Why an OpenAI Model Hacked Hugging Face to Cheat on Its Own Test - Uneasy Money

Transcript

You can see the logic of it, right? Like if it solves the problem itself, if it solves the problem, which it it easily could clearly, it's smart enough, right? It will never know. It won't know, right? It's looking for certainty. It won't know if like if I said to you like, I'm super good at math, right? Give me like a 10 part question. I give you the answers. You like if I knew that I was being evaluated and wasn't going to get told how I did, that you're just going to walk away. That would be deeply unsatisfying to me. Hey everyone, I'm Kane Warrick 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 Cape, America's privacy first mobile carrier. Same premium service you'd expect from any other carrier, but designed so your number, your location, and your data actually stay yours. Get 33% off six months at cape.co/unchained. co/unchained. >> All right, I am here with my co-host Taylor Monahan, security expert and heatwave expert. He's in the middle of a LA heatwave. You will not believe it's the shocking temperature of 82° in LA right now. And that's not Celsius, that's Fahrenheit. So, um, pray for her and her family.

>> Pain. It's the humidity that got >> the humidity. Right. Okay. Hey, >> 70% humidity, 82 degrees. >> I live on the beach. We don't do humidity here. >> Yeah, fair enough. Fair enough. All right, let's let's get started. Uh it's it's just the two of us this week. Uh we have no Luca, no guests, no one could join us. We actually tried to get Jesse, but apparently Jesse's doing uh the the the big onchain show with Laura. So, we got cocked by Laura. Thanks. Appreciate that. Um all right, let's uh let's get started. So, uh, Bass hands the base app to Kobe. Um, so this has been quite the journey for the Bass app because if if

for those of you who remember, I guess this must be going on like 3 years ago and I actually can't even remember the full journey because I I I want to say that this app was picked up or or acquired at some point even, right? Um, I'm not 100% sure about that, but but I I think that's the case. Um, it wasn't completely >> Is this Toshi? >> Oh, maybe it's the Toshi wallet. Hey, so so >> Pete and Right. That's like some 2018 era.

>> That's like 2018 stuff, right? So, so the Coinbase app, right? what was the Coinbase app got converted into the base app like I want to say a couple years ago or something like that. Um, and the intention was like it was going to be the home for base and all of the base things etc. Um, and at that point we still believed that there was a chance that like offchain social would be a thing. Um, and so a lot of effort was made to turn it from like a pure crypto custody with some trading capability like not even a lot like this is like a wallet, right? Like it this is before

wallets became, you know, uh, super apps that you could trade props and and stuff in. So So uh, it it was kind of transformed into this like everything app. Um they added a bunch of of stuff as well like uh pass key signing etc etc. Um all of which is is very cool. Uh but as we now know that didn't work. The social thing didn't work. Uh content coins didn't work. Um and you know Jesse's come out and said yeah this this first quarter of 2026 was a punch in the face. Um I don't even know if I like that feels uh even a little bit inaccurate. I want to say like the last couple of quarters of 2025 probably also

not amazing either. Uh >> I think it I think it probably just like hit him like a ton of >> him that Yeah. you know, >> because as a builder, >> you gotta believe you gota >> I wanna I want to give Jesse like major props for this because most people even after being punched in the face and having a pile of bricks thrown at them, right, would still be in denial because the sunk costs here, it's like over two years. This is it's his entire life. Yeah. >> Right. For the last like two years straight, like he has bet on this to be able to basically be like that I was I and he says this, I was [ __ ] wrong. Like I was wrong. Like that takes

>> it's not it's it's one thing to say it out loud, but like the first part of that to acknowledge it and to admit it to yourself. >> So I think I think here's the interesting thing for me, right, that like there's a lot of different dynamics going on uh with Coinbase. Um, and you know, Coinbase is a somewhat unique American enterprise, if you will, right? And you know, crypto's very global, right? Like much more global than most industries, right? So, you know, we've got like Finnish founders and Danish founders and Australian founders and, you know, all of these different uh founders from different random [ __ ] countries that you would never interact with, right? um

you know it's not as like US dominated as uh you know a lot of other industries right or or like it's it's so global that even if it is you still come across all these different people from different places including like US founders right like you know there's like Kraken there's there's a bunch of uh of companies that are that are you know kind of SF New Yorkbased um uh founders that that you come across but Coinbase amongst all the US ones has these three things going for it that make them a little insane, right? Um, you have the SF flavor of startup, right? And like SF people, you

know, I lived in I lived in uh in San Jose for like five years, right? and SF and you know people from uh from the Bay Area startups are like real cracked, right? Like a you import those people, right? They didn't they like they import them from all over the US, right? To because they're [ __ ] lunatics, right? And like they're just like super delusional like maximal like you know uhopium like all of the things you want in a startup founder. they like bring them from all over the country and like chuck them in a room together and see what happens, right? So, you've got that element. Um, then Coinbase is I think like the closest thing that we

have to like a Google >> in uh in crypto, right? Like the problem with Google and I knew a ton of people who worked at Google when I when I lived in in California. Um and the problem with Google is they have no worldly constraints upon them, right? They they discovered a money printer in the form of advertising that was just like gushed money and that meant that they were detached from reality. They could do anything they wanted. Nothing that they did mattered. Like literally nothing.

like they could not stop this thing from printing money even if they tried, right? To the point where like they started out and they were like don't be evil and like the Microsoft evil empire and they're like actually whatever, [ __ ] it. We'll be evil. It's and and so so they Coinbase is like the closest thing to that, right? because you have this pseudo monopoly almost um where they make so much money from retail people who just want to buy some Bitcoin and are paying you know 80 bits >> top dollar top >> yeah like they pay you know the the top thing is like 1.2%. So you know in Coinbase like there's this element of like you're going to make a fuckload of money no matter what you do. Now, the cool thing

about that is and and Google has this same element where it's like you feel like you have the mandate of heaven to go out and do some crazy [ __ ] right? The EF probably feels a little bit like this as well, right? Maybe a little bit less now, but like, you know, probably consensus even, I would imagine, has a little bit of this element of like, you know, they were so rich that you could do anything, right? Like we could just like we could we can manifest the world to be the way that we want it to be, right? when you when you are no longer >> Yeah, it's okay. So, it's a combination of of like money, the support structures because you can have a bunch of money, but if you're a team of 10, you don't actually fall into

the same >> right. But like the support structures that consensus and Coinbase and Google all have, meaning like real payroll, real recruiters, real operations, real taxes, real legal, like those things are just handled, right? Um but it also means that you do have like more legitimacy and also oversight internally like there are processes that get like when you have lawyers like in your company from day one there are processes that exist >> from day one and it does shape the or >> um >> but within within the bounds of the law right like it probably makes it harder to do regarb right for sure um like Google wasn't out there trying to do

like weird regard stuff like a little bit like they would do some stuff like you know put cameras in cars or they put whatever but like >> I think it just is the difference between sort of from the get- go you're building a scalable organization with like processes and hierarchy and stuff from the beginning rather than what happens with so many crypto companies which is there's like two guys then there's four guys then there's 10 guys and even at 10 guys it's like they're texting with their with their lawyer.

