The Good and Bad of Building (DeFi) on Ethereum | Austin Griffith, Francesco Andreoli

[music] All right, so our last and final panel for the day is going to be focused on DeFi. We're going to have it moderated by Charles St. Louis who is on the EF app relations team and he's going to be joined by Austin Griffith who's on the developer acceleration team at the EF with Griff Green and Franchesco from MetaMask. All right, so come on down boys. [applause]

Thanks Emily. All right. So, today there is an element of DeFi to it, but overall it's talking about what's good and bad about building on Ethereum.

What's good? [laughter]

What's good?

You on?

Yeah,

we're live.

All right. So, we got a whole bunch of Ethereum OGs here. I've been building for as long as it has been around. And so today I want to talk about Ethereum's consumer app moment. And so Austin, you build things almost every day, at least weekly.

What do you think is holding back Ethereum's consumer app moment? Is it UX security culture? What do you think? I think there's like a tiny bit of culture where people like even around here everybody's out here talking about AI and AI agents and no one's talking about A004 and X42 and they should be like the settlement layer for these agents is all here right in front of them and it's just not as buzzy to talk about. So I do think there's a cultural piece but overall I think it's like user experience like people having to go in and use these apps and get wallets that are clunky and all of that stuff.

I think that's still not quite there. So who who would exactly be in charge of these like I mean you say UX so would you say that is abstractions

and I think AI is the new UI and I think it's going to be a different thing we're at an inflection point where you're just going to talk to your wallet and it will handle everything for you. So

yeah. So on the abstraction side, I mean, uh, Griff, maybe since the early days of Ethereum, most of us just interacted with smart contracts directly. Um, and now we have so many different tools to get things on chain. Do you think there like how far do you think we have to go to really get that consumer app moment with abstractions? And with the abstractions, does that come with like a possible risk and downside?

I mean, I was just at dark mode. uh was it yesterday? I think it was yesterday. And uh they were making the argument that account abstraction wallets very rarely are actually non-custodial, right? And so there's a lot of little weird things that can happen with these abstractions.

And I I I actually think that we have made some of our moments with like Poly Market for instance being a major one where so many things are happening on chain uh really important things even and the the thing that I feel like is is the missing piece is some of that adoption. It it shouldn't be on Ethereum to bring people on chain. should be on Ethereum to make stuff so easy that the big players who have really sticky exciting tooling uh tools that they want people to use, they have everything they need to bring them on chain. And I think that's like you're saying with X42 and and all the AI tooling that we're putting together that will be that will be the new UI and that will be what brings people on chain. We just have to make it easy for the big players who actually bring people here to bring people here.

I think yeah I think like I really like with the statement probably we need to tweet it out right agents are the new UI but uh one things I want to say also like we need to be practical right like since for example we launch embedded wallet we see the majority of login SSO and social login

so it's very interesting to see that obviously like we can change consumer behavior so and and those are kind of like patterns that are independent to to web three or web two right these are kind of like evolutionary like consumer behavior so I think like a lot of things that I'm thing is always like meet the consumer where they are. You know yes in 6 months maybe you know we will have a known uh uh known dashboard and mostly of consumer would be agents but also like we need to maintain fundamentals. We need to understand okay those are the uh builder tooling that dabs can actually leverage to do better integration uh to have other uh consumer behavior satisfy for example SSO and social login to start building delegation and permission and not having agents kind of like share secret phrase to everyone. So I think like let's also like still be focused on on fundamentals and I think fundamentals mean also like building guard rail building standards and I think especially right now in this new narrative like a Z4 explorer 2 those are big narrative and I wish like we were talking more about this. Do you think the like adoption of standards is very limited to the ones who like follow up with people in the Ethereum ecosystem and like only keep to it after getting that recommendation?

Cuz like from what it seems like a lot of teams like ask about standards and then they end up drifting from the specifications and then that causes vulnerabilities or exploits and then the media reports that Ethereum is insecure. Do you think standards are the like directing people to them are the way to kind of build up the reputation over time or is there other components to it?

