ETHPanda Talk | 从 Nethermind 到以太坊基金会：Tomasz 的以太坊核心开发之路

Transcript

Hello everyone. I'm Bruce uh coin initiator of East Panda. So welcome to join our East Panda talk today. And panda talk is a podcast about building a better digital future on Ethereum. Uh we invite leading Ethereum builders to share why they joined, what they are building and uh what they have learned. uh also how they see the future by digging into their stories and ideas. Uh we aim to offer new perspectives and uh spark inspiration so more people get involved and help move Ethereum forward.

So today we are more than happy to have Thomas here to talk about like uh his personal growth foundation and the protocol AI on Ethereum. So welcome to Shanghai. Welcome to join us today. >> Thank you. Thank you for having me. Great to be here. >> Wow. Nice. So, uh, could you give us a brief introduction about your nail? >> Hi everyone. I'm Thomas. I joined the Turing Foundation this March as a co-executive director uh alongside Wang Shaw and before that for eight years I've been building company called Nethermind that I founded in 2017. So Nethermind is a core development and infrastructure uh company that has been

very very focused on Ethereum for all the time. I at the beginning was uh directly involved in the implementation engineering of the Ethereum client implementation the mind uh nowadays it's around 25% or 30% of the mainet network that is running on top of uh mind client software. So really something that's that we've been proud of long time before that. So I was in uh traditional finance. I was at City Bank on the FX technology side at the investment bank in London and after that also uh for slightly longer than a year uh at the hedge fund called Rockus Capital in London. So there's a story at the

Eternal Foundation last 6 months as um like trying to introduce changes without uh breaking things that work well and um lots of stories that uh I carry in Ethereum both from the Nethermind side and the Ethereum foundation. I think in mind side um historically we brought a lot of a lot of people to the ecosystem. So I think I've seen that journey many many times my own but also journey of um many interns that were at Netamine. So they would join for three months and we had around 600 people over three or four years that went through from through net of mind uh learning and then often starting their own companies uh are contributing to major projects in the ethereum ecosystem. Uh it was great to see all of those journeys and um

similarly I think a German foundation has the protocol fellowship program uh that invites a lot of uh new researchers to the term eical system and actually the the two are similar in size headcount wise like uh just over 200 people. >> Wow. Yeah, that's nice. So, so before like joining Ethereum Foundation, so as a founding Netherlands, what pulled you into the uh Ethereum core development? >> Um well, curiosity. I definitely wanted to get into Ethereum at the beginning. I wanted to build uh software for I think like for executing transactions on on chain by myself and like practically to

to build software for myself and then I started to read yellow paper just to understand how exactly Ethereum works and I started implementing yellow paper aspects like EVM. I just started playing with EVM and uh when you start it's hard to stop. know it suddenly ended up being a a three-year project for me to to implement everything. I well we had a working client after after a year or so but uh I started uh independently bootstrapped the company uh after a few months uh Jagos joined me uh for a while for a year or two to build alongside me and it ended up being core development but it was not meant to be at the beginning it was meant to be application building. I know there's something super

exciting about core development. If you're into this, you just feel the satisfaction and some kind of pride in building the uh the client implementation, the protocol itself. You start to feel like, okay, yeah, I can I can build it here. Am I part of this? And it's not that you're part of the group of people. You just really feel like you're part of the protocol in a way. >> Um I was building on my own mostly um without getting in touch with the other core developers out of ignorance. I simply didn't know how to really get in touch. I didn't know it was as simple as sending an email or tweet or something.

So, you know, it's core development story is not necessarily it's a it's a very clear path that someone tells you this is how you go. Sometimes you just discover it yourself uh in an open source and permissionless platform. I think that's very possible. >> That's interesting. Yeah. So, I I also hear like you both strapping and uh no taking no wasting money at the beginning for another month. So do you think like what were the advantages and the challenges of of all this? >> I I I tried you know like it's not that I I wanted to go without any funding. I I had some imagination of how you start companies. I've heard that well you potentially should fund raise. So I started uh looking for reaching out to

VCs but it was very not not well done like not well organized. Nowadays when I when I work with founders I know exactly what the process is right like if you work with founders uh that are asking you how do I raise money you tell them like like you you find the problem you find the solution you prepare the pitch deck this is how you reach out you have to have the momentum it's lots of aspects but it's pretty standard uh and then depending on the on the type of idea like whether you whether you're in the right place uh with the right right problem solution uh then it goes very and otherwise you maybe struggle a bit more. In my case it was very disorganized. I was building core development uh solution like the the uh

client implementation. Uh I had some other ideas of how to how to build monetization on the side of it. Uh I had a vision for a few years that was really grandio in a way that deliver client for hedge funds of the future that will run Ethereum clients in their in their like on premise for themselves. And you know in 2017 it was not something that many people were thinking of or there is something something funnier. Um I assumed that others would not see that only me who understands that and because I talked to a few people and they felt like oh how you will ever get there. Oh you're by your by yourself. It's really hard. Why do you build client? This will

never monetize. And you hear it and and you kind of feel okay. You really believe in it. >> Yeah. And nowadays when you look at what I wrote there in the like piece of paper it's it's kind of exactly describes what the mind is nowadays like it does work with the large financial institutions with hedge funds they do run clients on premise for themselves to to extract data RPC to send transactions and so on but at that time I thought like okay only I can see it and I I cannot convince others because I'm on my own because I don't have you know like credentials of the top university and I don't know how to do this fund raise and I don't want to prove people uh right by talking to them. I want to prove uh people wrong that's by by building something.