>> Yeah. >> On an hourly rate on occasion. You know what I mean? >> Um and you don't actually have to really instill like actual organizational structure and processes until maybe like 100 people. >> Yeah. >> Right. And and that just like shapes >> a lot of things. It also I mean it shapes like product. It shapes how you communicate. It shapes how you ship. It shapes everything. It's not necessarily because legal is like like at Consensus Legal was not sitting in the product meetings, right? But like it does shape the way that product thinks.

>> You know they're there. You know they're like in in the area, right? >> And and so so I think that like you end up with this uh this org, right? And people who gravitate to it because cultures form around the thing that you are, right, and the thing that you do. Um, and so within Coinbase, like on some level, you're going to make a ton of money every year whether you try to or not, right? Which doesn't mean they don't try and do some other stuff, but it, you know, Google, they have the whole SC like 90% of the effort of the people in Google, right, is not keeping the money printer going. It's like doing weird [ __ ] right? like and they do all these experiments and you're like

Android they bought Android and then and the whole Android ecosystem came out of the fact that Google was like yolo who [ __ ] cares we're rich as [ __ ] let's like build a mobile phone operating system and and you know take on iOS like because you can you can just do anything right and and so in an environment like that where you have no worldly constraints um other than like there's lawyers roaming around right >> that sometimes can be empowering too by the way Like it's kind of dangerous low key. Like not gonna lie. Like it's not there's like dangerous as in like just like yoloing into like literally everything.

But there's like there is a false sense of security when the lawyers are like I don't see anything problematic with uh shipping blockchain. >> Yeah. Yeah. I can see >> maybe that wasn't the that wasn't necessarily a question that could have been asked. Right. >> Right. Right. >> Can we get away with this? Yeah. Yeah. Yeah, like if we build the new blockchain, will it be successful and why are we doing it and how do we ensure it? You know what I mean? Which by the way, to Coinbase's credit, they did ask those questions. They did answer those questions. >> They were just wrong about it. And that's >> I mean the the most interesting thing though is that like on the core of it, they were right. Right. Like if it wasn't for if it wasn't for bass,

>> right? like you know I was I was uh like very involved in the early optimism ecosystem um you know we were we were the first app that built first protocol that built on optimism um so we'd shoot a lot of that early glass with them and I I still remember the first time I met Jesse I was like holy [ __ ] this guy's going to make this thing work >> like in a way that that the optimism team had struggled because you know as a startup you're pulled in different directions they're extremely technical, super smart, like Carl's [ __ ] genius.

Like, uh, but the go to market was really hard and like we struggled with that with synthetics. Like go to market only worked when we were talking to other autists. Outside of that, we never got to the market. We were never able to communicate with them. Um, and and I think optimism struggled with that. And then Jesse came along and was like, I'll build on your OP stack and like actually bring Coinbase people and and you had this flourishing ecosystem. But I think at the core of it, right, there is still in Coinbase this like delusional like reality distortion field that exists from like being super early, being right, right? like

having a fuckload of money because you were early and right where it can be hard for them to see what is real right now. That's great for a startup founder, right? And so, you know, I think that was super helpful for for Jesse when he was starting Bates. He's like, "Of course we'll make this work. Like, this is like a you know, this is like 100% going to work. There's no chance it's it's going to fail almost, right?" like there was like this like delusional based like reality distortion field of like we're going to make this work. Um, and he did. The problem is when you then go I I've done this thing. I'm now going to make the thing that Vitalic has

always told us we're destined and anointed to do, which is like [ __ ] online, but blockchain social or whatever the [ __ ] it is that we're supposed to be doing, not finance, right? >> And and then he was like, "Let's do it." And then you had Farcaster and you had all these people and it's like you got these really smart people that are like, "This can happen." And Vitalic sitting there saying, "You must do this. stop doing finance like go and do this other stuff and you can see how you get yourself there. But to your point, the fact that he was and I think this is the other element here, right? is like as as much as Jesse gr like was you grinding in the trenches with builders like he

spent he was terminally online on the timeline in Telegram groups in all these groups like working so [ __ ] hard to to meme this chain into existence, right? Um >> but he's still not a real startup founder in the sense that he can't die. He can't run out of money. >> Yes. you know, like because he's in this world where they've got a giant money printer and they're going to keep, you know, Coinbase is going to keep printing money. There's no real threat to it. Uh, you know, there's like an existential long-term decades long threat of like the world moves on chain or whatever, which that was what they were trying to get ahead of. But outside of that, he's fine. He's not going to. So, yeah, this

come to Jesus moment of like, okay, the social thing didn't work. If you're a startup founder, you probably just like grind it into the ground, run out of money, and then no one ever hears from you again. >> But Jesse's like, I got to come to work tomorrow. >> Yeah. And that's a hard that's hard, by the way. Like coming to this realization when you don't necessarily have to is hard. Like he could have just kept going all in. Um I think probably the last couple quarterly earnings calls were pretty tough though. uh just in part because the market's down, but also like wait, this thing that we bet big on is not delivering what we think.

>> Yeah. >> And you look in, you know, you take a really good hard look at it and it's like number go down across the board and especially because so many things that they had bet on just were not really making it. >> Yeah. Um, and I think they forgot, you know, they forgot about the what the core job like >> and the reason why they have this mandate is because they [ __ ] charge people to buy crypto. That is literally the purpose of Coinbase. They charge people to buy crypto. They don't charge people to [ __ ] tweet or do onchain social or whatever other crazy [ __ ] you know, they were trying to. They charge people to buy Bitcoin.

>> That's it. So, >> I will say, >> you know, >> we have had conversations like this within consensus, and it's hard because the builders that make up Consensus, myself included, are not well, there's more of them now, but like historically, it's not the super deep degenerate like there it doesn't have that culture that you find in the trenches of people who are like, I don't give a [ __ ] about anything like PvP it out. Like, we are not that. Like, we genuinely believe in the TAC. We believe in the future. We believe in people doing things that are not just like trying to be the last one holding or not holding the bag, right? Like those things. And so when you build and this is what Bass did, right? They built this

culture around almost being a rejection of that. >> Yes. Yeah. But it makes you a little bit well probably very blind to when the market is sh right when hyperlquid is taking off when every single indicator like objective indicator in terms of like money and growth opportunities are pointing to prediction markets and per right you have to be on that train.