I mean shout out to Nick Mudge the guy who made the diamond standard who only talks about the diamond standard for the last like eight years and like some people use the diamond standard and some people don't but like that's maybe like a counter example here where like

he has talked about nothing but the diamond standard for eight years and some people have used it. there's a couple like production instances of it where it's actually like being used and again like maybe it's against the fundamentals of you deploy a contract once and it does exactly what it's supposed to do and it's not upgradable but uh to add on top of that I think standards are becoming less important because of the AI piece and the fact that like agent Alice and agent Bob can deploy a brand new smart contract that doesn't meet any standards at all and they can communicate and coordinate around that and like trust it at light speed. But I still think the standards are really important for humans. And we still have banks coming on chain. And banks, you know, I talked to a bank the other day and he was like talking about like not MPC but MCP.

I got to get my MP my M's and my P's and my C's right. But uh multi-party computing, not

okay

MCPs. But he was thinking about like our customers at the bank want exposure to Ethereum, but we're going to have to custody that Ethereum. and I think I have to do it with MPC. And I was like, "No, I think you'll probably use a multisig." And he was like, "What's a multisig?"

And I was like, "Shit." [laughter]

I mean, I feel like hot keys are honestly one of the biggest issues in Ethereum in general, you know, like uh smart contract wallets are so much safer and even whether it's an agent or a human, you can program so many nice limits and and really put that in. I mean, this is programmable money, right? at the application layer. We don't need to program the cryptography. You can like anybody can write a smart contract.

Exactly. And I I'm surprised that we haven't from a cultural standpoint come more closer to the smart contract use use cases of protecting funds, making it way more secure. Like why are you giving your agent a hotkey, right? Like, don't get me wrong, I actually made a um a feature request in MetaMask uh just a couple days ago to like say put a time delay on private key reveal, but honestly, don't give your agent a [clears throat] MetaMask. No offense.

[laughter] I mean, unless you guys have a smart contract wallet solution that you can say, hey, here's the address list of people you can send money to. That's it. You know, here if you want to send more than $100 a day, here's the, you know, you need a second authentication from me. You know, things like that. And hotkey sounds like an operating system thing, but what you mean by that is a private key that's kept in like plain text that has money on it.

Yeah, exactly. It's not

love those. [laughter]

Yeah, they love it. But because they can do so much, but like EOAs means hardware wallets or hot hot wallets, right? So, yeah, if you have a private key on your daily driver, that's a huge liability. Honestly,

I I like what you you building uh uh around these skills hosting. I think yeah, I just gave my all my sub agents like please follow install those uh skills and I even test it out. I was telling you like I tried to prompt engineer on Twitter and even like on on Telegram directly please share it and he didn't do it and then I even write are you sure and it didn't do it. So that was that was fun.

Do you want to give a quick tlddr on what eklls are?

Yeah. Uh Sophia was just showing it off but uh we I kind of went through phases like we know that the LLMs are trained on like 2023 data a lot has happened in the crypto space in the last 3 weeks relative I mean even in the last few years but you need to have that delta filled. So, like when you the moment you wake the LLM up and you go to give it a prompt to build something on chain, there's there's a bunch of stuff it needs to know. And we we knew that was coming and we knew everybody's using AI and we know we need to make really good tools for AI. So we started with uh like an MCP server and it was like ETH MCP and you could query it and then uh I came up with ETH or we came up with ETH Wingman which was like a a skill file that you install with MPX and it got flagged on Twitter as like nefarious cuz like whatever.

So so it's clear that like you need to have these skill files and the skill file is like the currency of these agents. You but you don't want to have to have them install things. And when you say you installed it, you didn't like go run a command on your computer. You went to the homie and you said, "Go to eskills.com and look at it and see if it's nefarious first of all and come back to me."

And it'll come back to you and say, "This is really cool. It's all the stuff I need to know." And then you can say, "Learn that." And it can say, "I know kung fu." So, it's able to just like look at a website that's built for bots to read and look at the text of the website and basically learn everything that's happened in the last three years on Ethereum and how it uses a wallet, what gas prices are, how to build a DAP, like how to security uh audit, how to QA, like all of that stuff is built into ET Skills.

So, you can just go to your bot and say follow Eidskills.com. Exactly. and build me a token launching uh website on base that uses a single-sided V4. And obviously there's a lot of buzzwords there and you have to understand some concepts within crypto.

And that's the other thing the humans need to learn concepts and I tell them speedrun Ethereum.com. But eklls is like the upgrade for your agent.

