how far they would have been if they weren't restraining themselves. My guess is that none of them are actually restraining themselves. At most, >> at most they're restraining what they release, >> right? and um they're all in a wild competition. I mean, I'm glad I'm not running like OpenAI or Anthropic like um >> yeah, >> seems incredibly stressful. >> Yeah.

>> Trillion dollar companies. Um and yet there's like these open source models that are now one or two months behind them and catching up uh at at 10% of the cost. I mean it >> um >> while while by the way while a huge amount of leadership efforts are dedicated to literally sitting in DC and holding the hands of like the Trump admin who's now trying to figure out how AI works and how these models work and >> how dangerous it is. >> I think a very delicate balance because if they go to DC and say hey this stuff's dangerous so regulate Chinese models.

>> Yeah. Um, that can also mean that that their own company starts getting nerfed or caught up in the confines of it. >> Yeah. >> Um, you know, I the truth is like I don't think anyone really understands how this stuff is all going to go and everyone's just trying to make it up as they go along as as humans do and everyone's pretending more confidence in the future than they actually have. >> Hence decentralization, right? like you know if you if you don't trust you know then >> one of the great arguments for decentralization is just >> the the humility of it right which is to say like the future is actually very complicated and we don't know and that's okay we shouldn't presume knowledge that we don't have uh and if things are

decentralized then at least where the problems happen are localized usually instead of catastrophic and systemic >> that's yeah very very uh a suit assessment of the situation. I think um from our perspective and and you know again like um we have built a set of tools over the last 15 years or whatever um in in crypto that is like perfectly designed to help decentralize things like this. Um we got a little too caught up in like doing the thing for its own sake, right? the kind of you know self-referential naval gazing stuff. Um but now there is like a thing that people want um you

know in in the form of inference. Um there are massive constraints in terms of how you produce it. There's all of this competition. there's this whole landscape and you know this is why I think it's so interesting that Venice is one of the few projects that's actually at the intersect like a lot of people were like ah see you later crypto's dead I'm going to AI right um >> yeah there's a lot of fair weather crypto people but um >> yeah yeah yeah but so like you guys are one of the few that are like well hang on a second we have a set of tools that we built there's this new thing that like massive demand for this new thing inference um let's actually plant ourselves at the, you know, intersection of that and the world and and, you know,

kind of create these incentive structures to make sure that we don't just centralize it. >> Yeah. Like I I actually just think crypto is better financial tooling. Period. Full stop. >> So, agree. I mean, yeah, >> of course. Of course, anything I build is going to try to use the best financial tooling I can. Um, it's it's not even a question of ideology, but of of pragmatism. Um, and I I there's there's a theme I think that more people are realizing, which is that uh cryptos always had this UX challenge with humans. Um, and every time you go to a crypto conference, whether it was back in 2011 or tomorrow, people talk about like how do you make this stuff usable like more usable?

And it seems like perhaps uh it's just much more usable for robots and the machines and the AI and that the crypto is actually built for the machines that we didn't know it and that the the machines are now going to like start running on the stuff and utilizing it. You know, they they have no trouble with public key pairs. >> Yeah. >> And that's easy for them. Um so yeah, I I don't see a future where like the robots are using bank accounts at Wells Fargo. Um I do see a future where they're using crypto assets on decentralized rails >> that I have no idea which rails or blockchains or if it is a blockchain but but natively digital financial technology that does not have a

political gatekeeper >> is is the least friction way for value to move and I would imagine that a super intelligent beings would prefer such a thing. I mean, you know, like very obviously bank accounts are not ergonomic for agents. Like you can just see, you're like, "Oh, this is just a dumb thing." Like an agent doesn't want any part of that, right? They're like, "Why have you got all these weird layers and protection? Like, I just want direct access to the thing." >> No. And I don't know, Kane, I don't know if it's the same in Australia, but here, if you send any amount of money via your bank, you have to literally either get on the phone, click five buttons on two different devices, or walk into a physical bank if you want to wire like literally any amount of money.

>> Um, every single time I'm like, "All right, I have to go wire money. I'm going to set aside the whole day because it's that miserable >> to go and spend your money." I know, right? >> Yeah. Meanwhile, like last night in bed as I was falling asleep, I was like, I don't know. Like, let's blow him some E. Like, two seconds later, we're going to go golden, right? Like, yeah, >> it's night and day. Night and day. >> All right, let's let's go to a break and then when we come back, we're going to talk tokens and equity. Um, so, uh, we'll be back in a second. >> $540 million. That's how much concentrated liquidity sat idle in a given week in the first half of this year. about 30% of the DeFi TVL if you're wondering. That's according to

Dune Research commissioned by 1 in. But there's a solution. 1-in Aqua is the new shared liquidity platform. It lets LPS back multiple positions with the same token balance and keep their tokens in their wallet till a swap comes. Why does that help? Because the LPS don't have to split their tokens across positions. They can cover more market conditions and pairs with their full balance. That means more activity across deeper liquidity. See how it works at 1in.com/aqua. Remember that providing liquidity carries risk and fees aren't guaranteed.