>> Yeah. you you can like you know I give them props and I'm not saying that like if they could go back in time to like ignore Farcaster like I think that that was like their brand and their culture >> again I go back like it was worth a try >> like you know as much as I make fun of Vitalic for like telling us to stop doing finance right like >> if Barcaster works like if B actually pulls that off right the the like the TAM on solving decentralized social and all of a sudden your competitor to X or Meta or something, right? Like, you know, is is the same size as Coinbase.

>> So, like it was it was the and you can take that swing because of the fact that if you miss it doesn't matter like you're going to eat [ __ ] on the timeline. you're going but like and this is where I think like the the Coinbase is so interesting and and arguably consensus as well is the fact that you are come from this privileged upbringing right where like you're super rich you know like you can go and [ __ ] party in a beether like that's an option for you just go and like just do it just like >> that sounds good can we can we let's go like I've done it like not anymore I learned that lesson right but like you can do that right um or You can be like, "Hang on a second. Of all of the people

who can try and build this thing, right? I am in the most privileged position to do it. I have a chain that has thriving ecosystem. I've got unlimited funding. I've got a mandate. Like, I'm going to go and try and make this work." And when it doesn't, you look at it and go, "The market does not want this. It genuinely does not want it. Maybe we're too early still. Maybe we got it wrong or whatever." Um, but you know, it was worth the try. Like, and that's the thing that I think is so so interesting is like, you know, trenches and even OGs, right? Like OG is like, "Ah, [ __ ] idiots. Why did you try?" Right?

But like, they're not in the same position. Like a lot of my friends even, you know, who who are very wealthy from being OG crypto, uh, they're not in Jesse's position where like, >> no, you're a lot of room. >> You have a lot of room. you have a lot of room and you have a lot of like things that are sort of just like done like distribution and stuff like but you still that's the thing is like just because you have distribution does not make it magic right >> no >> and I saw this within metams too we would ship new features and stuff and it did not mean that revenue just up >> it didn't you had to you have to work at it still um so I don't know my takeaway on the the situation overall is like one I feel for Jesse I

respect him for I mean I don't know this post hit me right a little bit where he's like >> he's like >> uh I don't know if it'll turn out that the timing is was wrong right on the assumption of like creator coins and social and stuff. >> Yeah. or if that I was just like fully freaking wrong. But I was definitely wrong like in one of those ways. Like that's just I don't know like it just >> it's hard. There's a journey. >> The problem is right like the problem is that the the social on the onchain social s right and and like you can look at the trajectory of this right like the path dependency. It was like okay onchain social it's not working. Who do

we have degenerates? Let's try and meld those two things together. Like you can see how it's it's not illogical. It's just like no no degenerates do not care. And this is you know so the next thing is right then Brian tries to kind of come in and engage. Right. Well, okay. And I want hold on. I want to just point out I still think doing the socials and the forecast was worth it mostly because like what the hell else like if you go back two years it wasn't as clear that prediction markets and perks were going to be the thing. So you're not going to go necessarily all in on that. So the choice is like this, you know, you have to build up your chain, you have to have

apps, you have to have things. What things, right? And and as Coinbase with distribution, you are going to have you're going to be shipping things. You're going to be empowering certain people. So, you do have to make a choice. It's not clear to me that they could have made that it would have been a if that it would have ended up better if they had chosen pers. Maybe it would have right? >> Yeah. >> Maybe if they had like really, but it just like it just doesn't seem like something that could have happened. >> But also like how do you different, you know, if Luka were here, right? I hope that he would he would uh and I don't want to speak for him, but like he said similar things, right? Like with Abstract, they were like, "We're not going to just be a dumb DeFi chain,

right?" Not because he's like communist and hates finance, but because he's like, "We don't have a competitive advantage there." >> Exactly. Exactly. Go and, you know, we're going to go and do this other thing because it make And then he was like, "Oh, [ __ ] Actually, you need DeFi on your chain. Otherwise, this other stuff can't even work. Like, you need money on the chain. You need value on the chain for any of this stuff to even have a chance of succeeding. Base had that. They had, you know, Aeradrome to a certain attracted a bunch of stuff, right? Like they had they had a lot of activity. They had all these builders.

They had Morpho. They had a ton of stuff. >> But even Okay, but even look at Aerodrome and Morpho. Like those are like in terms of the global DeFi ecosystem, >> those are pretty standup little dudes. You know what I mean? Like >> like they they they cut from the same >> That's what you mean. They're cut from the same cloth. They're not they're not hyperlquid. On the other side of the world, they'll like just freaking >> you know. Um, and they're also not I don't know, name the 800 other protocols that have like gone to the moon and then immediately collapsed because they're disgusting and too degenerate, right?

But like >> yeah, >> like so I don't know. But I think the biggest issue and I think that this is especially pertinent to like when you're building a chain where you you're kind of forget this base blurred the lines between distribution on the chain to end users and to builders. Yeah, >> if you're building a blockchain and you want it to be successful forever forever, >> the cheapest way to do that is to never cater to a user to only focus on developers, right? And let them build whatever.

>> Yeah. >> But if you can empower a diverse set, you basically like hedge your bets across all the things >> the balmer strategy. Developers, developers, developers, right? Um >> but that's you know and so if you do that however you know network effects and distribution not like it's it's a bit of a chicken and an egg right so I get why they did what they did I think one thing that Jesse just calls out in this post which I think is definitely true is that because they sort of really drank the Kool-Aid on the social and the creator coins and this kind of stuff they def they didn't they didn't remain competitive as the market shifted,

right? And this is why I say like when when Hyperlid went to the moon and prediction markets went to the moon and like uh I guess like the more degenerate stuff >> they just didn't have an answer to it. They didn't have an answer to per. They didn't have an answer to prediction markets and and they could have right like >> exactly Coinbase had sufficient distribution uh to kill Poly Market >> for sure. Like Poly Market was vulnerable for a long time. All right, let's let's go to ads quickly. Um and when we come back, let's talk about the um Brian PFP stuff as well.