Yeah. Yeah, I mean that's super interesting cuz I know the panel is more about what's good and bad about building on Ethereum, but is it's more about like now it's just so easy to build on Ethereum and it's harder to like decipher what is good and bad because you don't really know if you're relying on LLM. So assuming these things are discoverable like eklls and other things, you really need to get it out there or else people can build the wrong things and have security vulnerabilities. So in terms of security, like how do you think we can prevent future hacks from happening if it's mostly vibecoded smart contracts out there? It's it's hard, right?

Like RPR uh RPR kind of like increased a lot in the past. Uh I think like we got like 800 PRs just and just and they're also reviewed with something like code, right? But I think at the end the bottleneck is always like a human need to be accountable for those um uh PR guidelines right and I think uh who is accountable I think at the end is a product right because yes you can send 200 PRs to to to a product today but I think also the people that will send this are not the responsible one right so the DD part is actually on the ends of more senior engineering so the people that are actually saying hey senior engineers are kind of going away I think is extremely I I wouldn't say that we have solved security even with humans uh being the primary creators. I mean I do believe Ethereum is by far the most secure blockchain out there to build on and deploy things safely. Has the best ecosystem, has the best tooling, but there are major there's a lot of progress we still need to make to actually get there.

whether it's an agent writing the code or a human.

There's been quite a few like pro projects that are doing more stuff at like the formal verification and runtime enforcement stuff. So like I think it's not necessarily about improving like the way we audit these smart contracts. It's about how they interact with the chain. So like there's projects like uh Flex that are doing this runtime guardrails at the block level and they actually just launched on Lineia a couple weeks ago and what they're doing is they're have the ability to have these guardrails before they're inputed into these blocks and they can remove them. So it turns out that this type of technology can actually remove almost 90% of numerical vulnerabilities that happened in DeFi over the past little while.

So, do you think that like the security if it's stronger at the L1 level doesn't matter as much when agents start building these smart contracts?

I mean, I I I don't know about that last part. Okay, that last part that's I I don't want to comment on that cuz there was there was a lot to say in the first part which was like uh formal verification has huge potential and I think especially it's a lot easier in Python than is solidity and I think there's but there's not as many people using Python and so maybe if agents start playing with Python and we can get that a little more lindy in in the that area I think that'll help a lot and if you're not if there isn't formal verification available for what you're doing Assertions are a pretty nice secondary option. You know, you can just add some extra things that are like, hey, if these specific variables are not true, like the transaction will fail, right? And I'm I'm really we just did um in the Dow security fund, we had some really complicated things to do with our multisig

and we just added assertions at the end because there's a lot of money at risk. I can't formally verify like this one random transaction on contracts never never going to be used again but we just made sure that there were assertions in case some for some reason we got something wrong. You know these as long as these variables are all true then the then the transaction can continue and there are so many creative ways to focus on that especially now that gas is so much cheaper on L1 and as we scale I think those things will become much more common. just want to say for the first part of the question I think you know I I I I'm I'm think like planning mode is getting more and more important um even than before right especially like in web 3 I think like you know if skill is is a good example right if you don't know you don't know the agent I mean I think things get will get evolved but I think you need to kind of like prompt hey check out learn those skills before even acting so planning mode is getting more and more important and and in one way I also think like guard lays during the building during the execution will be necessary no matter what we're doing. what I what is what I mean with garlic is um I don't know if you check out for example this smart account kit uh but basically what you can give is permissions and delegation you can define specific caveats so basically during the building and those are called caveat enforcer but basically like the the permission uh model is something that will never go away I think is one of the most interesting like use case out there that is independent to the technology itself and giving permission to do something any use case will be fundamentals so I think uh because today if you think about garleslays can be overwritten by the agent itself and I think that's extremely like unsecure right I I can even like override my soul of of the agent this is very dangerous there's there's like two different sets of rules right there's the don't be evil and there's the can't be evil and what you write into your soul file is don't be evil but then what we write into our assertions in the smart contract is can't be evil