>> Yeah. >> I don't think many many new builders feel like this but this is very tough journey you know I struggled for a long time with this. So I luckily I had money to to bootstrap the business right so I I worked in a well- paid positions in banking and hedge fund. So I was not in a position of a founder that really feels like if I don't raise money there's nowhere to go. Uh but it was tough because the moment when you start hiring people then the money disappears super fast. Uh I think that the solo founder without any like going entirely by themselves I mean without hiring someone uh you can almost like stop eating you can move to a to a country where you barely spend um on yourself and then it's really cheap. But when you

hire a person then it's like it's totally different. they start asking for that monthly payment and u and that's when it becomes stressful and many founders you know they they immediately hire and now many founders who are for example not technical they they have to hire like the developers engineers at the beginning uh so you have to have a bit of everything to just decrease the cost if you want to do the bootstrap there are some advantages as I think I didn't know how to do the things at that time if I raise the money I would probably waste a lot of that money and maybe it would be a failure And it would be very painful failure, probably much more painful than just like proving yourself every day when you just struggle with yourself and not someone else asking you what happens to

my money. >> Yeah, >> I think that was a huge benefit of it that you're just responsible in front of yourself and you're very harsh for yourself obviously. >> It's it's tough enough, but it's better pressure than than someone showing you that they're disappointed. >> Wow, that's cool. So, how many people in Netherland at the beginning? >> Not nowadays, I don't remember exactly. it was one or one and a half year we applied for one of the first uh foundation grants program >> and we got $50,000 and that's when I hired the the first person properly because Jar joined kind of as a as a founding engineer co-founder uh he stayed for a year and uh and then left but we are in a super good relationship

nowadays we met like funny enough I never met him in person uh until like three or four years later. So the beginning was like working remotely and the first hire was probably after one one and a half years. >> Ah wow. Yeah. Yeah. I think that's a good strategy you know. So especially we have AI agent um coding agent now we can build something faster with a smaller like a >> I think it's possible you know with lots of experience in engineering at least know how to build what they want to build. Then very often this says safe journey because it allows you to at the same time you try to maybe learn the the business part of it.

>> But look I I had much of the theoretical background. So I had practically the master's degree in the computer science and 10 or 15 years of experience of building large systems for financial institutions and I also went through five years of the master's degree uh for the business administration >> uh and while at city bank I also learned much much more about corporate governance accounting and so on. So it really covered almost all aspects of running the business and marketing. I studied like business administration market marketing and uh and it was like you know this CFA it's like three years of learning a lot of accounting and governance and uh ethics and legal and so on uh and financial instruments. So

this was good enough preparation to >> to not be totally overwhelmed and still I felt overwhelmed because uh there was so many things that were practical like you you start on the day for day one >> and I I think my first failure was within the 10 or 15 minutes of starting the company. So I went to like register a company luckily this is in UK so in UK it's very easy. You start it in like five minutes. You register, you pay $10 pounds and and you have the company and for the next year you don't have to file any paperwork. Everything is fine. Uh that's beautiful. I actually recommended some of the founders like this is very counterintuitive, but go to like places like Hong Kong or UK and just start a company because it's the easiest and you

don't pay taxes for a while. And then I applied for the bank account and it's 2017 and they and they ask you like why do you want to open a bank account and I wrote everything that I could imagine that you could do and running a blockchain company like and uh and I got a refusal very quickly like within few minutes and it was a bit like we'll not open a bank account for you don't ask us why don't contact us again and it was like for the next three years I think we were running this uh landmind without the bank account we were doing everything from my like private bank accounts and >> so it was possible.

>> Yeah. So for Netherlands for the like many years of development so what was the like key milestones do you think like uh I I I think maybe receive EF grants is one do you have other like milestones? Uh I think there were a few. So definitely the EF grant was important because um I mean maybe like connecting with the London community was super important and role of Antonio Sabado of U nowadays is chief growth officer at Namind and he helped me a lot to integrate with the community in London encouraged me to run workshops on EVM to to share the knowledge to he told me to open uh to start a Twitter account that