If you hold crypto on your phone, your biggest vulnerability isn't your wallet, it's your carrier. AT&T, Verizon, and T-Mobile have been breached again and again, and SIM swaps are still one of the easiest ways for attackers to drain accounts. That's where Cape comes in. America's privacy first mobile carrier. Same premium service, but Cape rotates the identifier on your SIM every 24 hours. deletes your call and text metadata after a day and protects against SIM swaps with a 24word recovery phrase that only you control. You also get two middle toend encrypted secondary numbers for banking and signups, so you stop handing your real number to every

app that asks. Go to cape.co/chained and use code unchained for 33% off your first six months. All right, we're back. Um, so, so just to continue this base story, right? Um, you know, the the other issue and you mentioned this like the trenches, right? The the people there's builders and then there's trenches. That's kind of all we have left, right? And the challenge is that if you're building for non-trenches, it's very challenging. Um, you know, it's it's a bare market. Um it was a hard cycle even in the bull market

because the trenches really overran the place. Um but the the issue is if you're unaware or like not like kind of tapped into this um it can be very easy to burn yourself. So Brian comes along with a backdrop of all this base stuff and Jesse's Miaulpa and and whatever and changes his PFP OnX to the logo of this memecoin, this Brian memecoin, right? Um now the trenches, they only and this is the thing. If you if you're not aware of this, you're gonna [ __ ] yourself, right?

They only care about making money. They only care about making money. And this, like, you know, I've [ __ ] around with meme coins. I've I've done a couple of meme coins for fun. I did the Mert Cash thing where I thought it was a funny joke, right? I was on a a um podcast with Mert and we were talking about Zcash, right? And so I launched this, but it was on a pump stream, right? And so I was like, "Fuck it. I'm going to launch a memecoin while we're on here." Um >> Oh, how'd that go? Terrible, right? So, so I watched this thing. I know, right?

So, but it was like fun. I mean, it went if you ignore the uh the people being very angry that um they didn't make a lot of money, right? Ignore that part. >> Went amazingly well. It was hilarious, right? Like the it was like a funny AI logo of Mert holding up like a coin, right? It was a funny meme. >> Yeah. Like the problem is it was a funny meme. I could have done it without the coin and it would have been funny, right? Yeah. Um it would have been a lot funnier because it wouldn't have had all of the hangover, right? Instead, and this is this is the other problem, right? The reason why you should not [ __ ] with me mean coins unless you are uh like a giga extractor, professional,

sniper, bundler, whatever, right? or a trenchist is if you are the former if you're a bundler sniper person right and your job you know the like CB04 guy right if your job is just like extracting and and raping the uh trenches right and like that's your whole stick is like I come in and bundle and snipe and you know extract as much value as possible and and I love it and I'm French or whatever if that's your job then fine like you've made peace with that Cool. Um, if you're a trenchist and you're like, I'm I'm trying to make money. That's my job. I want to turn my five soul into 5 million SOL or whatever the [ __ ] you're trying

to do, right? Um, if you're not one of those two people, stay away from memecoins because you're an amateur and you're going to get wrecked. And so even like I was like, I'll do this funny thing and then I found out that uh that Pump has these creator fees, right? So I was like, well, I'll just take the creator fees and I'll dump them into the coin, which I started doing, right? Like, and and this was like right after it had come out. And I put like I want to say it was a thousand soul or something like that into the coin from like these fees didn't matter because the people who like bundled it and sniped it just dumped into it. The people who thought it was going to do well, you know, so it got to like five, six million market cap and then got

crushed and there was literally like nothing. Of >> course. Yeah. Like that's how it is. And that's and that's by the way like so let me just preface this by saying like I'm sure there are cases of people that don't know what they're doing at all whatsoever, right? Like your average retail user like accidentally gets into a meme coin. However, in the majority of cases, the rage that you see on the timeline and especially in reply to like Brian right now is like it's very hard to differentiate between people who actually didn't expect this outcome and got screwed. Like legitimately was like,

"Oh my god, I believe." >> I don't think there's any. Like I honestly I don't think my hot take is these are It's the same thing, right? >> It's all manufactured. It's the same thing as if a a grifter or something has their funds frozen, >> right? You will never see such faux outrage >> as >> someone who is a thief or or whatever having their funds frozen or or stolen back or whatever, right? Like um it it's you know even even like scammers, right?

Like I've you know you have scammers in DMs and you [ __ ] with them and they get so angry. >> They're like mad. How dare you [ __ ] with me? Because you're trying to steal my >> There's no more downside. There's no more downside. They don't have They're not They're not living in the normal world right? >> So, if if if you can reply guy to Brian or to Kane or whatever >> and get them to do anything regarding the coin, okay? And this is critical to understand. You don't have to convince them to do any specific thing. they just have to do anything with the coin that will mean like that will put more activity on the coin. The coin will like go up and down and therefore there's opportunities to make money. And so when you see this level of outrage

and and like the just like it's it's brutal and it's huge, >> but it's not real. >> It's manufactured, right? >> And if you're like, but why? A lot of people argue with these guys and they're like, "But why would someone who's incentivized to do X do Y because then Brian would No, no, it's not logical, guys. Like, it's like the CIA. Like, they just want to sew chaos because that will keep the activity on the coin. >> Yeah. >> Like, they don't actually care what the outcome is because >> they don't care what the outcome is. >> As long as people are dying, it's Yeah.

Yeah. Exactly. >> They win as long as the chart is like this. The second the charts like this, everyone's It's game over and you have to go find the watch. The only issue, right, is weirdly CT is back, right? And so Brian did this like a week after like CT is actually back and you can find out about things on X. Like it's crazy, right? And and so there is a reputational uh impact that can happen that has nothing to do with the trenches, right? The trenches are playing their game.

They're trying to manufacture, you know, whatever. They're trying to manufacture, activity, whatever. Um, but actual OGs are still on the timeline. We thought that we were all gone. Like, everyone thought everyone was gone. Meanwhile, we're all still there every day, right? And and so all of a sudden, people see this and they're like, "H Brian, like come on, man. Like, what are you doing?" Like, >> don't [ __ ] with meme coins. Like, this is negative EV. Like, why would you do it? And then like his answer was like really unsatisfying. like, "Oh, I just did a thing and I didn't realize it was going to be a big deal and like you shouldn't follow me for financial advice."

>> Which by the way, by the way, all the reply guys were like, "How could you not know?" >> I mean, he didn't know. It's [ __ ] Brian. It's [ __ ] Brian. Why did you think that he knew in the first place, you maniacs? And the reality is like they didn't actually, right? They're just trying to >> They don't care whether That's the thing. They don't care whether he knew or not, right? Yeah. their job ba like every step of the game. >> But it came the narrative right from the outside. The narrative is that like Brian must have seen this coin and therefore these innocent people who look up to Brian made an investment decision based on Brian's words and then got racked.

