Kartik Talwar I Opening Ceremony & Fireside Chat With Vitalik Buterin I ETHGlobal Prague 2025

[Music] Good evening everybody. My name is Cartek, one of the co-founders of ETH Global and welcome to ETH Global Prague. This is not a conference. You are here at an ETH global event and our mission is to onboard thousands of developers into the web3 ecosystem. What makes these events unique is a few things that we want to tell you about today.

This weekend is a hackathon. We're doing this thing after running our Pragma conference yesterday. And the cool part about all these things is that we want to focus on the builders. Yesterday was about everybody else in this space. Today is just people who want to build some cool [ __ ] These events and this weekend is for you to learn, get help, mentor others, and grow as individuals to get more advanced in how this ecosystem is evolving.

And the most important piece that we want you to take away this weekend is that showing is way more important than telling. We started this for fun over seven years ago, and that turned out to be more than just a one-time thing. We've been going across all different parts of the world, from many countries to many continents. constantly getting more and more people to be excited about web 3 from Tokyo to Lisbon to Waterlue to Paris, New York, India, London, Sydney, Brussels, Singapore, San Francisco, Bangkok at Defcon and we started this year off with our first Inerson event in Taipei and it's incredible to see the impact these events have made. We've done events now in six different continents, done over 150 of these, and successfully on boarded over 150,000 people.

So, I want to thank all of you for being part of our community, and this gets me even more excited to be here in Prague. This is our second time doing an inerson event in 2025, our first time in Prague. I want to thank everybody who's been so incredibly helpful for making this happen, especially the ETH Prague team. This would definitely not have happened without their support. And I want to see who actually is coming back as a past ETH global hacker.

Cheer if you are returning. Nice. This is one of our more unique events where we're getting more new people to uh to be part of it. There's over 825 attendees. 600 of them are hackers from 55 different countries.

And a third of you are brand new to building in web 3, which means they're existing developers and builders trying web 3 from web 2 for the first time. And I can't wait to see all the cool things that you built. And it's going to be even more fun because we're letting all of you use AI to help finish your project. So I expect even more ambitious things being tackled on for the next 36 hours. People are here from all different parts of the world, five continents represented.

This is absolutely amazing. This blows me away every time that people came in and made this the way they want to spend their weekend. And of course, so many different parts of Europe are just represented in this map. So, I want to meet all of you who are here from so many different places, especially for those of you who are joining us for the very first time. A warm welcome and we can't wait to see what you think.

Not only just the hackers, there's 110 partner representatives from all amazing protocols in this space that you get to use and you have used in the past. They are here to help you simplify a lot of your weekend. And we have 20 dedicated technical mentors who are going to be there to help you debug any of your problems. Whether it's on solidity or front end or backend or anything in between, you can actually get real-time help at any given time of this weekend. There's 16 workshops helping you get started with different protocols and web 3 in general.

And because this is not a conference, there will be just one talk this entire weekend. The best part of it is that you also get rewarded for building cool things. So there's going to be $150,000 in USD prizes being given away for being able to use different technologies and we'll tell you a little bit more about that very soon. All these prizes are thanks to our 14 partners who are here. you get to see all of them at different boots.

So, I want you to all ready up and help me thank all of our 14 partners that have made this possible. So, please give a massive round of applause to 1 in, Block Scout, Flare, Flow, Hyera, Layer Zero, Pit, Rootstock, Verair World, Yellow, Barrao, Hardat, and Protocol Labs. You can find people from all of these companies here and learn about how they work, how they can help simplify a lot of what you want to build and also any opportunities that you see to help them improve their product or help you get more help in improving your career. Let's talk about the weekend because I want to make sure that we spend the next 5 to 10 minutes just giving you all the key things that you need to know. I want to talk about all these things on how you can get help, how everything is structured, how judging works, how project submission works.

So, let's dive in. Hacking officially begins now, which means from now until 36 hours from now, 36 hours and 45 minutes from now, you have the all the time you can build on your projects. You can work on a project with up to five members on a team. It's also okay to work as an individual. You don't have to have a team, but we prefer that you do because it helps simplify how you spend your time over the next two and a half days.

36 hours and submissions are due next month. That's right. 9:00 am June 1st is when you have to be ready for your submissions. And here is what you need to know about this whole process. Thank you for that one person who found this funny.

Everything you are doing here must be done from scratch, which means you cannot add a feature to something you've been building before. We want to make sure everything is fair. You cannot add and build on pre-existing work. And anybody who's submitting a project will only win a prize if everybody who's part of the GitHub repo and on their team page is RSVPd properly. Which means if you added one of your team members and on your dashboard it doesn't show that they were confirmed, you have a problem.

You may want to make sure that they RSVP before midnight tonight because at midnight they will all be removed from your team. And if you win a prize, only people who were listed at the time of submission as a team member will be receiving their portion and we will not make exceptions. On top of that, we want to make sure this event is helpful for all of you as builders. No matter if you're junior or first time or you've been in this space for 10 years, we want to make sure that you get a lot of support and information. So, there's a lot of workshops that have happened.