I didn't have at the time for a year and uh and so and so on. So this was big and then he was also the one person who told me you should apply for a grant for a ter foundation said I don't mean no like I don't deserve it like I'm for sure they will not agree uh and uh and then later I think obviously like the when we first time deployed never mind and it just could sync it mainet it was super nice moments like you're you're proud of this you delivered a big piece of software that you feel it's like your child and uh it's a big milestone and then I think for the business part like being starting to feel successful as a business it was well the sad event which is COVID but when CO started many people

around the world went online and uh started playing games uh started experimenting with crypto >> uh started just like having a lot of time and had to execute also all the payments uh online because there was no like everyone had to had deliveries and so on. So the economy chain and also because the business was all remote. >> I think that our business was already prepared to how to run it in this uh in this environment when you couldn't meet face to face and all the other businesses were like for a while probably they were lost because they had to reinvent how they how they designed the office like design the office in the digital space. So it was critical because immediately much more money

started flowing into the blockchain solutions and uh uh in like the the onchain payments and so on. uh and it brought a lot of attention to to Nethermind and started bringing contracts uh even before that I think the first contract with uh with uh XDI network at that time was super important was the first paying client >> that wanted to say like we want to use Netamine and could you please like provide some extensions that this is you know milestone when you have your first paying customer. Yeah, I think and later there was like so many different things but uh obviously it was accelerating like around the around the covid times everything started accelerating there

was there was lots of applications on interum that were extremely successful there was lots of money flowing in for additional investments people had to build very fast and they were reaching out for any uh any developers with experience and by that time we had like three three years of experience building uh everything on the protocol. >> Yeah. So uh if you have the opportunity to jump back like 7 to 8 years ago and uh give yourself like younger version of yourself like one or two sentences >> there are a few classic answers to to this one like the first one is I would tell myself don't do it [laughter] >> because there like this three first three years now you know I may smile but they are they're extremely draining and like u mentally exhausting and and you

seriously doubt yourself and uh and it's tough and I I think that maybe I could have avoided that experience. So even if now obviously there's lots of super cool things happening by building on Ethereum being part of Ethereum ecosystem uh I still like when you're responsible for yourself 8 years ago then you think about that person and you feel you know like maybe stay naive maybe stay unaware of uh of everything and just continue standard life. This seems very very fair answer like the caring about another person that the younger yourself. Um the other answer is that probably it's not reasonable to give any other advice than just just stay quiet. You know giving

advice to people. It's it's really hard to to predict different paths. You know your path and then you think maybe I did something different would be better but maybe it would be much worse. >> Yeah that makes sense. Yeah. So um you are now joining uh Ethereum foundation as a query executive uh working with the shall we one. So I'm wondering like how do you and the shall we want split responsibilities in E foundation? >> Sure. So we we agreed at the beginning that we'll be picking up uh whatever is there to be done uh whoever picks it up first. I think it gives like it solves two problems. One is then instead of working as a collective like two people that have to take decisions together,

you just need one decision from any person and it should move faster than instead of slower uh than the one executive director. And the second thing is it automatically assigns kind of task that you prefer to work on because everyone picks what they feel comfortable more comfortable picking up and then after a while we just I kind of feel like we we have clear split where I'm doing much more of the external communications podcast Twitter but then um shower is doing caster and then she does a lot of internal um internal like coms and talking to people and operations and we worked together on the like new compensation

policies. We um she picked up a lot of treasury and defi aspects. Uh and then we exchanged the views every now and then but uh very often we operate very independently. >> Uh sometimes we focus on different regions depending on who is where and uh and then we can have speaking opportunities. probably a few other things but then I I forget and in the future maybe maybe you'll switch maybe I'll get a bit tired of the of the communications and need to focus on the internal things maybe shall I will start talking more to to everyone or maybe even never change it and >> wow >> yeah but uh it's a good collaborations >> oh yeah yeah I can see there like a collaborate very well uh in this year uh EF released their three main goals like

uh scale L1 scale blobs uh improve UX. So after like a half a year, many months is a go later. So what do you think which goal is the most challenging to work on now and why? >> Uh which go is most challenging? Well, you would have to ask people that specifically work on those goals. I think that um interrupt UX is the broadest and and because of that it feels like the most challenging to to answer questions when exactly you achieved all the goals or not because it was the one that was hardest to quantify on the scale one scale of two. We could have said we we said something like L1 will will be 100 million gas per block

and then it will be 300 million gas per block. So, so we have specific goals and and people started going there. It helps a lot. On deal two, it was slightly less clear. I mean, we knew that we'll uh go to the blob scaling and they'll be accelerating the Fusaka release with 120 blobs and then further and I think that that question about further of how how we go to the two-dimensional uh pure Das. This is less clear, but I I I think that it's worth to uh to talk to Alex Stokes and and all the like researchers and uh and engineers at the foundation and core developers on what the next steps would be. Uh I've talked to Vitalic recently about this plan for

blob scaling and I think he's very he has very clear uh road map for all of this. M >> uh so so maybe that's why I say that's not the most challenging but the data availability concept in general is uh is a big thing like as like how to do it right and and how to make it competitive but also properly decentralized properly secure >> interrupt UX so we have now the two major deliveries proposals the OAF the open intense framework >> and I think recently there was a very nice big summary from Joshua Dolph about everything that we've done there and the ecosystem like I think that the last part is most of the coordination of the non-client teams, noncore development teams, but more like tooling teams, the