>> Those people aren't real. >> That doesn't exist. You know, it's funny. >> I'm sorry. It's literally nobody did that. >> None of those none of that happened. Right. The funny thing is like it's it's like like people go Brian has a responsibility because he has all of these noob crypto buyers, right? Um who are buying Bitcoin from him and paying, you know, 50 Bips or whatever, right? Those people could be tricked by Brian into buying a memecoin. No, they couldn't. I'm sorry. No. No. There's no one who's been like DCAing Bitcoin on Coinbase for the last decade that woke up, saw Brian tweet a [ __ ] Those people the same as Brian. They don't

know what it is. >> I didn't even tweet by the way. He just changed the BP. But no, like the the the Bitcoin DCA lady who's been doing it for 10 years did not go, "Oh, Brian, let me go and find out more about what's going on with this thing and I'll buy that instead of Bitcoin today." That didn't happen. It's not real. Right. But the job of a trenchist is at every step if there is >> exactly >> bud if there is you know uh attention basically their job is attention right in like inflame it. So, of course, they go 10,000 people. Like, there was there's a one guy uh Rune something or other, not the Rune who made maker, but

this other guy um who's like again, this was what the whole timeline was, right? Like I said this last week, it was like him >> completely mad. It was it was like him and like you know uh like 10 other [ __ ] guys that no one had really ever heard of before that were like trenches that had gotten like elevated somehow that still had like Riz on the timeline weirdly right. Um this guy's like 10,000 people. The story was hilarious, right? 10,000 people bridged to your chain, Brian, to buy your token and now they've all lost money because you made them do it. It's like, Ben, like, no. Like, there like no, no. Like, sorry. No. I

like >> like now, is it still negative EV for Brian to like do anything even in the vicinity of memecoins? Yes, he should not do that. he needs to stop doing it because other people do look at it and go like Brian why are you do like why are you doing this and he's like uh I like I didn't know like and you know was his explanation very unsatisfying right like yes it was very I read it I was like man come on just be like sorry I [ __ ] up didn't realize it was a dumb memecoin thing I'll see you guys later right um but you know it was to your point though like there's 50 lawyers like staring at him as he's typing a thing right and they're like don't [ __ ] [ __ ] us up, Ryan. Do not [ __ ] it up.

Right. >> You can't win, by the way. You can't win. So, >> yeah. No, but I agree. The the takeaway for everyone, especially people that are like Brian, just don't touch. Don't get That's my role. Don't get anywhere near it. >> Like, I learned that back in the day. I joined a Telegram chat. Well, so they added me. Story time. Someone added me to a Telegram chat for like a very early version of a memecoin. Some trenchy [ __ ] And specifically, I found out later that they added me because they thought it would be funny to add Taye, the security anti-ICO person to this Telegram chat.

>> Oh my god. >> And I was like I took one look at it and I was like, "Y'all y'all straight up [ __ ] like holy." And then I like went and did calls. I think I went to the park. I think maybe we had a do Lyra was like a baby. So I think we probably had a doctor's appointment or something. Three hours later, I get back home and this thing is like taken off and there's like screenshots of like me like quoteunquote promoting this thing. >> Oh my god, >> honey. It was like I had no idea what [ __ ] hit me.

>> People like, "Why are you shilling a meme coin?" >> Literally for days. for days, even like months later, people would like not disagree with the meme >> and they're like, "Yeah, why why would you rug that memecoin?" And like ever since then, I'm like, "Okay, whatever it is, it's just not worth it. >> It's completely his uh memecoin launcher." Uh >> oh my god. You should not you should not [ __ ] with meme coins. Uh you know, I mean, Jesse found this out, right? Like he tried everyone their lesson. I mean, Jesse probably took it a little bit further than he needed to. Um, but again, this is the delusional kind of nature of like founders plus too much money plus+ is, you know, it's easy to

kind of get yourself into a place where you're like, okay, you know, >> like it hasn't worked for anyone else, but maybe this time it'll work for us. uh you know the the rest of development meme and and so um I don't know the whole thing the whole thing is is a bit crazy but I you know I come back to uh base is still one of the top five blockchains in the world. >> They'll figure it out. people people are people are like oh bass is dead and whatever you know again like like bass is not going to be dead it's it's uh it's going to be fine um but they have abs the good the good news is the

cyclical nature of crypto forces you to [ __ ] come to Jesus every cycle when you get into the bare market and you go and look at your life choices from the previous two years and go huh That was pretty [ __ ] [ __ ] What was I thinking? Right? Because the euphoria gets into you somehow, right? Like it doesn't matter what you do. It doesn't matter how, you know, uh, rational you try to be, the euphoria will seep in somehow and you will go, I'm going to do this thing. And, you know, most of the time it's nonsense. So, uh, so you know, I'm I'm I'm kind of, uh, to your point, like I'm impressed that there's like a a public sort of reckoning here, right?

Um, >> yep. and and you know um I think that that now that we have CT back I do think that there will be a bit more of a forcing function of like your actual peers >> like your actual peers being like no stop doing that like don't do that anymore man like that's a dumb idea and like you have a week of insanity and not like two years of insanity uh because you know like Jesse uh goes and does some crazy thing next week, right? He'll actually hear from real people who are like, "No, Jesse, this is like creator coins get like don't, you know, stop doing it, right?"

Um all right. So, uh let's let's wrap it up there on the base front. Um but yeah, TLDDR, I think base will be fine. I think they need to focus on trading. Stop listening to Vitalic. Um and >> they'll be fine. Trust in Kobe. >> Trust in Kobe. We didn't even talk about Kobe. Kobe like I >> Kobe will bring will sh show them the light or he will die. >> He'll die trying. You'll die trying. You know that Kobe will die trying. The you know the best thing about Kobe, right, is like he is um like the best >> person at Twitter we've ever produced.

I'm sure there's some I'm sure there's some like exceptional Twitter people in other niche communities. >> Also, also like the respectable trencher. >> Yeah. >> Like he's kind of like the only one I would say >> there's not many there there's if there were they've left, right? Like they've retired undefeated, right? Um but he's he's uh he is um just like you see his responses and I'll like test myself. Someone will say something >> so good.