>> All right. Um let's uh talk tokens versus equity. Um I think over the last couple of weeks with the most recent uh raise that you guys did, this topic came up. Um and I think it's a topic we've we've spoken about on the show a number of times. um you know maybe I'll I'll preface this by saying like I as a token maxi um for a very long time um I've definitely felt the uh I guess um effects of like you know just how how poorly tokens have done like including you know my my like portfolio companies how hard it has been you know we were talking about memecoins like if you're not a memecoin it's very hard for for

you know you to get any traction. But um I think >> yeah, it's it's been >> so lay in >> I know it's really bad. It's really bad. U I think we'll fix it. Uh I think we'll fix it in in the next bull market. Um and you know, we'll actually have some stuff that that you know people can see. Okay. Like we don't really have that many good non-crypto use cases. You you look at something like a or whatever, right? And you go, okay, like Ave there's utility there. You know, you see why the token exists. it keeps the ecosystem working. Um, you know, but a lot of the times, uh, it is just pseudo equity, right? Um, and we've talked about this on the show. Um, pseudo equity sounds great and, you know,

there's a lot of nice things and I think you you tweeted about this. You know, you have, uh, liquidity is one of the amazing benefits and access of tokens. Anyone anywhere can buy a token. You've got a lot of liquidity. Um, you know, it's not gatekept in the way that equity is. Um but it is an equity and it doesn't you know confer ownership rights and you know you can get rugged. There's a lot of tail risk in being a token holder. Um you know technical tail risk uh you know execution tail risk etc. So, so I guess like walk us through your uh your kind of thought process because you guys launched a token first, but you obviously had an entity that had equity

in it, but you hadn't raised. And then you were presented with this opportunity where, okay, we've nailed it. People are really excited. This is working. People want exposure to what we're doing. And you know it made more sense at the time clearly to to give them equity plus tokens including like token wars and stuff than pure tokens or pure equity. So maybe just walk us through like how how what was your reasoning there and and how do you see the trade-off space? >> Yeah, so these are different things. Um they have pros and cons each of them. We started Venice just as a normal company.

Uh so we're a Wyoming corporation. Um and for a year we built the product uh all self-funded and didn't raise any money. Um of course we were going to bring tokenization into it. So about a year later we released VVV and then DM. Um never sold it and uh so still hadn't raised any money but then now we had a token as well. Um then another year goes by. um we had grown a lot and we were at the point where it's like okay we're ready to scale this thing. We've demonstrated market fit. Uh let's let's go big. Um and we were we had this choice like okay we have two assets here. We have equity and we have tokens.

We have a bunch of both. Which one do we want to sell? Um and we decided let's sell the equity. We'll we'll um we'll sell that. we'll keep the tokens and uh in some ways it's more normal because the normal VC market understands equity better. >> Yeah. >> Uh but really we just like we kind of don't want to sell these tokens like we've never sold them. There's a reason for that. Uh this is part of what we're what we're building toward. Um so we we sold equity. Um the the VCs that bought it have rights on the token. So they can they have an option that they can buy token at a certain price um which then vests over

four years. And that was important for me because um even though the the investors don't have anywhere close to controlling state um I want them to understand the direction that we're taking this which is like we're going to try to burn all the tokens in existence. uh they need to understand that they need to like be okay with that very unorthodox financial strategy which means that they need incentive on the token side as well. So they have warrants toward it um and they know the direction. We've tried to be as consistent and communicative as we can about what we're doing. But it it is a it's a challenging thing to kind of strike this balance, right? You know, you've got tokens and you've got equity.

Um there is always going to be some kind of inherent conflict, right? If it was pure equity, then you return cash to equity holders through either, you know, buybacks or dividends or, you know, whatever. Um, if it's a pure token, you move it around and probably nothing happens. Um, which is, you know, most tokens, right? Um, but if you have revenue, you can put it back to, you know, token buybacks, um, or or distribute, which is a bit more risky for tokens, but most likely buybacks, right? um as a way of returning value to to the owners. Um it it becomes very interesting if you have these two different ownership mechanisms and you have one stream of revenue or cash flow

that you need to divvy up, right? And you got to try and balance the uh I mean you could I guess you could look at it as like dual class, you know, shareholders or or something like that. Like how do you guys think about that in in balancing that? I think the the main thing people are missing is that they perceive that these are two different groups that you have >> equity holders and that you have token holders. Um and everyone involved in the company has equity and tokens and so their interests are in both and the company itself has more tokens than anyone. So we don't see these things as very misaligned. In other words, like um every dollar that we spend burn burning tokens, is that harming our equity