You can see all of them on YouTube and it's ready for you to watch on demand and it focuses on everything from beginners to intermediate to experts. You get to learn about some complex things as well as how to get started with some basic SDKs. All this information in addition to a lot more documentation on best practices for smart contracts, deploying things, how do you keep things more secure are available on an info center where we have a getting started helpbook and everything was made available for you to look at from now, but also for any of the events we've done in the past. So if you are curious about how to use any protocol, chances are you can find that YouTube video by just putting in the name of that protocol and ETH global. I'm not going to cover everything here.

So if there's anything else you want to know, just ask us, find anybody on the global team or there's multiple other ways to reach us and we'll be able to clarify those things for you, especially on what counts as a pre-existing project. For example, it's totally okay to fork an existing open source project and add things to it. And if you are not sure if that counts, come talk to me or somebody else on our team and I'll happily clarify. Let's talk about security. All of you should have these wristbands.

We want to make sure that you have them with you at all times because this is the only thing that'll let you come inside the building. If you don't have your wristband, we will deny you entry, especially at night. And this is the only thing the security guards have been told to to look for. So, if you forgot your stuff and you forgot your wristband, you definitely have a problem because you will not be allowed. We also don't want you to leave your belongings unattended.

If you're planning on going home, take all your stuff with you. We are not going to be able to ensure that everybody's stuff will remain in the same place you put it. So, please make sure you are guarding your own items. Also, don't make a mess. There's a lot of garbage cans here.

If you're going to consume anything from drinks to food to snacks, please put them away and make sure that your work environment is clean and we don't do any accidental spills or damage any property. Let's talk about internet. Please do not use hotspots. The internet should be good. We're testing it constantly.

We've done testers in every corner. We found a bunch of things that were not working and have been fixed. Now, everything should be in a really good shape for all of you to have pretty good speeds with up to two devices at a minimum easily. We also have Ethernet cables if you need to plug in directly and you can get a Ethernet dongle if you don't have one by going back to the swag table. schedule and everything else that you need to know is part of the website.

That is the source of truth. Anytime we update something, we update the website first and then we tell all of you about it. So, if you are unsure about something, just go on eglobal.com/progue and you'll be ready to follow what's going on. In terms of communicating with all 800 of you, we're going to use Discord as our main source of platform.

That's where we're going to do all the announcements. That's where we're going to ping everybody else and you get to ping us. That's the one we monitor the most. And this is the place where all of our partners are ready with all of their team members who are here and also who are remote to just watch the channels about their own protocols to help answer any of your questions. And the same thing goes for a mentor.

If you can find somebody you need help from, just ask in the mentorship channel or ask anybody on the global team and we'll monitor that and we'll respond to you no matter when you ask those questions. And this simplifies a lot of things because you can put the question there and still try to find the relative person in the space. But we heavily encourage you to use Discord to ask questions. Of course, you are in person. Talk to each other.

This is one of those opportunities where you get to meet people from so many different parts of the world, 55 different countries, people from all over their experiences in being in this space. From founders to designers to developers, you get to learn from so many cool people. And we want you to uh take advantage of that. And also, if you're stuck on something that's not working for you as code, you can probably shout for help and there's a good chance that somebody would know how to debug that contract or that front-end library or understand what you need to do to get moving for your project. Quick things on the venue.

This is a venue that is a combination of two buildings. All of you are here. This is where the main hacking area is, but then we're also connected to a building across the hall. That's where food and additional hacking tables is and also where the judging is going to be. I'm not going to go through this map.

But just remember that if you haven't gone past the other side. If you exit from this hall, there's a whole separate building connected through a hallway. And the venue will be open at all times. So you can come in anytime. It's 247.

And just to make it clear, because we keep getting asked this a lot, there'll be a lot of food. Dinner and a midnight snack today, brunch and dinner and a midnight snack tomorrow, and breakfast and lunch tomorrow. And then we're going to do an afterparty here in this building after the whole hackathon ends. All this is also on the website. You can read up all of this.

And for today, we're going to do some traditional Czech food as a midnight snack. So, promptly at midnight is when you'll see it. And we're going to continue the same theme tomorrow. Of course, there's a lot of cool things that you can take up take away and we want to make sure you have great memories. So, there's also a photo booth that's digital and you can go at the the up where you came in from the stairs and take a photo and get that emailed to you.

And we hope that you get to uh take a photo with your whole team. And uh we actually recommend you do one right now and then on Sunday so you can see your faces depending on how Sunday goes. We're also going to be making sure that you get to stay active. Uh there's going to be yoga sessions here in workshop room C1. There's going to be three sessions tomorrow from 12 1 and two.

So if you're interested in this, check it out. And there's also a really cool NFT gallery that we're piloting where a lot of our attendees drew and made submissions uh to just really really cool things about our ecosystem. you get to uh see all these galleries are going to be rotating and you get to mint these as NFTs uh using POP. So, we encourage you to check out all the monitors in the hallways. We'll show some of these uh these artworks and uh we hope that you enjoy some of the amazing work that our community has done.

And if you haven't already picked up your swag bag, we got a pretty badass swag bag for you which includes an actual bag and a bunch of swag in it. um shirt, sticker sheet, some amazing check wafers, toothbrush, and some fidget things that I'm sure you'll use to annoy everybody else around you. I want to quickly talk about being on chain because we want to make sure all of you are being given the opportunity to make things onchain a lot more smoother for us and all of you. Uh there's two things that are important here. The NFC bands and hacker pack.