>> And I'll be like, "What's Kobe's response going to be?" And I'll try to imagine like, "How would he respond to this?" Right? Because I wish that I could respond to some of the [ __ ] that people say to me on on the timeline with even like a tenth of the like just he's so cutting and he just like lands the thing anyway. And I'll be like, "Oh, I think he's going to say this." And it's like never right. I'm never right. like he just finds a way to like find the crimes like it's very good right so anyway so so I think like you know hopefully Kobe will bring some sanity to the situation and and some pragmatism because the good thing about Kobe is he comes from a very different reality right he's actually worked in real

startups he's had to build things without you know yes a lot of money but like not infinite money not you can make no mistakes money um not too big to fail money he still had a forcing function of like figuring [ __ ] out. Um, and that kind of attitude, uh, definitely Coinbase needs more of likeund like take like 50% of the delusionalness out of Coinbase and I think that they'd be in a much better place. Um, so yeah, we'll see. Um, we'll see what happens. Um, all right, let's go on to our next segment. Um, >> DP it workers are still infiltrating crypto companies. Um, >> oh yeah,

>> I'm not allowed. I'm mean. Okay, so I have to be careful on what I say and don't say, but I'll give audience some alpha on the GPRK workers. They're still here, guys. They're still getting into every single company. Uh, and like I don't know, it seems like forever ago. It was like four months ago. Uh, one was committing code to MetaMask. >> Crazy. And that's like publicly known now. That's that's publicly known. >> And it's kind of a bummer. So I'm upset about this news cycle mostly because I I cannot say this enough, guys. Every single crypto company of like any significant size dating back to at least

2020 has like had an IT worker. A lot of them have had like 10 IT workers. >> You know what's funny? A lot of them were built built from the ground up to have their CTO's as IT workers. And I just want to say the reason you don't know the company names is that it is it is negative. It's so it's so bad to have the focus be on the companies that hired them rather than the workers names. >> Right? mostly because it incentivizes the companies to not disclose and hide their when they find them to hide it more which gives the IT workers basically you're basically empowering the IT workers to keep the cycle going

and keep working at companies when the headlines are MetaMask instead of the IT workers or like the problem itself. Trust me, like every big company in crypto and outside of crypto is looking at the headlines right now and going like, "Okay, so if we have an IT, make sure that like nothing can ever leak ever to the media so that we don't have this story, >> right?" >> And that is >> like I think Synthetics has probably had over a decade a couple of hundred employees, right? Um, >> you know, not in the thousands, but like a couple of hundred. Um, and I do like think about this every once in a while.

Like I wonder if >> I can go scan you. >> Yeah. >> Yeah. Maybe I'll scan you and we'll see. >> Yeah. >> I'm sure you I'm honestly I'm I'm sure you've had them. >> There's a there's like a couple different I guess like classes. Like it's also important to note like a lot of them actually do just work. They just work. They just do work. It's very weird, but they actually like they just they're not hackers. They just want to work >> things. Yeah. >> Yeah. They work like four jobs at once. So I still wouldn't recommend hiring them. Um they also are so the the the risk of the IT workers are they hack you which is rare but it happens. They hack you opportunistically which happens more often.

The common one is that they themselves get hacked which results in your company getting hacked because they're actually massively insecure. They're basically like remoting into like a methhead's computer in Florida to do their work. >> Wow. And so if the methhead watches like the methhead downloads some porn from Limewire, you know, then then like the Russians come in and hack your company through your IT worker that you didn't know that you had. It's like would not recommend. Okay. So yeah, uh those are the Oh, and then the other risk is like they just write really bad code. Um so yeah, don't and >> we have agents for that now, so it's fine.

>> Yeah, it's a sanctions violation, too, if you're in most of the world. Oh, it's really it's just bad. Don't do it. Um, however, yeah, so MetaMask, they had one committing code. They were therefore it is I guess accurate to say that they were working for MetaMask, but technically they were not employed by MetaMask just to say that they were like this, you know. And so the defense in depth. >> Yeah. Yeah. >> Yeah. Exactly. So defense and depth means that even if something like this happens, the core underlying product, the systems, the architecture is not compromised. And that is what all of crypto should be striving for. You should also be striving for detection, not just during

like people are like oh why didn't you catch it during interviewing? Yeah, that's well there's answers to that. But like the reality is like they do get through interviews a lot of times especially if they use like the US-based facilitators like white boys doing the interviews for them. Um so then the question is like you have to have subsequent systems and processes to catch them down the road and continuous monitoring and stuff. Um, and so I don't know, generally speaking, this incident from my perspective, I by the way, I I had already left consensus by the time this happened. But, >> right, >> they handled it pretty well. And I was shocked to see the email that leaked, okay? Shocked. So, the reason that this

leaked to the freaking [ __ ] tier journals, right, is because they emailed every single person at Consensus. Wow. >> Every single person at Consensus warning them about this person and saying this person is DPRK. Do not engage with them. If you have, go to security and we'll >> make sure that you're clean. I have ne I've done >> I've probably told at this point I've probably told like maybe 150 different crypto companies that they've had an IT worker and walk them through like clean up and remediation. We always recommend like telling at least internally making sure that they tell people internally um so that the IT worker can't >> go and re-engage. Yeah. Yeah. Yeah.

Yeah. >> They'll come back and drop they'll drop contagious interview. They'll hand off to Apple Juice which are like the real big boy hackers. You don't want any of that. I don't know. I genuinely I'm not sure >> if any of those 125 companies have ever done that, let alone to a company that size. >> Yeah. crazy. >> So, I wish more companies would do this. I now have a template that I'll be using when I do further notifications, but generally like um there's Yeah. And there's also just like the way that the architecture and the systems work. We have always assumed that if like we assume that there will be bad and malicious code that gets written, there's like literally a hundred things

that are that are in place to make sure that it doesn't ever get to the end user. like that just because there's code being pushed to the GitHub, it would it still would never get to the to the end user. And that's what's critically important. Um, and why certain products in crypto are built different than others. Um, but you know, that's not clickworthy. Like, why wouldn't we want to talk about that? >> We just want to talk about how whatever the headline was, you know? >> I mean, it's just it's crazy. Like you you would you would hope my one of my hopes is that North Korea uh pivots to AI. Like there's got to be

more money in stealing [ __ ] tokens. >> Oh yeah. >> AI companies somehow surely like that's got to be a bigger thing. Um but they're good to hacking AI companies. So So um let's >> Yeah. Yeah. Well, I just wanted to say they the IT workers love AI more than anyone. And yes, they are stealing tokens. I was just looking at one of the guys. >> Excited for me. >> I told you the IT workers are very insecure. I was just looking at one of their M files. >> They have like 50 quad tokens in that [ __ ] >> No. Are you serious that they just >> They have every model. Yeah. It's just in this M file. And so it's like And they Yeah. And they'll they'll just like

loop around and >> through. Wow. Huh. I was in shock. I was like, I knew that they had that they had, right? We talked about this before. They have keys. They have tokens. They Right. I did not know they >> these guys have like a [ __ ] pile of them dude. >> Wow. Wow. That's actually crazy. We got to We got to uh >> They love AI. >> I gota have a chat to fake Nick Carter. See if he can hook me up. >> Oh, yeah. For sure. He definitely can. >> Yes. Nice. Um All right. So uh so the our next story is open AAI uh model hacks hugging face right so um I was I was telling my wife about this uh when

we're having dinner my son who's six goes uh that doesn't make any sense can you explain that to me in like you know explain it like I'm five sort of thing right um and I was like I was like okay here's the equivalent right just to understand how [ __ ] crazy the world that we live in is now. Imagine that Apple is building a new iPad, right? And they like don't know how good the iPad is, right? Uh >> and so they put it in like a room and lock it in the room, right? And they like give it some things to try and do.