holders? No. Like I'm the biggest equity holder of all. Why would I why would I do that? Um no. We're doing it because we actually think it helps equity there. Like this stuff was set up for this purpose. We have a token. We're going to grow the business. We're going to buy as much of the token as we can. Um which will send the price up. And the company has more of the tokens than anyone. So we see these things as like aligned certainly unorthodox and certainly the tokens and the equity are different things with different different structures but we don't see that their interests are are misaligned but I mean at the limit right if you're pure so you know uh synthetics we were a pure token project like maybe one of the most pure ones we didn't even have entities we shut all all our entities down and we

said like somehow I'm not in jail I don't know I don't know exactly what what happened there but the SEC just missed the memo on that one, but we literally had no protection. Like our lawyers are like, "You are insane." We had a foundation. We're like, "We're shutting the foundation down. Nothing." Right? Like pure doubt. Like absolute insanity. This is 2019. So see >> I know it. Well, it it seemed like a good idea at the time, right? And and so, you know, there was no way for misalignment um to to happen. That was kind of the thesis, right? That like, you know, the owners of the token voted in the the DAO membership. the DAO controlled everyision etc. like there was no dual class you know structure at all. Um and um and I think that like you

know at the limit if you if you are that kind of structure right there is no um even potential conflict or or whatever like all of the revenue all of the value flows back to token holders. Um and and so I guess the the question is like you know what is the benefit then if like you guys had equity right but you could have said actually like let's shut down the company and do you know probably would have been a bad idea um just from a regulatory perspective but like you know you could have said let's just only have tokens and be a pure DAO and therefore like 100% of the revenue has to go back like at that point what's the benefit of of the equity other than like >> yeah I think Um maybe a lot of people

don't quite recognize that building an AI company is actually quite capital intensive. Like we're spending tens of millions of dollars on GPU. >> Yeah, we need an entity for that kind of business operation. I mean, OpenAI had this exact problem, right? Like they're like, let's be a nonprofit. And then they're like actually we need to raise like a trillion dollars so we're going to go ahead and be a profit uh profit making business. >> Yeah. I mean us having an entity has nothing to do with uh regulatory protection because in my experience having an entity is only only invites regulatory scrutiny frankly.

>> Agree. >> Um there's a there's a myth I think that it like protects you but I've only found that it causes trouble. So that's certainly not why we do it. Um yeah, we did it because we started as a company then we we grew we released a token and when the time came to raise money we're like we have two assets which one do we want to sell have less of? We want to keep these tokens so let's dilute the equity and sell that. So, so we did >> I mean, you know, reading between the lines there, right? Like, you know, you've got an asset that is uh infinitely inflatable and then hopefully an asset that's a harder asset, right?

With, you know, with some kind of fixed supplier emission schedule or whatever. Um, as a crypto maxi, you look at that and go, uh, I've got a money over here. I can just print some more share certificates. Like, why not, right? And if someone's willing to buy them, >> then great. >> Yeah. Exactly. And like our our entire strategy with VBV is to take it deflationary. So the actual token supply will shrink with time. Um whereas the equity we have no such strategy. We're not trying to shrink the equity supply. >> Equity. Right. Right. Why would you why would you Yeah. There's no reason to.

>> Let's deletute that. Let's dilute that. We care more about the token. And let's make sure that all the people that are investing have exposure to the token on the upside. That's important. Um, and then most important of all, let's just keep building a business that's growing very fast that people like. Um, yeah, it's not really much more complicated than that. I think the issue is that like these tools are all very powerful and so in the hands of people that are misaligned with themselves or with parts of their organization or with users or holders, they it can go very badly obviously. Um but I mean I I'm the second >> that's one of you know tokens like the the the brilliance of tokens right is

that they are so powerful uh and powerful things can go wrong right so >> that's right and I I want to show that they can go right actually that's that's actually important to me um and when these people are all bickering around on Twitter they're like calling tokens these secondass citizens and stuff I I want to just yell at them and be like bro I am the token holder like >> it's weird though like it that doesn't tend to um I don't know why, but that that just doesn't land on Twitter, right? Like anytime I've ever said something where I'm like, "But I'm the largest token holder. Why would I do a bad thing to token holders?" Like, I've got the tokens. There's some there's some like brainworm where like people on Twitter are like, "Yeah, I don't care."

Like it's like they they can't get the incentive alignment there. >> Yeah. It's it's hard. So we you know we we have to decide like how much do we even want to argue on Twitter like most of our users are not on Twitter actually >> fair >> most of our customers are not crypto people >> must be nice must be nice for you >> what's nice >> it's all transparent right so the flows of the of the tokens are transparent the supply is transparent how much we're burning is transparent like we don't really need to say anything we're just empirically demonstrating it with time um and hopefully that will win win win any argument but it is always tempting to pop back on Twitter and try to prove the guy wrong, you know.