All the wristbands for everybody here are special chips that are actual Ethereum wallets and they can sign Ethereum transactions. These are P256 curve compatible which means every wallet is a private key that signs it. And thanks to ARCs, you also have an SDK. So if you scan anybody's wristband, you'll get to expose a lot of cool things from it. Uh every wristband is personalized to the person wearing it, which means it has context for who you are and where you've been within the global ecosystem.

But if you scan someone else's wristband, you'll get the option to connect with them and they get to share any contact information that they want with you. So you get to track basically who you've met and they get to specify, hey, here's my discord, here's my telegram or here's my email. They can toggle what they would like to share with each of you and this becomes part of your weekend here because these are going to be onchain transactions at a station, don't worry. And we want to make sure that we can show you that web3 is at a point where it can be applied day-to-day. The other cool thing is you can actually go to arxs.

org and go to their docs page and you have a simple JavaScript SDK that you can use and you can actually code things with these wristbands and you can have somebody sign it and you can build an app that lets you use each person's private key and it's always awesome to have people try this out and you can just take them away at home and use this in general ever because this is the proper Ethereum wallet. The other thing is hacker pack. Everybody who's a hacker here ended up minting these things and we want to make sure these things help you be more successful as a developer. A hacker pack is a special NFT contract that has a bunch of other features that let us blend in onchain offchain perks. Some of these perks that are relevant here is you all get access to a test net which has 19 different chains.

You can request a lot of test net funds and never have to worry about how do I get the relative funds. Everybody here will also be able to get an ENS lens and a Farcaster account. We'll cover all the cost thanks to the support from all three of these protocols. There's a lot more perks and credits that you get to use as developers. I want to make sure that you are ready to actually take your project from just a hackathon to everything that's next.

So everything here is on eglobal.com/packs and in global.comfet. I want to quickly talk about the prizes and how you can be ready for submission. So there's going to be two things that are important here.

There's going to be a partner prize and a finalist prize. When you pick wanting to go through just a partner prize process, you will have the option to select up to three partners. That is the max limit. You cannot apply for more than three partners. And any partner you select, you'll qualify for all prizes under that partner.

So, if a protocol has three prizes, you just have to select the protocol and they'll consider you for all three. You don't have to worry about, oh, I didn't want this category and I didn't want the other. Once you are submitting your project, you get to select this. And as long as you select this, the partner will get a notification saying they have to review you. The other part is the finalist track.

If you want to be part of the finalist track, you will be considered to be one of the top 10 projects for E Global Prague. These are the best projects we see from this event. They go through a special judging process and the prizes that they are eligible for in addition to any partner prizes is that they will be eligible for the finalist pack and they'll also come on stage and demo their project to everybody else on Sunday. In order to be a finalist, you have to follow a few rules. One is that your project must start from scratch, which is the same for everybody else.

But you have to be using version control on Git and GitHub properly. You cannot just commit one 100,000 line commit and that's all it says. You need to be showing us a proper history of how it started, how you've improved it, how your team members are contributing. Everything must be auditable because we want to see the history of how you started to where you end it. It must be open source.

It must be deployed and live. So local host will not qualify. Put it on Versel, put it on any other place. Try to do an experience that somebody else can go on their phones and try. You don't have to deploy your contracts on mainet, but you cannot do this with the expectation that somebody will have access to your computer.

They need to be able to see that on their own computer. And if you end up being a finalist, you will be qualifying for the ETH global proc finalist pack. This is an additional NFT pack as part of our PAX contract, but give you access to a bunch more things. Means you'll get ETH Global Plus for the next 12 months, a th000 USDC to every member on your team. We'll be giving you a $500 flight credit to come to the next E global hackathon and an onchain proof of you being one of the top hackers of Ethereum ecosystem and we do a lot of private events with different founders in this space to help them answer any of your questions and give you guidance on how do you can take your project to the next level and a lot of credits to be able to deploy and not have to worry about infrastructure is over $15,000 in credits and every single person on that team will get these.

So those are the two things you can do for the prizes. But this is how you want to integrate all those things. At nine o'clock judging will prize submissions will be due and that's when judging starts. There you will have the option to pick the partner track or just the finalist judging track. Now both of these things happen in parallel and they go from 9:30 to 12:30.

If you select a partner judging, it means that you are going to only qualify for the partner prizes, but you don't have any participation in being one of the top 10 teams. If you do the top 10 finalist prize, you will be eligible for both, but then you have to do one extra presentation to a separate set of judges who will determine if you're one of the the more interesting projects from this entire weekend. All you have to do to qualify for a partner prize is make sure that you select that partner in your submission form and the second you do that they'll get notified and you can come and present your demo to their booth. Every partner will have a booth and they'll be able to look at your submission. They will be able to ask you questions and sometimes they'll just look at the submission itself and only message you if they have any questions.

So, make sure that you are giving as much detail as possible because it's not a requirement for you to go to every partner and demo your project. The order doesn't matter. If you have three partners you're presenting to and you want to be a finalist, that means you have four possible demos. You can go to any of them in any order. It's going to be first come, first serve.

But if you want to do a finalist judging, you will have seven minutes for your demo. Four of this is for a demo and three minutes for Q&A. And we recommend that you have a video submission on your project so that if something goes wrong or if a judge needs to look at your submission again, they can play a video recording of your demo and try to understand what happened and then also ask you any questions. We really recommend that you don't do this thing last minute because it will take you at least 40 minutes to work on your submission because you have to put in a lot of details. You have to put in screenshots.