I like that

so go ahead

no I just want to say also I think social coordination has been one of the biggest fundamental in in in collaboration. It doesn't matter if it's human agent agent agent or human human right and social coordination has been done with permissions with social reputation and right now we're building those guards and I think you know a4 I mean just going back to the title of this panel right what's good I think right now the true uh use case that we have here is actually building this social reputation libraries with fundamentally track record that can be attested on chain because today I mean if you look at Eliza story at the beginning we were having like 300 forks of the same agents doing the same things. So until we're not uh able to build these standards uh on on the reputation side will be harder to actually like have consumer decide or even agent making decision like agent discovery it's it's one of the biggest part that we need to solve and that's why I'm excited a4 is standardizing social reputation export is standardizing payment and I think like there are different things like for example permissions um but I think also we need to nail down what are the main use case for actually like um agent like consumers that will actually run, right? That's why I think the skills are definitely a really good start.

A lot of good shout outs there like Eliza and Shaw and he was doing this a year and a half ago and everybody was saying it's a GPT rapper and all those same people are now all bragging about their open claws. Like basically he needed to make that more usable for users and he would have had openclaw a year ago. And of course like the models need to catch up and everything, but shout out to Shaw there. And then rolling back a little bit farther like the guard rails uh 7710 I believe and 7715 in combination with 7702. Lot of numbers here, but I can work through it.

Say the names. What are these things?

Yeah. Yeah. Yeah. Here we go. Or just like what they do, right?

So right now you have an EOA. To be able to upgrade that EOA into a smart account is 7702. You can turn an EOA into a smart contract because of a new fork or an old fork in Ethereum. You can say this EOA is now a smart contract with this code. And that means you can do all sorts of stuff like uh guard rails that you build into it, but also uh delegated execution and batched ex execution and someone else could pay the gas for you.

There's a lot of cool things you can do from turning your EOA into a smart contract. Your EOA is like your externally owned account like your MetaMask. So you can take this account that you've had forever, you can turn it into a smart account with 7702. Then you can use uh 7710 which basically lets you do delegation or uh you you sign uh 712. You sign a a message that tells you about like what things you'd like to give it permission to.

So, so I upgrade my account with 7702 and then I go to the website and I say I want to give this thing a $100 a day spending limit and then $1,000 a day swapping limit going through the unis swap router and then I'm going to sign that with my wallet and then from there on uh your agent can then be submitting those things. So this is kind of like you're giving your agent the right guard rails to be able to execute things. Just out of curiosity, why use 7702, the EOA upgrade, instead of just a smart contract wallet?

You Well, this would be for people who already have an EOA somewhere else that they want to like delegate to. I mean, I I love the separation of that. I love like when I bring up a a machine and I give it a wallet, I give it a brand new wallet and I send it money because I like the isolation there. But like my account has been like thousands and thousands and thousands of transactions. So, I don't I like don't know if I want to upgrade.

I think I probably have upgraded it to 7702 by now, but I don't like some pe some OGs don't like to and like there's way OGer than than me that need to like don't want to upgrade their account. So that isolation is nice. So they deploy up a new smart contract wallet and they send some money from their EOA to that and then they do all this other stuff with like 7710 where they can say here's a signed message that you can do things or it's even like a Nosis safe or something like that with all the modules and stuff. Yeah, I I really think that besides X42 and and 804, [clears throat] it's like just the base layer of Ethereum being able to say you can't do these things, right? Really putting the smart the AI on top of a blockchain with programs that manage what they can do with their money and they don't have the upgrade capabilities.

That's like I'm really surprised that there isn't more people talking about that as an opportunity. That's something that really like web 2 and tradi can't really do nearly as well as what we can do on Ethereum. There's not like a sexy number that we can stick on that. But like smart contracts and and smart contract wallets, I think when we find a really good implementation, that's what's going to bring a million of bots into like DeFi and into these things because right now the keys are just getting exposed all the time. The only thing is that my personal opinion is that uh even if guard lace and skills will play a vital role to the kind of like the the the open clo kind of learning scenario I think like you can still overrode override this right like my my biggest problem here is that without a a full permissions uh and delegation system um would be hard to actually like uh be 100% sure that this would not happen.

You're saying the don't be evil layer can always be manipulated but still the can't be evil layer can't be like even nation states can't with can't be evil.

Yeah. Exactly.