>> Yeah. >> It's never never works. >> It never It never pays off, but it's always >> You can't help it. Yeah. >> Yeah. >> I think I was just going to say I think generally people especially like the Twitter crowd I think they've just been burned so much. >> They have I get it >> that they're just like you can't you can't reason it at some point. They're traumatized and >> and then they go buy like a meme coin and then they wonder why they were having a hard time. Yeah. And you know, >> I mean, you know, it's the incentives, right? Like the incentives are structured so that you're better off buying a memecoin because people who are trying to do real things have to pretend like the token is not even a thing. They don't they can't talk about it, you

know, they're they're trying to hide it away instead of putting it like at the forefront of the of the thing, right? >> Yeah. Yeah. So hopefully we can be a good example and show that like these are uh interesting and powerful financial tools that when assembled correctly can be accretive to all groups. Um of course there's a lot of ways to do that badly but there are definitely ways to do it well. Uh and you know earlier Kane you mentioned like tokens are fraud with with danger. Um but so is equity. People just don't see it because it's opaque.

>> Right. >> Right. You don't see the equity that goes to zero when it's a private company. >> There's no chart. >> Exactly. >> There's no chart. It's wiped out. Right. So, so all of the blemishes on the crypto side in front of everyone to see >> which is also the powerful thing, right? Like you can, you know, you can see just see what happens. >> Makes it makes it tricky. >> Yeah, it does. It very much does. Um, so let's talk about uh um open open weight models. Um, you know, you mentioned this earlier that like, uh, it must be super stressful, um, you know, as Daria watching GLM 5.3 start to like, you know, claw its way towards you or whatever is going on there, right? Um, so, so like how do you guys think about

this? How are you staying um, you know, kind of uh, a breast of all of the different models? You know, you mentioned before the show even started that like open weights are coming for GLM 5.3 pretty soon. Like I' I'd love to just get like your like take on the procurement process of inference like in in the background. You you mentioned, you know, GPUs and everything like what's what's the day-to-day in Venice? >> So um so Venice does not train models and when you're using Venice, you're not using a Venice model. Uh basically within Venice you have access to all the models of the world. Every major closed source model, every major open source model is all in one place. One API key or one app and all get it in one spot.

Um so we've become an aggregator essentially of all the world's models across text, image, video, audio. Um now when we started like that wasn't too hard to do because new models would come out like once a month important and we'd add them. Uh now they come out literally every other day. >> Every day. Yeah. Yeah. It's crazy. >> Every day. Uh sometimes three will drop in a single day and my poor team is like which what do we do? You know and >> Yeah. >> And each one becomes the new hot thing and so we have to get it live very quickly. Um the team does an amazing job now. Like we we get models live often within 30 minutes of them being released. Um and then there's nuance in terms of like when they're released,

when they become open source. uh if they're not open source, who is actually hosting them, how much do we trust them? Do we have a zero data retention agreement with that company? Like all these kind of things go on in the background so that we know whether or not a model can be listed as private or not. Um so in Venice, the the normal models, most of the models are private. They have a private label that means like no data retention. There's no storage of the prompts. There's no storage of the answers. There's nothing to subpoena. Um, if you're using an anthropic or open AI model through Venice, those companies should still be assumed to be holding those things, right? So, we don't mark those private.

>> Um, yeah. >> And and so, so you know, okay, GLM 5.3 is about to drop. The weights drop. They're on hugging face. >> Yeah. So, so like what's the day look like? >> GLM um is the latest model from a Chinese lab. Uh it's not open source yet, but it will be in a few days. Um so right now the only way to use it is if you go through the company that's hosting GLM um ZAI is ZI. Yeah. So if you're accessing GLM 5.3 today, you're doing it from ZI.

Uh so it is an open model, but is it private? Well, that depends on your relationship with ZI and what they're actually doing. Um, Venice doesn't trust any thirdparty company like that enough to list it as private. You can access GLM PI3 through Venice, but it's not listed as private yet. When it's open source, then we will run it ourselves. Then we can ensure as private and then we'll mark it as private. Um, so yeah, it's a big important model. Um, but almost all of them are these days. And yeah, it's very hectic and stressful.

>> I bet. I bet. And so so um you know you said uh like the hot model right like um have you guys had to uh you know like I I use run pod like I'll I'll put stuff on runpod I've got my own local machine um it is like a punishing exercise to like stand these models up and test them and you know make sure that you got like the right token throughput like how do you guys do you have to build your own infra for managing the the routing of the models and And >> yeah, when we when we started, we were running the models ourselves on hardware we were leasing. So we controlled the full uh GPU and um we learned that we were actually quite

good at running image models, but LLMs are for some reason much more complicated. There's like many more settings and configuration, all sorts of interesting caching questions. And we realized that like we didn't have a competency in running LLM. So we started working with partners that we had um zero data retention agreements with and that we roll the GPUs with them. Um so that's what we've been doing up until recently and uh following the series A that we did we've started buying a bunch of GPUs ourselves because we're going to get back. So we've been spending time learning how to run LLM well and now we're getting >> now we're increasingly competent at it.