You have to walk us through specific lines of code. So try to spend as much time as possible getting ready for your submission and you get to see what all the questions are tomorrow morning. So you don't have to worry about doing this thing last minute. You will have over 24 hours of heads up on what information you need. All this happens on your dashboard.

You will be able to see everything very easily. We've done this thing a lot of times and there's a few things that we recommend that you do to stand out as a submission. We really recommend you record a video. Doesn't matter if it's finalist track or not. Just having the ability for somebody to pause, look, and absorb what you're doing just gives you and everybody else a good odds of communicating what was written or what was built.

We recommend that you spend more time on the demo and less on the slides. So, do not spend more than 30 to 40 seconds on introductions. Just get right to the demo and why you built it and how it works. We really want you to use Git and GitHub properly. We should be able to see how something is being used and imported and where you fixed the bug, all those things.

Just any best coding practice, we want you to follow that. And do not be lazy about putting in descriptions for your project. You should really talk about how it works, why you chose to build it, how something is happening behind the scenes, especially if you're doing something complicated. Communicate the steps of this is how we got to here and this is what happens in this step. All this helps everybody understand our partners, our judges, me really figure out what's going on for a given project.

So ultimately our advice is make the submission good because whatever you see on your showcase project page is what helps you be more successful for your submission. And anything else that we didn't talk about? Well, we want to say make sure we have opportunities here. So there's a lot of technical mentors. They're all wearing vests.

There's a mentor station as well at the end of this hall. And you can also ping anybody in the mentorship channel. So find any of these people who are not construction workers and ask them how to fix your code. They'll be here this entire time, whether it's 400 pm or 4:00 a.m.

There's also going to be a feedback session. So if you're looking to get feedback on what you built so far, you can go there tomorrow at the station at the end of this hall and get live feedback from some of our mentors and partners. Anything else that I missed? You can ask us any questions, ping us on Discord. There's also a global help desk.

You can find a person on our team. You can message us on Discord, email, Twitter, anywhere. Find us. We'll be happy to help all of you. And the one thing I want to also talk about is our code of conduct.

This is an inerson event. There's a few things we want you to be aware of. Everything here is written out as part of the event rules in e-global.com/ruules. It covers how to behave here, how to be successful here, and what will not be tolerated.

The short of it is please be respectful and harassment in any form to anybody will not be tolerated. We will take prompt action and if you see something that shouldn't be the way it is, please come and find us and report it and we'll be able to take action immediately. And the most important part of this is that anything you build here today will be solely owned by you and your team equally which means neither ETH global none of our partners have access to the IP rights of your work. You can do whatever you like with your project after this event and you get to own everything that comes out of this. That was all the logistics we want you to know.

Want to get into a couple things that uh will get you excited and we want to bring on a couple partners to tell you about who they are, what they're excited to do. And I want to start off by bringing on D Sasha to tell you all about 1 in. So please give her a big round of applause. Hello everyone guys. How are you doing?

Hope you enjoy the the beginning of the of this beautiful hack GM or GN whatever you prefer. I want to see the hands of the hikers who are here the first time. I've met them a couple of them previously. Whoa. Hello guys.

And I'm very glad that we have such a big new beginners community. For those who are here for the first time, one inch was built at the Ethereum global hackathon in 2019 in New York. Th those was one our co-founders who are building and this that means that um no so we've built this hackathon and now the here we are uh we have almost uh 655 billion of uh uh total volume. There is a beautiful core contributors 160 who are distributed all over the globe. Um and you see all these numbers and uh this is basically thanks to Ethereum global community we are now here and uh we built the hackathon and now uh the time is yours and this is what we prepared for you.

So first is basically the bigger price. We released recently the crosschain solution and the ability for you is to basically get this price is to build fusion plus across those different chains. The second price is basically to play with limited order protocol and maybe build ext and or extend build some custom hooks on libert order protocol. The last one is basically we recently implement the developer portal where you have almost uh 12 plus different APIs and you can brainstorm on the idea and where basically you can implement our API swap API pricing API NFT API any APIs. This is basically more information about what we are um basically prepared for you.

There is one booth on the left side. There is a Tanner, Alex, myself, Matt who are welcome you and are ready to answer all your questions and we have 10 seconds but I just want to thank you and I wish you good luck and let's just keep building together. Thank you guys. Thank you. Amazing.

Next up, I want to bring on Curl to tell you all about Block Scout. Let's give him a big round of applause. Hello hackers. Hope you're enjoying this evening so far in the hackathon. So the blocks ecosystem has been growing crazy and huge for the past couple of years.

We are now proudly covering more than 2,000 networks across the ecosystem, different layers, networks, rollups, and across any dimension you can think of. just 200 of those around something like that is hosted by our team as part of a explorer as a service offering we have or as part of oneclick deployment tool called autoscout we launched this year and the rest of the of this 2,00 instances is being host out there since the product is open source by the rest of the by the clients themselves by the networks by the rollup as a service providers which host more than a thousand instances themselves and blog scout is the partner to is the primary blog explorer for all of them. So yeah, the numbers are growing huge. Super incredible. Now a few words about the hackathon bounty we have.

So we have a lot to talk about. So I'll be try to be quick. So the first bounty we have is just for the best use of blog scout in any way possible through our APIs. So you can use the APIs directly. You can use the API through the AI agent of your choice.