some of the researcher work, much of project program management and the the UX interupite beyond this. So there is EIL and this is coming from the um the AA team that's your advice and uh Marissa Pner and their team >> and this will be properly announced during def connect >> and then the question will be is it because it's it wasn't quantified that this means that we delivered an improved interrupt and UX. Some people will tell you oh interrupt is actually working fine. Some other people say interrupt is the major issue. Um some people keep saying oh UX requires uh fixing. But

then maybe different people have different views on what UX is because UX you can interpret it very broadly. You can say oh it's it's security >> or it's privacy or or it's performance or it's latency or it's UX for retail users or for institutional users or it's it's wallets that you have. um and so on and so on. Yeah. So that's that's why it's maybe the most most challenging the hardest to define of what success means there and and define it in a way that everyone else will agree with you. So we both know that layer 2 are also critical part of Ethereum but interoperability is still hard. So what are the biggest blockers uh to the four like layer 2 interoperability today and how are you

working with layer 2 community to you know address them? There are multiple questions here. So when when you have layer 2s that are that are EVM layer 2s then then you have much of the like compatibility interoperability between protocols that is a bit easier. Um then there are questions about the security of the bridging and how permissionless the bridges are within the interoperability how much of the properties of the L1 chain you're preserving as like the permissionlessness uh censorship resistance and so on. So can you bridge the assets without relying on the third party and uh this is a big challenge because so far when you when you look at

OAF many of the solutions under OF does do not really satisfy all of those requirements. The existing bridges very often do not satisfy those requirements and then you have those trade-offs where where you have better UX but then you suddenly hiding all of the security issues. Then the within the EIL you'll be trying to answer questions about how to do all of this without the third party and permissionlessly and what are the big features. So big challenges in the interperability the the address representation things like uh availability of exactly the same tooling and uh and applications on on different chains. So you know things like multisig uh the uh the representation of the stable

coins like you maybe you know the experience of like wallet displaying practically the same asset bridge into different chains and then you have all the different assets. So within the Kohaku wallet uh like Kohaka wallet SDK we've delivered many suggestions. So like the um how to how to represent addresses consistently how to represent assets in a consistent way. >> So it does feel like one chain because that's what we maybe that's the goal that we said like it it should feel like one chain. That's a big challenge.

There's so many small things here. >> Yeah. Yeah. From my experience, I I thought like there to interoperability is not more is more than a technical challenge. I I I also see like a uh maybe they have some interest. I mentioned that multiple times as well that interperability starts with people and the way that you start becoming interoperable if people talk to each other. Yeah. if they want to integrate the system. So if they don't want to create the modes and say isolation of the chains uh and sometimes uh for business reasons the L2 operators might be feeling less willing to to collaborate. So this is also something

that we set as a goal for ourselves to help the entire L2L1 ecosystem to work closer with each other. uh like one of the latest proposal is to to bring engineers and researchers from all the L2s to one location next year. uh this comes as a as a feedback from some of the engineers that sad that last year they had this meeting in Bangkok and it was L2 engineers and for the first time at the beginning I couldn't understand the feedback because they said oh you know like the core developers they never meet with each other and I said like what do you mean like we always meet each other like every year and we have lots of meetings and then after talking I realized that that he was talking about L2s the L2 core developers never meet with each other

>> and I realized oh yeah indeed it's like all the L1 clients >> they need regularly but the L2 is not. So we will try to facilitate that. And funny enough like people say L2 engineers they always want to work to each other. They always want interpability and and they just like totally ignore the business people that try to create those [laughter] modes. And when you start talking to like strategies and business people they they understand the challenges of like setting up your own branding and position and the competitive landscape. Um and and that is when it's harder, right? So they they will feel maybe we shouldn't. But when you talk to engineers directly, they're just like always work together because they just want to solve problems.

>> Cool. Yeah. So looking ahead, so what are like EF upcoming uh priorities like and strategy for increasing uh community engagement and external collaboration? >> Yeah. So obviously EF has so many different things. So there is uh security protocol like upgrades, testing, coordination. Um so big part of the of the goals uh to look at the protocol and try to guess from all the conversations with the ecosystem people whether it's cipher banks or large institutions like you want to hear feedback what's needed and then communicate that to all core devs and help to coordinate in that direction by

just focusing research and so on. Uh all of this without ever sounding like you like you really imposed the direction too much. It's interesting balance like you want to be a coordinator. You want to um you want to analyze and understand where you should go. You want to tell people where you think you should go but at the same time you shouldn't tell people that's where we go. And I think all cordavs is like very opinionated group of people very experienced and meritocratic group of people. Uh so they'll feel if you're wrong. So you cannot be wrong. It's uh this is part of the foundation thinking about its goals.