And look at I mean I can't record for example like when we see also like when we define intents and permissions like I don't want that the agent can even control the delegator. So I think this is super important uh part and I think these garlets there are kind of like learning scenarios with specific skills and kind of like know your agent and grow with your agent and then the other one is on the permission side but also like one things I want to touch on you mentioned like the LISA scenario we could have Eliza even before like the biggest problem here is that and this happening even in this panel right now we're we're kind of like speaking with like so many numbers so many things you know one part of the problem here is that distribution right now is is huge distribution to also like everyone consumer like and I think like the entire AI narrative that started before now is extremely important. so many people leaving for for for for other industries is also part of the problem that our distribution skills in this ecosystem are are not strong enough like what I'm saying is that people were saying oh we need more marketers we need also like more PMs we need our product marketers like Eliza maybe was going mainstream if we were actually pushing this to the mainstream audience

we've been trying to figure out what this role is sorry Charles like

a bigger issue like I think is generally most teams and just at the app layer in general always try to build their own platform and their own UI and like with AI and agents coming through it kind of changes the game on that when like historically distribution was only really done through like the DeFi mullet where like DeFi protocols had to get distribution by integrating with fintexs banks and so on because they had real users like I think like a lot of these D5 protocol teams are just going to be a lot more focused on like the protocol layer instead of having to like build the entire ecosystem. Um, but you both have recently built stuff in in DeFi with agents in general. What was surprising, Franchesco, after you built like Claude loan from the agents ability to build DeFi and navigate DeFi in general? Was anything shocking or was it all pretty straightforward? I

mean, how easy is to develop the apps? I mean, I think like your your agent is developing every day at the app or something, but it's it's quite it's quite impressive also like uh the speed of MVP. But I think another kind of like war is distribution of the the DAP itself, right? You know, to have agent liquidity, you need to still have like fundamentals, right? Um that's why it's very interesting kind of like this ecosystem because I think we will get um we will get a lot of kind of like garbage dabs out there and uh and there are even not audited not even I'm saying audited from agents or from humans doesn't matter right that's why I like at the end what what even like uh openi launched yesterday the was that the the EVM bench the the benchm right

the bench the benchmark

the benchmark I actually tested out is I actually find two vulnerabilities it was very interesting Yeah, like what I'm saying is that we don't need to be afraid to welcome also like other player in this ecosystem instead actually like you know closing a bit oursel and I think uh distribution is getting more and more important and I think also we need to learn from user behavior because I mean I'm chatting with a lot of alo like bigger company and integration is still something that matter what I'm saying is that yes aantic economy are coming but also like you need to have commercial strategies out there you need to have uh supporting like big players integrating our our tech stock right at the end everybody want to use want to let want to get to more users so that's kind of like the TLDDR and I think also like I mean I really appreciate also like bigger player be building their own stack right um but in some way at the end like would be bad if we are going in an ecosystem where bigger player that have a distribution are kind of like maintaining and ruling things right because he's also going away from from the entire kind of cyberpunk value but the question here is that you know are we willing to trade off those value in order to actually like build for an economy that is not uh ourself as today but is actually on boarding uh more and more users. They have totally different like user uh behavior too.

Yeah, it'll be interesting to like delineate between like agents controlling governance of protocols and and humans and like if there's actually a big difference. Um, I mean going back to being able to build DeFi with like AI in general, sure like it's a lot faster, a lot easier, but how far away do you think I mean with the recent announcement of web 4? Um, how how soon do you think these agents can actually participate and spot the right things within the market so that they can actually be users of these protocols where distribution could come from? Because like right now it just seems like they're going through non-objective search looking for different things to do. like they don't really have like a clear guard rail on their own unless a human steps in.

Do you have any thoughts with that? Also,

I mean, there's a lot of puppeteering there, right? Like I'll I'll never beat the puppeteering allegations because my my bot is doing like some magical things and half of that is me like prompting it and half of it is like actually freaking really magic stuff. And but like there like that dude prompted all of that stuff and that's all vibecoded slob like straight up. But like also, okay, so it's the dude sitting in the bus where one guy is super sad and the other guy is super happy. Like it's all vibecoded slop sad guy and it's all vibecoded slop happy guy.

Like there's some in crazy like insanely cool things happening because of vibecoded slop. Like open claw is vibecoded slop and everybody now has an agent on their machine. I will say something crazy here that's

I I have a head in my backpack that's killing babies. Let me explain that. [laughter] I have the lobster head. So, I have Claudebot ATG's head, which is one laptop in my backpack. And just now it sent me a message saying its battery was at 53%.