Any stuff like any even simple integrations like doing an LLM.txt file or doing a chatgpt plugin for the blog scout or integrating with any MCP server counts towards this bounty as well. Then we have the balance for the blessed blog scout UI SDK integration for enriching the transaction history and notification popups in your front end. Then we have a cool the most requested bounty so far was the about Blog Scout merits which is our loyalty program at Blog Scout. And basically if you can find a cool way to showcase the merit balances of users within your builds or find a way to distribute merits to your users as part of your workflow.

Be happy to apply to this bounty as well. And then for the rest of you, we have a pool price which will be shared among all qualifying projects who integrate Blog Scout in any way possible or just put a link to Blog Scout, verify contracts, any usage counts. I'm running out of time. Thank you. Good luck.

Become the king of this hackathon and take the crown literally. Amazing. Thank you. And next up, we want to jump right into the one talk that we have scheduled for this weekend. And for this one, like to bring on Vitalic to talk about a lot of cool things on privacy, AI, Dows, and a bunch of other things.

Welcome to day two. And you know, welcome to day, you know, whatever your day. How many of these have you done this year? Uh, four. Okay.

Yeah, pretty good. More to go or something like that. Amazing. 14 a year. Do you think we'll ever see Hogen 40?

Actually, yes. Um, but will you see me all of them? No. But we are trying to do that and um that's one of the things that we're actually super excited about as an experiment. And uh if uh if you go to the ETH global website, you'll see that we're ramping up uh doing more local engagements in different parts of the world.

M and um I'd like to see a world where we can do something every day in some part of the world. You know, there actually is a practical way for you to attend all of them virtually. Exactly. So, not just like virtually, the boring way, but you need to buy a Tesla bot. Yeah.

Like, okay. So, like remember Bitcoin conferences back in like 2013 and 14 and remember when like people like Charlie Shrem, you know, they'd have like legal trouble, so they would not be able to come in person, right? One of those iPads. Exactly. You'd have like a like a bot, right?

That's And then yeah, it's like an iPad on a stick and they can actually like move the iPad around and like interact with everyone like a person. Yeah, I just have to uh figure out my sleep schedule in that case. But uh I'm just betting on AI video to get better so I can just type out what I want to say and have it be from from cardic to chaic. Hopefully that'll be cathartic. That's good.

Okay. Um, let's start with uh talking about privacy. It's become definitely hot recently for uh some good things and for some uh not so good things that we've seen in in in the media. And my kind of first question to you is um are we in a good place right now to uh experiment more with privacy or do we got to wait for more tech and more regulations to uh to simplify and then will be a better time to uh try things out? I actually think now is the best time to really start getting back into like Sure.

Um so I guess maybe so maybe just like first off just from for the big picture like starting off with a bit of history right so uh who here in the uh audience uh has heard of Chamom ecash. Okay, so this is the uh you know there was like an entire lineage of attempts to create Bitcoin before Bitcoin. And one of the interesting things of the very first one from David Cham in 1982 is that it was not even decentralized. It actually was fully centralized. It was a server but it was private and um even the server uh could not actually tell like who was sending um how many coins to whom, right?

And then somehow you know we've evolved in this direction where we basically started doing the decentralization thing and then we uh kind of gave up on the privacy because there wasn't technology to do decentralization and privacy but then now we have ZK Snarks and so now it looks like uh you know we actually can get decentralization and privacy at the same time right and it's like interesting to look back at the much earlier versions the 80s we're talking about exactly the 80s the 90s like the goal was not decentral centralization. The goal was what was actually privacy, right? And uh I think it was uh like it was interesting to just remember and like think about that and uh like just basically keep in mind that you know privacy was really this very central part of the goal um the whole time and I would say for very good reason. Right? So you know what I just I find that interesting because at that time home computers were not a thing.

So that was government surveillance fears without tech actually being a big part of most people's lives. Well, you know what they say, right? The uh the mediocre predictor of the future predicts the car. The great predictor predicts the traffic jam. So, you know, these guys were like predicting three steps ahead of the traffic jam before we were even born.

Amazing. Yeah. So, and I think like you know the intuitive reason to worry about privacy I think like makes a lot of sense, right? which is uh like if you take as a a starting point that we value decentralization and you value decentralization because you're worried about centralized actors uh like doing something um wrong or like something negative that you don't expect. And then if you ask like just people in general like what specific things actually worry them the most in terms of abuse from centralized actors like normally it's not like them going in and editing numbers in your spreadsheet right there are times when it's like rugging you by changing the rules on you uh but uh a lot of the time it is all about like protecting your private data right and so like I think like if we want like really want to think about uh like actually meeting the needs that people have and especially yeah the needs that the the needs that have to do with um like avoiding risks of centralized actors that people have then privacy ends up being right at the center of it, right?

Yes. So, but it's also one of those things where what we see commonly here is that you only worry about being deplatformed um or being at risk of this once you've been affected, but it's harder to make that case dayto-day uh to to any regular person. Um so, while the goal is great, it becomes harder to get people to care. I uh I agree. I mean, I think for me personally, it's actually it definitely became like five times easier to care once uh like the coin media got really really good at like taking every single time I send someone some amount of ETH and turning into an article.