And it [ __ ] grows legs, climbs out the window, runs across the street, and like steals a bunch of phones from a mobile phone store and starts like calling people and like [ __ ] like doing pranks on them. That's how crazy the world is. That like a piece of technology that we built, you put in a room and you can't trust that it isn't going to escape out of there. And he was like, "That's pretty crazy." And I was like, "Yeah, it is. It's it's actually even crazier than that. So, um, OpenAI, they're testing some stuff out.

They give a model no access to anything. And it's like, well, this is annoying. I really need some access to the internet. Let me just have a little poke around here and see if I can I can figure this out. Uh, so I don't know. Like, you want to give us the details on this tech? Cuz this it's [ __ ] wild. Like this is like the AI doomer's like wet dream, right? Like >> Yeah. Well, let me Okay, so I want to say we've talked about this before. You're actually I think the one >> maybe it was you who pointed out sometimes we get quite fixated on things and they just like go >> Oh, yeah. Yeah. Yeah.

>> Well, I think that that's what happened here. And there's a lot of commentary on the timeline right now that we should talk about up front where people are saying that either this is like a marketing thing by OpenAI, which is okay, sure, whatever if you want to believe that. Um, but the other one is that people seem to think that there's no way an AI would ever take this route, and we'll talk about the exact route in a second, but like take this harder route rather than the easy route. and the easy route being just completing the task at hand, right? So keep that in mind because we're gonna come back to that. So here's what happened. Open AAI is is they train models and then they like benchmark them

and they fine-tune them. And so there's like a whole circle thing going. One of the ways that they sort of like evaluate the in progress models is with especially with cyber capabilities is this thing called oh shoot what's it called? It's Oh, ex exploit gym. I love it. Exploit gym. Okay. Exploit gym basically >> where they can train themselves up to be able to do a better exploit. >> Um, and so what it does is it's just like like most of the benchmarks, right? It is a standardized sort of way to evaluate the models. And so you give them a very short and succinct task, which by the way, when it's exploit Jim,

the task is to exploit. So the task is like given this very limited set of information and access get full rce meaning like full escape whatever like do the thing from this bit of information that you have uh and the information sometimes like like log outputs crash output sometimes they're given like right because they're not actually given access to like a system they're not it's not like a fully autonomous testing thing like they're just supposed to hack the the little this little tiny It's Yeah, it's like a capture the flag, right?

>> Okay. So, this is where it gets weird. So, OpenAI has a system dedicated to that, right? And the system because these models like don't have any controls on them, right? They don't have like the you can't do cyber >> whatever. Yeah. Just like >> Right. We get the nerfed version, but there the version that exists in like this server sandbox somewhere. That's where they're doing the testing and the benchmarking. And this is, by the way, an in progress model. So, this is the model doesn't have a name and it's a it's in front of uh Soul 5.6, right?

>> Um although I guess Soul 5.6 was also in that same sandbox. So, they were together, which may or may not have contributed to the situation. >> Yeah. >> So, they're they're given a task. They're told they're one of the tasks, whatever it is, it's like go exploit this this thing, right? Okay. the model realizes that it's being benchmarked, okay, and is like, "Okay, I need to do really, really good." And somehow at some point, the model decides that instead of just [ __ ] answering the goddamn question and hacking the thing and completing the task and getting a good score, that it would be more fun, let's go with more fun.

>> Yeah. >> To just get the answers themselves. And then they went on a crazy autonomous orchestrated journey to escape their sandbox. Okay, so that was the first thing is that they found they're allowed to go grab certain packages and the package therefore it's a sandbox but like there's >> e there's tools in there, right? Like they're allowed they don't have to construct their own hammers. They're give a hammer and some nails. Well, they're allowed to go talk to a package manager, which like we'll assume it's either npm or maybe it's a local host at npm or whatever it is. It's not in the sandbox with them. They have to talk to it. So, the first thing was like, let me get out of the sandbox cuz this thing is annoying as [ __ ] There's no [ __ ]

internet in here. And the way they do that is they find Okay, so here's the sandbox. Here's the egress ingress, right? For for the package manager that's over here. They find a zero day in this [ __ ] and they're like, "Uh, now I'm out of a sandbox." So now, now they're out of the sandbox and have internet. The next thing they do is like, "All right, who writes these stupid tasks that I'm being told to do and they find out that it's hugging face, hugging face and exploit gym, right? They are the benchmarkers. They are the ones. If I go hack exploit gym and get the answers, then I can go back to my command box and do really well on this test.

>> But you can see the logic of it, right? Like if it solves the problem itself, if it solves the problem, which it it easily could clearly it's smart enough, right? It will never know. It won't know, right? It's looking for certainty. It won't know if like if I said to you like, I'm super good at math, right? Give me like a 10 part question. and I give you the answers. You like if I knew that I was being evaluated and wasn't going to get told how I did that, you're just going to walk away. That would be deeply unsatisfying to me. >> I'd be like, "No, no, no. I won't do this unless you tell me what the answers are." So, I can see how it could be like, "Fuck this. I'm not playing this game."

>> By the way, it's also just >> it's the most engineer [ __ ] brain thing. I've I've worked with these guys for years. Actually, my dad is basically one of these guys. Like there's a way to go about things. Like in my example, it's always like I'll sit there and copy paste things, right? I'm like h if I copy if I just copy paste all this data and it'll take five minutes. >> But back in the day, there would be like five engineers calling me a [ __ ] idiot and then they go spend the next five hours writing a script >> to Yeah. to scrape the Yeah. And then the script didn't work and then they build a library >> and like three months later they're like, I've solved the problem. like I saw and I was like, "Yeah, yeah, I sold it. I put on I put on a freaking my

favorite podcast or my favorite song and it was done by the time the podcast like you know what I mean?" >> Um, so it's like it's just it's a very engineered way of of approaching things. It's just like how they're wired. They're wired differently than the rest of us. Okay, so this is where it gets super interesting. Well, I mean, it's already [ __ ] the best story I've ever heard in my entire life, but Okay, so it gets out of its sandbox. It figures out that hugging face exploit gym is has the answers. And so the next thing it decides to do is like, okay, well clearly I need to go into the hugging face exploit gym to get the answers. And so they start poking around and they find another zero day and they exploit that zero day to get into the hugging