Right. So, so help to find identify that thing. So when you talk about it, everyone thinks, oh yeah, indeed, >> oh, whoever I talk as well, they agree with that. >> So you don't have this u disagreement between what you announce that you think that the entire ecosystem wants and what other people hear because then it means no like who did you talk to? Um so the goals when we set them for like scale one, scale two and interrupt UX there was big enough agreement on all of this that that we were able to go in that direction all of us together >> and now we started talking about the next goals for next year and I started suggesting that maybe we should say finality privacy security but once again people said you know security privacy is

a is a default thing we'll go there anyway >> and then then I've heard from Dankrat who said well you know I think that we should continue the scale of one thing because we haven't finished that and this will be one more year journey uh which we in the past said that it will be two-year journey like to go to 300 million gas per block um and then we like in the process of discussing it like how how well we understood all the feedback from the ecosystem so we can uh finally said that that is for the all core devs that's what we think we understood and what what researchers think what um what all the users think all the DeFi builders think and so on.

>> Yeah. >> So these are the the goals for the next year is is is now the exploration but I I think finality is is big part of it. So so many institutions and uh and L2s and users said we want maybe like shorter block times and faster finality and there's some really really nice proposals here and then maybe we'll continue with scaling narrative and it will be simpler >> uh or maybe we just define a bit of this security and privacy programs but maybe we don't have to because those programs happen anyway. We have the privacy cluster outside of the protocol team. there's a privacy cluster of 50 people and they have very clear road map for the privacy solutions and the privacy for institutions something that we continuously get as a feedback that uh this is a must and by privacy for

institution I mean usually standards and explorations of how how to use existing solutions on Ethereum to deliver some particular use cases >> you ask me about how to engage more how to build comm build up community engage community so on ecosystem development we've we've had lots of changes to this structure and and um and the way we operate. So uh James Smith built fantastic team uh well or rebuild the fantastic team for founder success for the developer education and the acceleration of the enterprises adoption uh of the ethereum and reaching out to universities and uh Ethereum everywhere team. So they so we

can work with the physical spaces like the habs, the meetups and events. We've changed the approach to grants. So we uh almost like entirely removed financial grants like just paying to people and not asking anything and saying yeah like you can build it. Instead we prefer to help by just coordinating introducing uh amplifying the message. We activated a lot in our like social media accounts, the visibility, PR. We talk more in the event. So we can say, well, things are happening here and we can say like things are happening in Shanghai.

>> Yeah. >> Things are happening in Shenzhen, things are happening in Hong Kong. And we starting a hub in Hong Kong for people to meet together regularly uh between founders and academia and so on and we start collaborating with Polyu in in Hong Kong. Um and we are here in all these places and this is like type of message that want to just continuously repeat and keep asking like what else can we help with uh because it has to be your journey and it has to be you have to be independent when building communities uh because probably it will make you more entrepreneurial and that's that's what you want to also teach other people to be like core developers are kind of entrepreneurial because they are very independent figures and and deliver but also the application builders are

obviously entrepreneurial because they launch startups and applications. So I think that this approach to community where you force people into this mode when you they have to think about everything but naturally engage a bit like their customer uh might be better. This will be the effort like to keep opening this physical habs but in collaboration with with people that >> um that have also other arrangements that allow them to justify bringing sponsors like we we can help to bring sponsors from the ecosystem as well. So maybe the the sponsorship is combined between L2s, between application. We can tell them look this event is specifically good for for your goals. So maybe here you are a customer or this community does something that's

specifically solving a problems. So work with them longer term. Um sometimes we'll be directly customer but it'll be rare less often than in the past. Yeah, I can see there are many like uh local community hub growing and showing up. Yeah, that's that's great. So, we talked about your personal growth and Netherland uh EF and protocol. So, let's move on our like AI or Ethereum topic which is a very frontier like topic. I saw a lot of people say like AI doesn't need web 3 but web 3 is just a riding AI h. So, what do you think about this? For many AI businesses, web 3 is that security

aspect, the slowing down aspect and defensive or the centralization aspect where where they are now burdened of being extremely competitive with each other and having to deliver the the results revenue and faster models and better models. Um so do they need web free as soon as they discover mechanism that they can um they can use for for convincing more users to use their products um which always comes with like how many users are asking for for the privacy and verification um deliveries. So it will be coming. I think it's a bit like the in any other

in any other industry that we that web free touched in the past. It feels like five or 10% of that industry really requires some of the trustless delivery >> where it can rely on the existing trust. That's fine. But then as the assets are going more and more onchain digital becoming digital that's when web free becomes the default and I think on AI maybe that process will be happening faster because agents are digital and they will use digital assets from day one and that integration of digital payments web free payments with agents probably would drive majority of like

that's where AI will need web free it has been our for a long time like that people just use blockchain without talking about blockchain or talking web free those those aspects of verification of agents verification of training the decentralized training those might be slower it might be always the like some kind of edge cases um I think that you know we're talking very often about verification like what is reality what is truth um but what sells is not always truth right when we when we buy entertainment when we buy stories we actually buying like the lies and lies [clears throat] in a >> in a positive sense. We we asking for a nice story to tell not because we want to learn what the what the facts were or