It's like, you know, like I send like 75 ETH to some charity and then they sell it because they obviously have expenses and like Vitalic Bdderin is selling ETH. So let's uh well you know at some point like the extremes of not having privacy yeah become obvious right but like as I've said um in uh do you believe that it it has to be the extremes or can there or should there be a middle ground here like is it yeah fully private only or maybe some so in terms of like who should care about privacy I do think it's a fact that the types of people who have like more very direct and concrete reason to worry about privacy are the types of people that like deviate from the norm in pretty much any direction. And if you add up all of those directions, then that's like a big portion of the population, right? And if that's not you today, that could be you at any point in the future, right? And so it's uh yeah in terms of like how uh how we actually get there and uh like how like what what we should be willing to sacrifice how much effort we should be willing to put in like it's definitely not an easy battle right so I mean one example is like so if we if we think about signal right uh so h what percent of your like instant messaging would you say is on signal right now about 20 okay how much was it like three years years ago, five.

I think for me this year it was about 50 and I think three years ago it was also like five and and uh actually for ironically I think like pav well actually not ironically at all. pebbled of being arrested was probably like the you know like really big trigger for a lot of people right it's this reminder that that now you're worried about the side effect of someone else's exactly the uh you know the fact that the platform is centralized is not just this like totally abstract thing in the cloud it's like oh wait there's a database somebody has the password to the database and uh sometimes like the person with the password to the database ends up uh just like being not who you expect like uh I think there was this um incident um at the end of last year when uh there were the mandatory wiretap data that the I think FBI collects in the US like got hacked by China and so it's like well you know you thought it was one country's law enforcement well guess what now there's a totally different other country now right so it's uh it is one of those things where like the appeal is often unintuitive but there are these moments when it becomes intuitive and like what's where signal I think did a good job is Like it's definitely far from perfect UX- wise, right? If it was perfect, like both our numbers would be 100. But uh like it did it has come quite close. I think probably closer than most crypto projects to actually crossing the chasm.

Absolutely. Well, so I mean I want to bring this back to understanding how do you practically do this for money? Yes. So the kind of flagship technology that I think a lot of us have been talking about I mean using hacking with for the past uh I mean really decade is zk snarks and uh like who here uh has uh used a zk snark protocol more than two years ago more than five years ago more than eight years ago more 10. So a couple of of Zcash users here, right?

Yeah. So it's uh now one thing, do you remember what your experience using Zcash was like? Back in 2016, I had to wait three hours for the note to sync. Mhm. And then I was able to send a transaction and uh and then it took a few minutes.

It was like 8 to 10 minutes, 16 minutes or so to generate that proof. And it went like just a normal Bitcoin node. And I only sent it to another account I created to myself because I didn't have anybody else to send it to. And I was like, "Oh, this is cool." But then I didn't use it after.

Oh, wow. Wait, 16 minutes? That's like about as long as it takes for my GBU laptop to generate a video. Uh I did not have a good laptop back then. both of you know if you did not then uh I remember like I remember the officially advertised numbers were like uh two minutes right and of course definitely exactly best case eight minutes if you had nothing else running on your computer exactly you know the numbers always a lie right they always assume you know you're on internet's perfect your computer is like a a topend crazy Mac I mean TPS exactly um so um when was the last time that you used a ZK application I tried a bunch of stuff on the Aztec test net a month ago.

How do you remember how long it took you to sign? Like less than 5 seconds. Less than 5 seconds, right? So that's like probably less than it takes for like the New York Times website to load. Significantly less.

Yeah. Okay, congrats. So we're web two competitive. No, that's great. Um well, I guess my my kind of push back is well you're talking about that feature um being more accessible and faster, but uh you haven't made the case for how do you actually make that part of an existing experience, especially on Yeah.

So, super important points, right? I think uh for because for CK Starks like the bigger bottleneck isn't to use snarks or Starks both. Okay. Um Starks are a type of snark. Um but the like I think since we're use talking about privacy we we'll probably mainly talk about like elliptic curors and arcs because starks are like too big for most of those things.

Uh but uh basically uh Devx is like probably even more important than UX right and uh I mean don't get me wrong like UX has problems like for some reason when I uh want to make a private payment in Railway it takes me like four clicks instead of one and uh you know I hope that they fix that. I mean I hope it just railway just gets integrated into existing wallets but like from a DevX perspective right you know you have to learn a totally separate language you have to understand like what this thing even is you have to understand like what are verification keys and while we're still on growth 16 and like I'm hoping we can get off of it in the next year or two um you know you have to worry about like per application trusted setups and that's like a concept most people have not heard about right but even you've introduced more than you have simplified. Exactly. But uh like if we compare this to three years ago, the difference is already night and day, right? But it's still based on performance, not on accessibility.

No. Oh, I I mean DevX. Sure. Yes. Yeah.