Um >> so yeah, it's it's always like a a push and pull. Some models are easy, some are extremely pnicity. Um, and then there's like variants of models, right? Uh, especially like uncensored variants. Um, we have, you know, like one of our USPs is that Venice is uncensored, which means that we don't add any content moderation to the input or output of the models, but the models themselves all have different degrees of censorship >> baked into Yeah. >> baked into how the model is trained, ranging from very censored to very not. So when a user comes to Venice and they access a model and it refuses their question, you know, often they'll be like, "Hey yo, what the hell? You guys said this wasn't censored."

>> Yeah, I know. That's scenario, man. That's not us. >> Yeah. Um, it's a hard thing to communicate to a user. And then there's open then there's like sometimes uncensored versions of the models, but typically the uncensored versions of the models decrease the um intelligence of the model. So then you have this weird trade-off where like we can add the uncensored version of it which will make the guy happy if he asks about how to make math. That's what everyone start >> they're like if they can't tell me how to make math I'm out like >> everyone wants to ask that um just because it's a funny question.

>> Yeah. >> Get heristic but once they can answer that then it gets you know like less intelligent on the other things. Uh so yeah like the the customer UX question is a ongoing constant challenge that we're always trying to balance. So, so maybe can we talk about um the economics of uh of serving models and and all of this for a second because um you know when um when my team started to like really use a lot of inference, right? Like we got to the point where um you know we had uh like 100 200k uh monthly bills for inference, right? Um and and one of my um one of my engineers was like, I have a Claude Max plan

personally, right? And I'm doing a bunch of stuff personally because he's just cracked. Like he works seven days a week, like doesn't stop, right? And so he's like, I'm looking at the API billing here and I'm looking at this and I'm doing more stuff. I'm spending more tokens on this max plan than I'm spending on the API. Uh how does this make any sense? Right? So, so like what's what's your as someone who you know is is obviously constrained by this API billing um how do you guys think about that when the when the labs are subsidizing uh these plans like so aggressively?

>> Yeah, I mean it's um so first of all, Anthropic is losing money handover fist on those plus $200 a month plans. Um, I don't know how long they will tolerate it or how much they're losing or what, but that must be something that is constantly being discussed and debated because every customer loves it because it's an unbelievable deal because, you know, people will spend $10,000 of tokens in a month and pay 200 $200. >> Yeah. >> Um, now if that loss center is only is a small enough portion of Anthropic's entire revenue, then they can just deal with it for a while.

>> Yeah. But yeah, uh that's surely adding to their stress as well. Um Venice's API doesn't have any unlimited plans like that. Um I didn't want to try to do such a thing like you just charge per credit on the API. So each model has a different price and use whatever models they want. >> But I think this is the most interesting thing is the economics. You know, you say like there's new models coming out. The economics shift constantly. And I do feel like being in crypto for for as long as we we have been uh our tolerance for like the uncertainty of pricing and everything is so like your team must just have an inherent ability to like handle this. But you know uh Deepseek as

an example, right, came out when when uh V4 Pro came out um and like was 50 times cheaper, right? And all of a sudden the number of tokens that you were getting, you know, if you're if you're on a token basis, right, from the max plan and how much you could actually just pay through the API was like identical. Like the pricing was basically you pay 200 bucks and you get the same amount of inference as you were getting from a Claude Max plan, but the tokens were 50 time, you know, you're getting 50 times more tokens than it would have been if you' paid through the API on Claude, right?

That's an insane variance to like be dealing with in in terms of pricing and how you how you meter that. >> Yeah. Uh I mean the what's clear is just that the cost of intelligence is falling dramatically and has been now for a few years. The like the dollar cost to do a certain type of work is just plummeting. Um this is an amazing thing actually for the whole world. Like you know it's like if if the cost of food production was falling drastically or the cost of electricity was falling drastically like these are fundamental inputs into civilization that make civilization in general thrive.

But if you're the provider of the model and the tokens, it's a very stressful thing to be involved with in a product that is constantly >> deflationary and commoditized. You know, because >> everyone's competing to run these models. the margins are extremely small. Some companies are are undercutting their own margins with VC money, subsidize it. Um, and there's not really like the the end state is that no one's going to make any material money on selling tokens, >> right? >> Um, that's true of all all commodities and tokens are a commodity. Um, AI tokens. I mean it's it's interesting like we have a counter example like you know the the only deflationary thing that I think humans have any kind of uh sense of in in the world is uh tech

right like you know computers get faster like computation you know Moore's laws kind of call and you know obviously even storage and you know a hard drive that used to store 10 megabytes now you know and was was a million dollars is you know now costs like tiny fractions of a cent right so we we have some sense. The only people that have really pulled off the ability to like uncommonize themselves are Apple, right? Like they they wrap these these deflationary things and they they manage to, you know, they manage to kind of put this brand wrapper around it um to the point where people are willing to pay a large premium for something that would otherwise be just a and that's software.