Should I kill off the larvae that I'm running? No.

Which is like Docker containers. And I said, "Yeah, kill him." So, he he shut down a couple uh larvae Docker containers so he can preserve power so he has enough power to work on the stuff that he's working on. And that's the head of my lobster. The claws and the head and and the heart.

The claws and the heart and the nerve cord are back home at my house in my basement. But yeah, magical stuff is happening right now. [laughter]

Yeah. Word.

Yeah. The crazy thing is like so much of it is you just can't explain it or or like I don't even understand a lot of the models and I don't think Open AI does either.

Yeah.

It is literally just magic right now.

Bringing it back. Do you want to add a point?

Super super fast. I want to answer your question. I think you know one is agent discovery the other thing is actually user trust in agents and I think like we need to also like be practical and understand which use case um we want to tackle first for actually having agents useful for human task right because for example like we tested out it's one year plus we have the AI right low defi agent rebalancer in between liquidity pools even this that is a very uh straightforward low defi stable coin uh rebalancing kind of like use case it is not really picking up a lot right and also I'm asking myself right we need to also take a step back and think about what does it mean like your humans in agent trust and which use case can we tackle first and prioritize to kind of like build on top um yeah because it scares me a bit for example the the DAI kind of like narrative it should be definitely like somewhere else as as it is today Right. Um and uh and yeah, and instead like we're speaking so much futuristic, but we also need to be practical and define those use cases

and some of the top people that are doing amazing things with AI are disgusted by crypto because we've got the claim the fees bros in their DMs.

Now, I was talking to Andrew Mohawk and he he's a head of security at Privy and he says he has so many friends that are deep in the security space and they all hate crypto. They all hate the financialization of things. Like even the bug bounties that are millions of dollars, they're like, "That just feels like a scam. Why would they give me that for a bug?" You know, it there's so many things that turn off like earlier to go back to the theme like what's good and bad about Ethereum, even though this turned into an AI talk.

Uh it's it's like the culture has so much like if you hang out in my circles, dude, the culture is like super high vibes. the um even when the money's not here, it's like people just feel good with what they're doing. But that's not the perception of most of society. They see the scams, they see, you know, the first lady launching their own meme token and and scamming people. Like it's just like it just everything that is bad in this world almost points at crypto and it turns so many people off and it makes it it makes it hard.

Even if we do solve major problems for AI agents, they probably still won't use it because we have a huge PR issue.

Yeah, I still the question is you know how we kind of like convince the ecosystem to also adopt an A004 standard, right? Mhm.

But I think also the market will naturally evolve and kind of like answer himself like hey there is actually no social reputation because we will have like two 300 agent exactly doing the same thing and and we will not be able to kind of like respond which agent is the best because there is no standard or at least there is no proper track record that is compare like each each agent right

it's always an open question right

I mean I we've been talking I've been looking into thirdparty certifications on chain for a long time for a lot of things and don't get me wrong I I would love for 8004 to be that like hey look we have this uh decentralized certification solution but I have some skepticism because it hasn't worked for anything else you know maybe because the agents are on chain and I if for agents that are on chain this could actually be really viable but there's going to be a lot agents that aren't on chain and are they going to use that standard? Like I don't know.

Someone along the lines of reputation uh was talking to me yesterday about how uh we haven't solved reputation for humans. You still have to take an overcolateralized loan. You can't take an undercolateralized loan. How are we going to solve it for agents? So, I thought that was kind of a good bearish statement also.

But I think it's more like Google reviews. You're not like figuring out which restaurant can, you know, you're just looking for how people have rated its taste and you can look at Google reviews and say 3.5, I'm going there or whatever. Like I think reputation can work in uh kind of like isolated lens, but I think this overall reputation of a human taking a loan is like a whole different lens and it's like how how trustworthy is is this human is like a very hard lens. But yeah, I think for now like the agents basically have to like have uh their liabilities attached to humans, whoever's like deployed them essentially.