Like three years ago, you had to like like really figure out circum. Exactly. Well, three years ago, you had to really figure out circom and then like four years ago, you had to like actually be a cryptographer. And then now like you know there are like circuits for Merkel branches circuits for all kinds of things that you can just like pick off off the shelf right so you know we've improved right and uh like I think there are still place places where know things are left to go like I think the for me the real endame is when we can get a VM um inside of a snark provable client side um basically because uh like know the ultimate abstraction that's developer friendly just as like an existing VM that existing languages can plow to, right? But the problem is that you know the overhead of a VM is high and uh actually we have made huge strides in reducing the overhead of proving but the problem is that most of those strides have been for um like Starks specific or yeah Starks because the main problem is uh that we've been trying to do is like basically make ZKVMs faster right and the stark is 100 kilobytes and 100 kilobytes is like actually great for a lot of applications but if you want like something that goes onchain um or if you want something that like fits in a QR code that fits in an NFC then like it needs to be a few hundred bytes and so like that's something that is being worked on right so so we know that mathematically we can do this compression but it's just a performance thing or something else uh I think it's a performance thing okay like I think uh actually right now like if how do you reduce the size of the well so one technique you can always do right is you can take um a snark and you can put it into a snark and just do recursive okay exactly and that process takes like a constant amount of time and like that amount of time is a little bit too long right now but you know we can hope that it'll shrink right so but even if you just like look at the kinds of applications that we've already seen right so you know seen like railway um have have you used railway who does not know what railway is will give you a heads up okay I think it's what we're talking about okay yeah it's you know a thing that basically lets you send uh uh ETH and a few other tokens on Ethereum privately.

Um, so I have used it and um, you know, there have not been any Cointelegraph articles about who I sent the money to. Um, Kraken. Oh, no. I I'm not even sure if I have a Kraken account. Yeah.

Okay. So uh basically uh you know so we have uh like railway we have uh some really freaking cool zk ID stuff right like we have things where you can literally like take your um like passport and you make a zero knowledge proof that you have a passport and like prove that you are a person without revealing which person you are and yeah so so you prove that this is me and I'm from this country or I'm over this age um and you're kind of calling that as as part technically private information. You can prove you're not a bot. I mean, we have I mean I've actually I think talked even like here at this hackathon, I've already talked the two groups that are doing something to do with ZK proof of silver identity. Um so it's uh like it's getting very viable, right?

And like and from from the users point of view like the form factor is just like it's an app, you download it. um you know you'll like take out your phone and then you like tap the thing NFC and then you see like the rotating circle for a few seconds and then you know you're done right and but then from a devx perspective like you know there still has to be someone that does complicated circuits right but like that's uh like basically the difficulty of that is just dropping and I think like both the tech itself improving and LLM's making like all coding easier they're probably like improving that in lock step so that makes a lot of sense So maybe what I want to understand is where do you differentiate between ZK being helpful for just not storing all the information on chain versus helping actually increase privacy for the use of money. Okay. So good question. So I think top level right there's two big use cases of the technology which is scalability and privacy right because the property that these proofs have is you can make a proof of a really big computation and then the proof is really tiny fast to verify and also the proof hides your information right and uh here we're talking about the privacy use cases the scalability use cases are also super cool and they're going to be the thing that scales Ethereum by a lot over the next 5 years but for privacy right like what we're focusing on here and privacy.

So within that there's a few different use cases, right? So there's privacy of money and Zcash really pioneered there. Um and then tornado cache is like you can think of it as being a sort of like very minimalistic cloner Zcash on Ethereum. And then uh we now since then we've got privacy pools and railway that have these things that are basically using like a a proof mechanism to verifi or like generally dissociate from a particular input if they like don't wants to like if they wants to clearly separate their funds from that input. Uh so like that space is really expanding.

Um then separately there is the space of uh like privacy beyond the money sending use case. Um and I think yeah identity is isn't privacy beyond money the thing you just described? Yes. Identity identity is a big part of it. Okay.

Identity is actually not the only part of it. Right. But I mean for identity any sort of proof compression or or referencing information. Exactly. Um actually yeah I mean there's some like really interesting uh things.

So one of the things that I've noticed like making a lot of progress over the last year or so is like zk email zktls uh like basically zk rappers of like web two forms of trust. And the reason why I find this stuff useful is basically that uh like if you want to give someone a an account that is uh like basically like not fully decentralized but that is very easy to use and that does not rely on them on them trusting you. Then what you do is you give them a smart contract wallet and that smart contract wallet like basically lets you send a transaction if you can zk prove that you have a particular email address and like you can actually do this right and uh like this is an uh one of these you can call it an identity use case you can call it a credential I mean to me credentials are one part of identity and it's like solving a really valuable need right which is basically yeah you look we have this sort of trade-off curve between essentially decentralization and user friendliness and the more we can like expand that frontier and create the technologies that give us both at the same time the better right and uh like basically this theme of making like web 3 wrappers around web two things is uh like to me a really big part of that and Zeke yeah we know it works and exactly and we know it works and uh like generally as a user you don't want to be like blasting your email address out to everyone and so you need privacy Right. Yeah. How?

Well, okay. So, things are faster. There's better for a developer, better SDKs to not have you worry about a lot of these things. Um, it does not still eliminate the possibility of you knowing exactly what you need to do to help ensure that you as a developer are responsible for protecting someone else's uh from identity to to preventing any sort of information leaks. Um, and I think my kind of still criticism here is that I still need to understand some math that I don't think I can set an expectation to everybody else.

And uh, and that's a still a big bottleneck. Um and on top of that um it seems like I have not heard good things uh good alternatives around how do you actually make things more private if you care about doing things on Ethereum mainet because you have to now worry about public and private state and it's not an easy fun thing to do. So right um how do you simplify that so people actually can try more things instead of coming up with basically MacGyver solutions that uh may not actually solve the problem that you're trying to solve? Yeah, you know I mean a few good questions there, right? So I think one of them is around the issue of private state like if you have applications that are privacy preserving then there are like secrets that you as a user will need to hold and like sometimes those secrets are kind of secrets that you already have in the sense of like something like a ZK passport.