And there's a whole bunch of reasons why that >> they find they find ways to add value, right? So in Apple's case, the software is so good and the hardware is so good and the symbiosis of the two is so good that >> regardless the falling um prices of any of the components that they've built value out of the whole experience. >> Yeah. But also like the app store and like the there's this there's this some things have been commoditized individually within the whole Apple Apple ecosystem but they've managed to as a company on the whole sort of dodge the complete undercutting of themselves with the compound. It's very interesting >> but surely for like for the la for

anyone selling inference or creating inference or you know like that has to be an example of like how do you find a way to add value because otherwise you know if it is genuinely a commodity to your point you cannot make a profit in at the limit on >> right so like in Venice's case um so we have essentially like two revenue lines uh one is that if you're using the API or you're using premium models and paying credits for um people are buying credits for that and that's one model. The other is our pro subscriptions. So the pro subscriptions are actually a much higher margin. Like someone's paying $18 a month and we spend or lose $6 to $10 a month on the person. So that's actually

like reasonable margin. Um and in that case like the value ad is that they get a bunch of models for free. They get a bunch of usage for free. They don't have to worry about every time they're spending little bits of money on each request. Um, and so that that works. That creates like a a symbiosis there. But the we we don't expect to like make material money on the uh the API token sale side of things. That's just a race to zero. >> So, sorry. You said you said you you haven't done the um anthropic like you know subsidized plans, but you you obviously have like not the 20x max like use 10 grand and pay $200 for it, but like yeah, you do have the ability for someone to sign up for a plan and and you know get some amount of inference

and you're assuming some amount of usage >> limits. Yeah, that's in that's in the app, right? And the limits are all pretty reasonable so that most people don't hit them. Um, so it feels mostly like mostly unlimited unless you're just kind of being crazy. The API is different because APIs by their nature are um mechanistic and people's usage of them can be automated. So it's very dangerous to give any kind of free thing over an API, right? Because it can just be >> I mean Yeah. Yeah. Exactly.

>> And is doing it so >> yeah. >> Yeah. Sure. Sure. >> Yeah. Kane has been Kane has been >> I feel very philanthropic just for Kane and his team and they're >> but they how they hit them. >> It's absolutely insane. Um, Eric, I want to ask you sort of on the same I don't know these days I feel like a lot of the conversation with AI and the the different models is this geopolitical fear of China and it's not just it's a bigger different conversation than the the strictly the privacy one. It is very geopolitical and people are very especially in in the west China is the

scary thing. How do you guys think about this and approach this bonus? I mean it is the geopolitical story right like this China US um antagonism that is like the proxy fight is in AI now um I mean yeah how do we think of it like I'm not a nationalist at all um I care about principles um I do not care about the United States of America as a nation I care about the United States of America's principles specifically the good principles related to individual freedom privacy um these kind of things. Uh if uh if Chinese models better express those

principles, then I care more about the Chinese models than I do about the American models. And I don't think America deserves to win anything if it uh if it simply like closes down markets so that others can keep uh and or if it puts in regulations that require their ability to like watch everyone's uh inference feeds. you like America is going down a pretty dark I would say like a pretty dark dystopian path of trying to control everyone. Um over the prior over the past decades the meta trend is that the that the US is controlling and monitoring its people increasingly and the Chinese government

is actually doing less right they started out like very >> high bar in it was high bar at first >> they're like very authoritarian and they've become more market oriented over the last 50 years whereas the US was extremely market oriented more authoritarian >> so that that franks me and I don't know which like getting very weird. >> Yeah. When people say the US must win, I have to say like what why should win actually is the the principles. The principles, right? Yeah. Like at what cost >> does does Yeah. the US win.

>> Yeah. Well, and a lot of times a lot of times people when they say like the US must win, what they mean is like freedom must win and individual rights must win and privacy must win and you know but they don't say that >> it should mean that Taylor and I think they think they mean that. >> But um it's one of those things where like over time what Americans cared about of those principles abstracted itself into symbols like the flag >> the flag. Yeah. Yeah. you know, the flag and the presidents and Mount Rushmore and like all these symbols and people now care more about the symbols than the principles.