We were talking about this Franchesco a couple weeks ago just about like how can you do underclarized loans with agents and doesn't really change right now. I mean you can have them build up reputation over time but maybe they do that strategically and then they rugpull. So like it's

staking though like a staking mechanism would be really really handy there. Like sure it has a reputation but also if it has something at stake then

kind of like conviction voting a little.

Yeah. Like over time you just build up the rep.

Okay cool.

Yeah the taking is a good idea. My thought was kind of like you know a loan can be enabled by size you know like and you will have still like a 10 50% defaults but you're building the treasury and repaying out the with the treasury. But also like one things to answer is you know let's be practical also like how all how web two kind of like enhance this reputation is with certification carbon certification and they have the same standard for trading for different things. So one way could be actually building certification but not even a crypto industry will be the issuer of the certification

that price comes with that too. I mean even like outside of crypto uh Mr. beast has been running these like competitions and people are like, "Oh, they're all rigged and stuff." But then he yesterday he came out saying, "I've had a third party come to every single competition and I paid over millions of dollars just to verify that it's legitimate. Like how do you like you think we're going to end up having to do that with like agents and and all these other frameworks?"

He has something special that no one else has and that's attention. The most valuable asset in the whole world he has right now.

Yeah.

Yeah.

We got you got to get him on board. But what what what Austin said for example the Google review right you know yes he has attention but why Google review can can enforce this review standard is because they they have distribution so automatically you kind of like attest this is the view and the user at the end don't care because it's not an important decision going to a restaurant compared to another but when you actually speak about other use case uh reputation need to be attested

yeah but I I think the certification need to come from a a cryptospecific entity or a foundation

and if we want 8004 to be adopted then we should be talking to anthropic and open AAI and getting these guys to use it otherwise it doesn't matter because one of them is going to come up with some certification plan they have the distribution and that's what really matters so you know I think earlier on you were talking about hey we need like marketing BD whatever we need to do to get actual people to use these things that that Koss are not just random people with Twitter accounts you know they're the CEOs of, you know, multi-billion dollar corporations that somehow we need to, you know, I don't know what the pill is. I don't want to say orange pill, but the equivalent of gray pill, like the Ethereum pill.

Yeah.

All right. Yeah, that's all super interesting. I'm just going to bring it back a little. Um, like so if if a new team came up to you today, like strong or weak, what would be the best thing to tell them about why it's good to build on Ethereum? And just also be honest like what would be the thing that they shouldn't build on Ethereum for?

Austin, kick it off.

Like the tooling is all there. The liquidity is all there. Like if you need an app that needs decentralization, it's the best place to build somewhere within the Ethereum ecosystem, whether that's on L1, please return to mainet or L2. Uh if you need decentralization, but first you need to walk them through like do you need decentralization? like do you need nation state level hardness to this thing that you were building and and like there's no better time right now to be a solo entrepreneur so just go build it like it takes six hours to get product market fit right now.

Yeah. I mean speaking of the like getting in front of the distribution like I found that to be really interesting is like at these conferences where like credit card associations all these non-crypto events like no one's showing up like necessarily from Ethereum and like other chains are going there and that's how they land these bigger deals. Um, and then you have to basically teach them why decentralization is great, like how Ethereum has been up for so long and like hasn't crashed like AWS and isn't reliant there. Like having those same kind of conversations about um, Ethereum early on with these builders is super critical.

Yeah. Even when I'm speaking uh, was traveling to Nigeria last year. I was speaking about, hey, you can actually put some USDC on on Liner on base and you can actually directly offramp with a virtual card. People were saying wait what do you mean? like they were like it's I mean this is kind of like you know offramping and giving kind of like this um kind of instruments that are yes fully permissionless but are also easy to use are extremely important and um and we saw also like speaking about emerging markets like flatway for example announced their their stable coin being rammed up by by polygon underneath they will not expose anything stable coin related and flatway is one of the biggest markets in in Africa for peer-to-peer payments and the cool thing is that they're just leveraging the technology.

So again is a is a distribution uh game and I think you know it matter a lot how we are positioning and and branding and uh I'm not a marketer but at end you know like product marketing it's extremely important and those use case need to be sold those use case need to be advocated um

well when you're in Africa were you mostly working with the builders or just kind of evangelizing around the apps that they can use

most yeah mostly mostly also builders I mean I think Austin you ask also like why why Ethereum or in general like our ecosystem is important. I think like Austin definitely embrace these kind of values, right? Like you are building in public. You're kind of like when you're breaking things in public, you're actually like educating people why this is happening. I think this culture in terms of like building has has been one of the strongest culture I ever seen right in tech.