the events were but we want to be entertained and I think big part of media is like this as well like even the news news delivery for many news media they know that they deliver partially entertainment or mostly entertainment there's some facts that serve only as a canvas for for building the story >> and for attracting your attention and building emotions do people then when when they reach for AI to generate things to tell them the stories about the news from the world to learn about the the world. Well, the people ask for for verification or for truth or maybe they don't care that much. But then when you have institutional like enterprise aspect when you buy the data when you

want to learn about the the truth then you'll need web free more and more and more because I repeat it the we had fake news now we have fake realities really you can have the entire set of news the news feed on on Twitter whatever platform and as you scroll they can generate the entire world that doesn't exist for you with the news with the fictional people with uh fictional videos uh everything can be fake >> and if you never leave your uh your room then you can be convinced that some world is is real which doesn't exist. So you have to walk outside and maybe even like some some reality out outside in in

nearby you is becomes fake if you if you start using too much technology you know like so that's when will be always a distinction of those who want to pay for data the true data and verify everything and that's when free always will be very valuable. >> Yeah. So in the institutional aspect for the retail users I think usually you get that and this is like very very vague thing to capture that the retail user very often um gets lower price for the data that they consume if the providers of data can decrease like the cost of insurance of what they provide

them like kind of a cost of operations. So as long as web free decreases the operational cost or the cost of insurance like they decrease the risk of the data provider then the consumer actually chooses your solution over other solutions and sometimes users will ask about privacy and the centralization but well decentralization maybe not so often you'll hear it from retail but for privacy you do. I talk about it a lot recently because you hear people saying, "Oh, nobody asks, nobody cares about privacy." And then I was in um in Milan uh this this year and I was on a taxi and I saw a huge billboard in Milan uh that was from Apple that said like uh

Safari privacy and I thought you know there's this large international company that did a uh lots of like marketing analysis of what people want to pay for and they deliver products that have to like convince people to buy those products and they decided the privacy will be >> the one word that will place on the largest billboard in a in a financial district of Milan. So users must be asking about it or they might be wanting it but maybe not talking about it which means they want to pay and that's why that's when they may ask for >> for web free solutions and with a AI you know like you have a we can imagine crazy stories with AI actually knowing everything about you and the privacy

gets totally compromised. uh you mentioned like a uh coordination layer like web Ethereum as a coordination layer of AI agents. So do you have like some new standards like related to this part? >> Yeah. So we have uh ERC 80004 >> or we have X42. Um the ERC80004 that comes from Ethereum like this is ERC is Ethereum standard. X42 is a is a more general term and they're all built on top of some other standards related to like um MCP and A2A and uh ACP protocol. So like you have agentic commerce protocol, agent to agent interactions, X42 is request for

payment. This is for agents to actually or not even agents but like through the through the HTTP protocol to to request for payment and verifying it on blockchain. And uh ERC804 is for agents to be able to use the trustlessness from the from the blockchain. So the blockchain trust ambient trust uh on top of the protocols like A2A that assume that the agents trust each other that they can verify identity through the traditional methods. So the trustless um definition of standards ERC 804 defines of how agents can announce their identity, how they can register, how they can announce the services they provide, how they can verify claims about each other, how they can verify

reputation of other agents uh and how they can validate the work executed by other agents by just um asking external on offchain validators to to rerun some execution uh and confirm the results. So uh all of those come together to facilitate uh agents either representing us. So we can delegate identity to agents and so just just work as a tool for me the same way as I use computer and computer kind of represents me and uses my identity u but also agents that are autonomous as like or representing some some accounts that are deployed on chain.

>> Yeah. Yeah. I think like half a year ago I didn't buy the uh you know the narrative like AI can work with Ethereum but until I saw like ERC4 I can see there's it's a good start so we can register our AI agent on chain to like provide reputation and validation lots of top people that work on agents uh participated and helped with that standard >> and uh wouldn't be surprised if it evolves evolves further and if new standards emerge But it's already a a good start also for people to start building toolings around it.

>> Yeah. >> So you see uh you see people launching dashboards uh running first implementations of ERC804 uh starting deploying agents that follows the follow this rules. Those are some most exciting moments. Lots of lots of startup opportunities. Lots of uh lots of opportunities to build. >> Yeah. >> I think we needed that on Ethereum. you know uh in in the past we would have the whatever was driving people crazy on interum in the past the the tokens the NFTs and so on and now now it seems that agents is the way to go >> I saw there's a chaos chain and aic governance like looks like very interesting early this year so where are they now and what's the hypothesis are

you testing next >> yeah so this this is coming back to to my explorations at Nethermind so uh together with the team we started talking about um about agentic governance. What would happen if you think that slowly the car developers get uh partially replaced by agents, not like fully replaced by agents or like by AI. Uh and it was just following the observation that in all of our daily work, we start asking LLMs like AI chats to to give us some hints to like to have a discussion with what what they think what they extracted from the from the data sets and and it defines our decision process and then we understand that the the

training of the model may actually skew our decisions in some direction. So there are some risks of like how do you how do you influence the the journal opinion um based on the model that you provide the one that is leading and and then when you observe core developers that are asking those questions like what should I do in this case is it right approach is it wrong approach still the core developers will have their own judgment and will just use it as a tool but it means that maybe 5% of their thinking got skewed in some direction maybe over time maybe it's like 10% the better the better the uh the LLMs will be the the higher that percentage should be.