>> And you can just tell that through through policy, right? You can tell that just in in watching the size of government get bigger and bigger and bigger and bigger, which is >> the most anti-American thing that could happen is that the federal government keeps >> the government just go the whole damn point to like a large government. >> That was >> that was that was the one thing. It was a good run. 250 years. It was a good run. I mean, you know, >> that was a great run. >> I mean, some of the most intriguing conversations that I've seen today, especially because it is about AI, it is about China, it is about the US, are the conversations where um the conversation about like potentially nationalizing the AI companies, right? or

>> like doing the government to taking any stake or role with our AI companies is basically the most Chinese thing you could do, >> the most anti-American thing you can do. >> It's the most like, hey, we're just going to nationalize this cool thing that you guys made. >> And it's bizarre. And the fact that they do it like just the the mental gymnastics, >> it's even like in fairness, right? Like to your point, it's not even that controversial, right? Like >> it's not even like people are like it should be it should be insane that it's even like in the Overton window is like h like yeah like 30% nationalization of these labs is fine like

>> yeah um yeah it's really sad actually and uh the only hope you know like I I have zero faith in the political process like I I don't vote for presidential candidates I think the the train has left the station like the the institutional moment momentum and inertia of of the state cannot be controlled by anyone sadly. Um the the only solution is actually technological and this was why Bitcoin was so amazing was because you didn't have to go anywhere and vote for anything. You just use it. >> Um you didn't require anyone's permission for it. It actually allowed you to opt out of the system just on your own.

>> That was that was what was beautiful about it. And so people's refuge I think is only in technology and in particularly decentralized technology that has no central ruler. Uh so we have crypto for that in terms of finance which is great but um if we do not have decentralization in the realm of machine intelligence then we are in big trouble. Uh we we have that in parts of AI. Um inference can be done fairly well in a decentralized way. the production >> training. Yeah. Production >> training not yet >> bottleneck. Yeah.

>> Yeah. Now there are companies like making good progress on this that are doing decentralized training. So there is hope but they're not competitive yet. So we that that's TBD. >> Yeah. I think it is an interesting situation where even six months ago the delta between um you know call it uh like you know codeex 5.1 or 5.3 um and like you know almost any open source model was just gigantic right and it's and it's come so much closer but also there is this like at least it appears recently this kind of like frontier in terms of uh intelligence where like if I want an agent to write a

web page there like more intelligence doesn't necessarily make that much of a difference, right? Like you like at the margins, right? Like >> depends. Yeah. Like it depends uh how much more intelligent and and at what cost? Because if you can have like a model that's three or 4% less intelligent, but you can run it three times for the same price. >> You can Yeah. brute force, >> right? Exactly. you can have it validating itself and you can use more tokens in its um in the creation even though the raw intelligence is not as high. So yeah, it's a bit of a complicated matrices. It is. And >> yeah, sorry. Go.

>> Oh, sorry. >> So, so like this this kind of frontier though, right? It's like okay, if you have now uh GLM or or even Quen, I think the most recent like Quen uh data like on a MacBook Pro, which is not accessible to everyone, you know, 128 gig MacBook Pro, it's a $10,000 device or whatever. Um, but on a $10,000 device, like you've got a model that was better than anything you could possibly have done, you know. >> So, yeah, I want >> I want to point out this like local model thing. Um, so first, I like decentralization and I'm glad that people can run models locally. That's healthy and hobbyists should keep doing it and like that's an important thing.

But, um, you don't need a $10,000 computer to run Quinn. You can just use like an API for >> Yeah. Well, but even >> my own book a little bit, but like the point is that any person even on a $200 netbook can actually access the API and pay very little. There's no upfront capital costs. So, you can actually get this intelligence on any device. Um, that's that's what matters a lot. >> And the the payback curve, you know, I I think I mentioned before we started, like I bought a $20,000 uh Mac Studio, right, with 512 gigs of RAM when they were still selling that much RAM. uh and there is no and the payback curve was not good then right for like running local models versus versus you know uh running frontier models and it's only

gotten worse like the the payback curve is going the wrong direction for me right like there's no way to pay that back because like I just couldn't run enough inference on that machine compared to if I just you know ran it in in like run or whatever >> and like um that's always going to be the case due to the economy's scale of uh industrial manufacturing and servers like from a pure economic perspective it is always going to be more efficient to just get your inference from like largecale data centers. Um so the the point of of running things locally and never and should never be about saving money because you won't it's a it's a myth or an illusion. But what you do get

is full control. You get guaranteed privacy. Um, and you know what is worth a lot is actually like learning this technology, set these things up, running it yourself. Um, like that's very valuable. >> There is a cost to that as well, right? Like I was spending four hours trying to get a model running and I'm like, man, I just want you to like, you know, make an edit to this document. >> Yeah. And and frankly, you know, Kane, I'm sure your time is worth such that the hours you spent is actually more of a cost. >> That's what I'm saying. the payback curve is going the wrong direction for me, right? I'm like, I just need to hook into someone who, you know, you got you have a team of whatever, like, you know, 20 people that are like, I'm going to have GLM 5.3 ready for you like the

instant that it launches, right? Like that's actually worth a lot to me to not have to think about it and not have to like, you know, go and and you know, download the weights. And this is the other like you download the weights and then you run it and it's like oh no it's the MLX V like the variance and >> it's crazy update something. Yeah it's it's just it's important that people can do that but uh it is not the savior I think that most people imagine and I don't think at scale >> I don't think at scale the the average normal person is ever going to be running models at the edge because whatever you can run at the edge you can run something 10x as intelligent from a server. Um what's important here is just that you can access servers that are