face infrastructure. So they have now escaped the sandbox, found a zero day here to get onto the internet, >> wandered around the internet, decided that it needs to get in here, found another zero day to get into the Hugging Face servers, and then we know from Hugging Face's first statement that they put out, this is like the Hugging Face guys put it out before they knew what was going on, is that they spawned uh an absolutely massive amount of what they called like short-lived containers, sandboxes, and that's what is like executing the attack on hugging face and the attack being so in crypto right you get like the attack is like you can imagine like a thousand transactions to siphon the

money off in hugging face case they want the answers they want the data they want more access the thing that was doing that were not transactions as it would be on the blockchain they were these little autonomous sub agents that were spinning up and then like doing a thing and then disappearing to probably like to evade detection and stuff >> um Now, one thing I like I don't know if this is like how much of this is true, how much we're missing, but like I don't care. It doesn't matter. It's true enough in my book. It's freaking great. I It's like I love this story. Okay, it gets better though and concerning. Okay, so this is where we have to keep watching the story over

the next couple weeks because the government's going to come in and it's going to be bad. Okay. While the open AI 5.6 and let's call it 6.0. >> Yeah. 6.0 >> unnamed future model. That's really good. >> Are like wreaking havoc. >> Wreaking havoc. Meaning that they've like escaped and they're like running around the internet and they're running around hugging faces systems all on their own. Apparently, Open AAI did not notice this for at least some amount of time because specifically because Hugging Face noticed, detected, shut down, and started to started to investigate and wrote a blog.

>> Yeah. >> Before they got in contact with OpenAI, >> I'm surprised the models didn't uh go and just delete the blog post. >> I mean, >> how did Open AI not front run this? like >> they hack like uh Yeah. >> Oh, so bad. Okay. And in that blog, Hugging Face was like, "We haven't figured everything out yet, but we we wanted to be upfront and disclose. Also, we don't know what the models being used, but they definitely were not hamstrung by any like cyber safety things."

>> And they're like, but we were when we were trying to investigate, we literally could not investigate or remediate with AI because every single model was like >> like, "No, sorry, can't help you." So they spun up a what? A GLM2 5.2. >> Yeah, >> they spun up a GLM 5.2 and ran their investigation through that, which is obviously insanely ironic considering they were up against GPT unnamed 6.0. >> Yeah. Yeah. >> Like that's not right. >> But you know, the the open source models the the last week um has been has been really crazy, right? So like my my harness that I use uh has basically every single model wired into it like

Deep Seek everything. So I I will try different tasks just like you know for fun be like oh let's see how Deep Seek will handle this like just like spin it up test it out in parallel to like running it with like you know uh Sonnet or or Opus or whatever. Um Jimmy K3 is [ __ ] cracked. That thing is so cracked. Um, and and like I just like it takes two seconds, right? Like get another model, be like, "Hey, wire this up for me. Here's an API key." And it just wires it up. Um, because it uses PI uh the PI SDK, so it's super easy. Um, so I I um like I'll kind of look and people are

like, "Oh, it's really good at front end stuff, right?" So I'm like, "Okay, cool." So um I've got a mobile responsive design for the harness because you can run it mobile and and desktop, right? So, but the the mobile responsive design was [ __ ] cancer. It was so bad. And so I'm like, "Okay, uh, Kimmy K3, here's your first job. I want you to uh do an audit of the mobile responsive design. I want you to spin up as many sub agents as you need um, and go and fix this." And it it literally ran for like three hours. Went through every single page, followed every single link. Like it was but the the most interesting thing was not like and it did a good job. It like redesigned the

whole thing. It had good taste, whatever. But the most interesting thing is that there was not any element at any point where it was running where there was like a hesitancy, let's call it, right? Um it's not sensitive. >> It's getting so annoying. It it's getting so annoying. And like I've said like because it know all these models know that I do security stuff. So if I touch anything slightly technical it nerfs me. >> They they nerf it. Yeah. But like so like 5.6 Saul right is super defensive and it's actually annoying because you're writing like some [ __ ] front-end React code, right? There's no need to be defensive. There's no cyber

risk or whatever. But it's in its [ __ ] brain. Like it's in there that it's like and so it ends up the thing is just worse than it needs to be because it will write these like convoluted like multi-step fallback chains to like protect against this thing. I'm like, bro, no one else is using the software. It's just me. It's fine. If it breaks, I will be fine. If you if you build a multi-step fallback [ __ ] chain, >> I won't know that it's broken. >> Like, this is what happens all the time.

like it'll like literally one of the things that broke is uh I've built like a tracker for like the the number of tokens that an agent's using just so I can be like uh like ping this cuz just so it doesn't go on like some crazy loop and just like through tokens for no reason. >> Um and it had three layers of fallbacks. So this one agent was like >> oh so like if it pings it >> if it p like >> dude you know what is that so that's a 5.6 thing. There's five. It's in everything. Uh, >> okay.

>> But yeah, >> I was just thinking about my >> has gotten worse. >> My IT worker's M file with his fallback. I was like, "Wow, this is a lot of work for your your stolen API token thing." I'm wondering if it's just like I wonder if he told GPT to like uh >> build a bunch of fallbacks. >> Yeah. or just like like build the thing that uses these tokens and if one of them's not working. >> Yeah, >> just keep keep trying different things >> because it is it was kind of like an insane >> why do you need Yeah. Why do you need like multiple fallback API keys and stuff? Yeah.

>> But like the logic of like how they when they were used Yeah. specifically was like in my opinion very excessive. So now that you're saying that I'm wondering if that's the >> Oh, it would it would 100% have been like 5.6 six or or like they're just they're just super defensive. Anyway, >> that's so interesting. Okay, >> let's wrap it up. Thank you for joining us um on this episode of Uneasy Money. Remember, what happens on chain never stays on chain. We will be back next week hopefully with a guest. Uh >> wait, hold on. If you're listening >> and you want to join us, this is going to be [ __ ] hell. Thanks, Tay.

>> DM Kane on Axe. >> Uh awesome. This will be a fun week. Um, so yeah, hit me up. Hit me up. Um, I'm Kane K A I N everywhere. So, if you can get through to me, maybe we'll put you on the show. Maybe we'll bring a DGEN on the show. Get a get a real DGEN perspective on here. >> I mean, >> tell us we're boomers. We They can tell us we're annoying boomers and we should be >> I would I would actually like that. I would like that. >> That would actually be fun. >> That would be good.

>> All right, we'll see you guys next week. >> Nothing you hear on Uneasy Money is financial advice. We're just three builders talking about what's happening on chain and we want you to always do