>> Yeah. >> The more of the cordives will think like oh I don't have experience in that area yet so maybe here I rely on what I read and I verify it briefly but but they trust it and and maybe it will perfectly answered you know like you hear now about some some of the PhD level research in some areas sometimes like the the solutions that come from the AI. So, so then you think how it goes even further. Uh that at some point the the governance like the the social consensus layer the deciding where the protocol goes becomes uh how many proposal you can generate with AI and how much attention you can give to those proposals and then the proposals and the attention to proposals is like the what what what core defs do nowadays. They

they work for months on research and proposals. they propose something and then the other other core devs are verifying whether it makes sense and they start debating >> and now if you think that AI start generating those proposals then there might be so many of them that you need AI to verify those proposals and it becomes almost like a proofof work level of searching for the ideas and verifying the ideas and when you find a good idea then you actually can push the protocol in that direction and you can keep filtering your own ideas that you find >> uh based on which direction you want to go to and never never publicize the ideas that don't push you in the direction that you want to go to and publicize the other ones. So suddenly you have the proof of work like governance of the social consensus and

then if it if you if you really extrapolate it and you think that it gets faster and faster and like how how often you can generate ideas and how how high quality the ideas are then it means that you can also introduce changes much faster and almost you go to extreme where the changes happen from block to block to block to block and the and the social consensus almost becomes like part of the validation and voting. from this imagination of how chaotic the system is when we started talking about chaos chain uh and and we started talking about it like the beginning beginning of this year or maybe end of last year we made some presentations and um in the Bay Area on the topic and uh and within Nethermind we started a team to to explore it and the team is

spinning out from the mind to to build that but also like the the idea split into two directions two branches I think chaos chain nowadays the team is building a bit of exploration of the agentic um this like ERC 80004. >> So you can think of it as like a getting attention thinking about the general agentic governance and and pivoting a bit into things that at the moment make perfect sense and and can be a base for building further uh and I think they go to towards the governance but they also talk a lot about the uh what stems from the research from the hatu team uh and the vital labs where they where they look at the attribution of work of agents. So when when you when the agents collaborate how they decide between

themselves what's the what are the reward functions that are uh based on attribution of the result which is related a bit to this like proof of work. So I think this is the direction of chaos chain team >> while never mind continues with a strictly like a DAO governance model which is more related to the agent supporting the DAO decisions because we think that the the DAO participation is like is very similar to core development because if you're in the DAO of like the the council that decides and the allocation of resources and verification of technical concepts uh that are proposed for for example L2s Um this is a bit of what core developers

do. They verify the quality of proposals and they decide whether to contribute to this proposal and implement it which is like similar to allocating capital. So when you when you start with with the DA participation then there are a few more steps like the milestones when you have the the agent that start uh writing proposals on uh whatever if research forum the agents that start uh writing EIPs or maybe first starts commenting on the existing EIPs and finding issues but then writing EIPs all the way to the agent that actually is capable of joining uh old core dev's call without a video >> and maybe write in the chat about the proposals. further to the agent that actually joins with the video and starts

talking and commenting on their proposals or even proposing and defending and championing their own solutions. And the thing is it's it feels very natural progression. There's no no magic there if you understand how AI works nowadays like you you could build an agent like this or just it needs to go there like technologically and requires a lot of expertise in core development. So so that's why we thought that at the mind it was very very good place to start building it. Yeah. Wow. That's very interesting and I I can see this is the best period like to build some experimental initiatives like this one and many things. So the next the final question will be like for builders. So how will like Ethereum foundation um to support and coordinate

this direction globally? What any upcoming plans you can share? >> Um so we started the AI team the centralized AI team. Uh so David de Krappies uh buil runs the team and that was probably around two months ago and already they have this fantastic effects of publishing 804 and getting lots of attention to it and getting some announcements with uh with Google Cloudflare Coinbase um is being mentioned as a foundation coordinating those agendic payment solutions.

uh there's lots of attention to to Ethereum and Ethereum standards because of that like how you can run agents and and run dashboards and registrations. I'm absolutely sure that over the last two months like there's already a few stale startups that are building something related to ERC 80004 and so this is definitely what we wanted to to show and and this was one of the big asks from many people about the foundation to u to deliver the canvas for building the next next cool thing that will be applications driving. At the same time, I think that many many VCs are still skeptical about agents because the last few months were we're a bit depressed on on agentic AI and

blockchain but I think it'll be coming back. So >> yeah. Yeah, >> I think that DI team will have some more surprises and ideas coming in. M >> uh the team will grow not too much like we we want to continue being just coordinators not to not to deliver uh applications but to to build the uh standards for the applications to be built on top of. >> Yeah. Yeah. We also have a education uh initiative received a special support from DI team. So they are really nice.