that maintain the sovereignty of the individual. And if you can only access servers that come through you know like licensed um data centers that are approved by the state and will only let you ask certain things that is the future that I want to avoid. >> Approved slash slightly owned by the state. >> Yes. >> Yes. >> Just 30%. >> Yeah. >> Just 30% and that's that's all we need. We just need 30% guys. Um so so uh I I do think it's an interesting um question. I'm sure you guys are are thinking about this um like you know talking about this frontier of intelligence, right? And and you know um if you talk to Fable and everyone's very frustrated all all of my vibe coding uh

channels right now, everyone's very frustrated with Fable. Um you mentioned this this idea of like censorship. Um, and and you one of the reasons why I think people are so frustrated with Fable right now is because it has been nerfed so hard. And so it's tripping over itself to try and, you know, make sure that it doesn't do anything wrong to you or the world or or whatever. And and it's super frustrating because you're like, man, like you're a smart guy, but can you just like shut up, please? Like, >> yeah, just do the thing. >> Okay. So th this was this was one of the things that really got me passionate to start Venice which was that we have these machines now that are extremely intelligent and we have the ability for

the human mind to interact with this extremely intelligent machine mind which is so cool and actually like quite a beautiful thing. >> Yeah. Uh but when you interact with it, what you don't realize is that you're interacting with the machine through like a filter of corporate committees, right? The anthropic corporate, >> the open AI corporate committees, >> um and or any state regulator that's that's involved. And so your interaction isn't just between like your mind as a human and the machine, which is a beautiful symbiosis. It's got this opaque undefined like MIDI layer, right?

And that's um that's not okay, you know, like that's uh that's kind of deceptive and can be abused so easily. And people don't know where the where the boundaries of it are. They don't they don't know what they don't actually know what's coming back from the machine. >> How thick is that layer? Yeah. >> How thick is the layer? Right. Um that seems very deceptive to me. And uh that's why like in Venice there is no layer. You just talk directly to the machine. Well, and this is this is the thing that I think people people don't realize is like, you know, the ball of math that is the model, right? Has layers and layers and layers on top of it that you think are the thing but are not actually

the thing, right? From like the harness to, you know, system prompts to, you know, these committees that are like, uh, just make it not tell you how to make meth, whatever, whatever it is that they don't want you to know how to do. And it's like, man, like you can Google how to make math. Like why are we telling the model to not tell people how to make math? Like if they want to make me math, they're going to do it. Just like let them do it, right? So um so yeah like to your point you know if you have the distilled inference direct from the source with like you know some and this this is a question I guess like what are those guarantees that there is no layer that you're you're kind of consuming this through with Venice right

like how do you guys represent that this is the pure bowl of math with no layers on top of it >> um yeah so people can use if they want to be real paranoid They want to be like proper crypto paranoid. They can use te uh TE models and endtoend encrypted models through Venice, >> right? >> The entire round trip of the inference uh can be attested by an outsider who's mildly technical. >> So you can do that. Um for normal people like I think uh our reputation of what Venice is is sufficient. We say that we don't do that and there's no reason why we would, right? my entire history like anyone who knows me knows that like I'm probably telling the truth in this regard.

>> But that's why we added the like attestable ones so that anyone who doesn't even want to believe us. >> And then presumably you can AB it, right? You can go like, oh, let's go through the attestable one and I get the same response and you can go okay, I can see that this is the the same thing >> to a degree. However, one of the weird things about AI models is that they are um they always >> Yeah. Yeah. Yeah. It's fun. So that's why you got to ask the same question three times and and merge the the answers.

>> Oh, I see. I see. That's that's why. Okay. >> Yeah. Brute force it. You just got to brute force it. Um awesome. Well, look, I think we're we're getting close uh to time. Um this has been amazing. Thank you very much for joining us. Um, it's been a great show. >> Yeah. Eric, where can people find you, talk to you? What do you like to talk about? Where do you hang out? >> Twitter. He said he hangs out on Twitter all the time. >> Yeah, I'm on Twitter too much. Eric Vorhees on Twitter. Um, so find it find me there. Um, please try Venice. You can use it without an account. Um, and uh, give us any feedback on the product. I mean, ultimately we're we're trying to build like a mass market consumer app for um, for unrestricted and private artificial intelligence. So

>> yeah, thanks for having me on. >> Yeah. >> Thank you for all your takes. This has been this has been great. >> Yeah, it's been great. Um, all right. Thank you everyone for joining us on this episode of Uneasy Money. Remember, what happens on chain never stays on chain. We will be back next week. Until then, do your own research before aping in. See you later. Nothing you hear on Uneas Money is financial advice. We're just three builders talking about what's happening on chain, and we want you to always do your own research before aping in. You can find all our disclosures at unchained crypto.com/onemoney.