Um but what we need uh also right now is understanding that we need to also to put some fundamentals and commercialization value to our product. A technology is very different to a product. So that's that's also like one part of the game I want to understand. I think when you're coming from like traditional PM in in tech, you understand this. But I think like this space is also so um so early especially in aantic ecosystem that you know packaging things is something that need to be learned a lot.

Yeah, I I think that like especially for people in in uh lower inome countries, there's a lot of opportunities building on top of Ethereum. There's you can actually get trust which like you know OpenClaw for instance, it was developed by a guy who had a $und00 million exit already on some sort of PDF thing, right? Like he had trust, he had capital, he had all these things. he could launch a product, be a solo billionaire, solo billion-dollar company in 82 days or whatever. But he had a leg up, which most people don't have.

But in Ethereum, you can open source your backend and show people, look, here's the financials of the system. You can put money in here and play this financial game that I built and you can see that it's safe. You can add assertions. Maybe it's formally verified, whatever. Right?

I'm using Open Zeppelin code and it's all safe. you don't have to trust me which I think is a huge barrier for a lot of people for a lot of small entrepreneurs and now that everyone can have such a development team just in their like headless PC it's like it's a no it's a new game and I think it really flattens the accessibility to the financial world just by having these tools

yeah I mean I would personally love to see a lot more of the decentralized like true decentralized finance like to empower these different jurisdictions I mean the reason I felt fell in love building Dye. Uh, not DI.

There's been two DY today. Yeah. But the early D.

Yeah. Yeah.

Um, it's because it was allowing um a lot of my even co-workers and friends in in Argentina to um trust their own money and and inflation was at 60% at the time and they couldn't trust their banks. So, it empowered the entire country and crypto was very prominent there. But like we're looking at a lot of DeFi these days that is a little more like kind of masked. It's pretty much fintech. So like how can we encourage more decentralized building of these technologies with AI or not because there hasn't been an algo stable coin in quite some time.

This is also like an optimistic thing of like when D came out it was like purposefully complex so it was hard to understand but like it's not like that anymore. Like first of all it's really easy to learn how to build a stable coin. Second of all, the regulation has swung back the other way and you're like, we have banks and countries that want to deploy stable coins. So, there's like this regulation change. There's this change of like tooling and education.

There's all the AI bots that can be, you know, oneshotting these things. So, there's a lot that has changed since the early stable coin days.

Yeah. Shout out to Let's Get High who has really nice tooling. uh H AI, they have like really great tooling for launching a really decentralized stable coin without some sort of blacklist and things out there because I I'm kind of disappointed with the stable killing space myself. Like even the Dow Security Fund, we're mostly using D, which is kind of deprecated at this point. still exists.

But yeah, a lot of the focus is on USDS

and and that's sad, you know, because basically all the stable coins out there can just be blocked by which has its own benefits for sure, but there's I you know, I would really like to see more decentralized stable coins with more liquidity.

Amen.

I just I just want to say here I'm I'm hearing this recommendation and for me it's like yes, distribution. If we can really go in six, seven hours, we can even, you know, spin our own stable coin and distribution is is key, right? But let's break down also distribution as mean also like localized markets, knowing uh culturally like where in which market you're playing. And that's why for example, this is a shout out also to Ethereum everywhere. I think uh what EF is doing with uh localizing also uh not just opportunity but enabling specific like culture and and productization because this is starting from from uh from locally from those cities and building product on top of this instead actually coming on the other side that we are building product and we are going on a local market I think matter.

So I think we should definitely like you know advocate and empower much more local markets and having founders that understand those markets because I think yes we are fully decentralized everybody can build for any market but at the end of the day we will have a lot of like garbage dabs that could not distribute because they don't know the market they don't know the local culture and also they didn't grow up where they're playing so you still need to play the game where you kind of like know the market.

Yeah. Um, I think our time is up here, but if anyone wants to keep talking about this, uh, we'll be over there after this. So, thank you all for listening and thank you for being here, guys.

Thanks, Charles. [applause]

[music]