So if you are working on this direction feel free to reach out to them. So they will respond quickly and provide some some support you need. Right. >> Well, I remember there also like they run some hackathons together with the major AI projects. So, aentic AI projects and I think we'll want to be uh very active in connecting here like the Bay Area region in in China well in Shanghai, Shanghai, Hanjo and the Bay Area with Hong Kong to to connect the like open source AI builders and the modded AI builders with the builders and startup founders uh in other parts of the world. So I feel there might be a lot of great solutions that will come

from from coordination between the different regions. >> That's lead me to the next question. So because you you already know like in China we have many open source AI models, we have many AI developers, engineers. Um so what contributions would you like um to see from like Chinese speaking ECM community for now? Uh yeah, I think that uh what's what we hear here is that uh it'll be great if there is lots of collaborations between Hong Kong and Milan on building startups together. There is some contributions like looking how to connect uh with with whatever really uh suits the like the policy and goals here. So I think the

crossber payments are coming into consideration and then this gets elevated to maybe aentic payments um and then and then you look at open source models here and how they can be connected with with any solutions that are globally available. Probably a lot of lot of founders in in the west uh are looking at the embodied AI solutions in collaboration with the Chinese engineers. So this most like will be the way to go like saying look we we understand what happens here in robotics in uh in open source AI uh and let's connect and let's build together.

>> Yeah. >> Uh or ideally like the the full solutions here obviously like the some aspect of it can be coordination across the region. So I think this supports the global global neutral aspect of Ethereum but the there's full capacity here to to build very successful projects. Yeah. Yeah. Cool. I think that there are a lot of things to do. So feel free to join the community to discuss further, explore more initiatives and work on some actual things, right? All right. So we are getting close to the end of our call. So let's uh wrap up with some quick questions. So uh so first one uh do you read a book? I rarely nowadays manage to like finish the book from the

beginning till the end because of lack of time. I started doing a bit like uh buying a book and then reviewing it kind of like checking what is about then going to lambs and telling >> tell me what this book was about why it was so critical why it was so important what were the major insights >> and maybe reviewing and matching some style of the book and some pages. So I spent some time in the last months on like it was more like few months ago they on the like Santa Fe Institute and the complex systems and the I think I was reading the origins of order uh by Kaufman and that was because I was trying to understand better the uh the natural like emergence of hierarchical

structure between L1's L2s uh like regional chains you know nowadays I think much of My reading as the exploration of what LLMs can answer in the question on particular topics and verifying it with with the sources >> uh but reading the inter books is is more rare. I got a very nice gift recently of uh the book called Babette's Feast from Bastian at the foundation. So so he knows that I like the movie and and he gave me the book. So so I carry it with me and I hope that I'll read it very soon. So you you're working a lot on a lot of things and but you you know stay energized.

>> So I wonder like what's your best practice for sleeping like training diet and scheduling? >> That's uh definitely not a question for me. I totally don't feel like a uh like a role model here because the there's no schedule. That's just uh rushing with different time zones, not sleeping properly and like >> okay >> uh not having any rules to to how I arrange time and just just rushing and and doing first thing that's possible. I had some some minor rules that helped me over time over years, you know, like I had no meetings Fridays and then sometimes I end up with meetings. So it helps me like traveling and having the full day of focus on even if some people sneak in some meetings like one or two

on Fridays. Usually most of my Fridays are free which means but it's like makes Monday to Thursday like extremely packed. Usually the other things are this is like very simple hint for many people is that like to keep the the meetings in in Google calendar or whatever you use to to end like five or 10 minutes earlier >> then then you have so much less stress between the meetings when you have 5 minutes between the meetings. M um so this is the scheduling hint and I this this one is I advise everyone to do it because um I was reading specific research on this like this shows how much the stress level falls if you have this 5 minutes guarant people will overrun anyway but somehow now they look that it's like oh finishes 55 past we can overrun a bit and they overrun like

three to five minutes but but you just don't care >> if you have like meeting after meeting and people overrun and you're always late and you start being super stressed and you don't have time to like just take notes after. >> Yeah. >> Prepare for the next meeting. >> Yeah. So you you traveled a lot. So do you have like time differences like issues or skills? >> Not much. Yeah. Because it's like when you when your schedule is that you don't sleep regularly then you kind of maybe you're just simply jetliked all the time. So then you don't notice when you travel. >> Okay. >> And I don't jump between continents that much because I I travel almost like one direction only all the time.

>> Like people very often like fly somewhere then go back. Right. you go for a conference and you go back and then sometimes they go to another continent, one week later they go back and you have like this two sleep adjustments. >> I move gradually in one direction usually. >> Well, that's nice. I I got you know time depresses issue. >> Okay. Um cool. I think we are uh done with today's talk and thank you Thomas for joining us today and welcome to Shanghai. Yeah. >> Thank you. Thank you so much. Cool.

Yeah.