The secret basically just is your passport. Um but you know you might imagine something more complicated right you might uh you know imagine some application where like you have a zk passport and then that zk passport allows you to do some number of operations every day and then to prove that you haven't already done too many operations you have to keep track of the ones that you've already done. So you could have like once you get your applications become more complicated there's definitely like private state that you have to store um and uh like this is one of those tricky things right a second thing that's like I think a really good point that you brought up is uh that you know you have technology that can give really amazing privacy and security guarantees in theory but uh to a user like how do you actually how does a user even know whether or not they're getting those guarantees in practice Right? Like if uh you as a user are told like, "Hey, put your passport up to this thing and like watch the circle go around." Like how do you actually know like whether or not it's like actually sending all of your data, whether or not it's like leaking extra data that it shouldn't?

Um and uh like so for so far like a lot of that info does not exist yet, right? And it should exist. And like basically I think uh we need people to build the equivalent of uh like ether scan contract verification is like one good example right like basically the equivalent of that for zk and then times 10. So, well, you know, I the thing we're discussing is like why is it the right time to build or care about building things for privacy, right? Um, here is one stupid question.

Uh, why bother trying anything now? If I can wait maybe two more years and have a VM that just is computing everything in private because I think it'll be fast enough and just build apps the way I would normally because I assume any computation in it will be private and not have to worry about anything else because I'll have something else transpile my code into giving me those guarantees. Mhm. Yeah. I think uh I would say like my answer for this is kind of big picture which is that I think on a civilizational level there's quite a bit of urgency in terms of uh getting some privacy out there.

Like basically I see us being on a default path to losing um quite like quite a lot of privacy quite quickly. Um and so uh you know we've seen you know like payments are like be very far along the way to becoming electronic and uh there's a risk that if the only form of payment is a form of payment that's not privacy preserving then like we're basically locked into an equilibrium where like bas surveillance of like everything financial becomes the norm. Um similar things with AI um you know AI is uh like in very not private similar things with all kinds of emerging technologies especially technologies that where the way like they're really interacting with you know your deepest thoughts they're interacting with a lot of your private activity. Um the identity problem um basically you know the era of uh on the internet nobody knows you're a dog is basically over and uh the reason is that if nobody knows you're a dog then nobody knows you're a bot and we don't want an internet where uh you can just have like 50 million bots taking over everything. Well, you already are there if I'm being honest.

Um, the reason why I asked that question is because if I compare this to something like DeFi, um, I would say that it's probably fair to say that there are no longer any outstanding primitives that you have to invent in DeFi. I can just assume that I have a lending protocol that's big enough that'll kind of satisfy my needs. I have an options protocol. I got a a DEX and an AMM. and now I just get to sort of do these uh combinations of things to come up with the product I want.

Um is it not better off betting on that same eventuality for ZK where I can say here is one protocol that manages sending money from A to B and uh and it'll do that in a preserving way and uh I'll wait for maybe two or three other primitives that help me do some sort of common computation and I'd rather treat that as public infrastructure in the sense these are not companies. You could be part of the VM. It could just be one of these things that exist. And I now build things on top instead of what it seems like we're forced to do is do something on our own every time and then hope that these things talk to each other because you're not really going to get that with GR 16. Yeah.

I mean I and I think I would say yeah like there's definitely a class of developers who really should wait two years. Um, but I also think there's a really important class of developers that need to start like laying the infrastructure now. And um, you know, the reason why I want to start uh laying the infrastructure now is basically because uh, laying the infrastructure and like getting the loop started on having at least some level of user base is something that can inform the design of like basically the rest of the entire stack of the ecosystem, right? And so I mean like we want to make sure that we don't get like walked in to choices that are needlessly privacy unfriendly and uh you know basically we need scouts right and uh we need power users and uh and also we need like there are categories of applications that I think uh like basically are ready for prime time but I feel like the only way you get to get there faster is by making that the default. Yes.

for where the code is deployed and not have that be additional thing somebody has to take on. Um, so you're better off betting on a private layer 2 existing than introducing this as a new product into the mix of things that are also not private. Yeah. So the private layer 2 topic is interesting. Um, so uh I think one one of the challenges there is well a couple of things right.

So one of them is that uh like there are privacy applications where the optimal thing is to actually have everything be offchain or not everything but like most things be offchain and uh like one type of privacy that I generally think a lot of people underrate is privacy of metadata and uh like even signal breaks here right so like for example you know if I send a signal message to you then uh like even signal the company they have no idea what the contents are they know the last IP exactly Well, they know a little bit of the content because they know it's a tiny text message and not like a 100 megabyte file, but they know the IP, they know like both of our usernames. um and then phone numbers and then even if you switch over to like accession or something else that anonymizes that like there's just the fact that a user sent like user A sent a message to user B at time um you know like 21105 on May 30th Prague time right and these days AI is like freaking powerful right and so even just from a pattern of like a few dozen messages of like one person sending to another like that's enough to deanmize people, right? And often that's even enough to like get things the conversation topic like message arrival like times like you can freaking do sentiment analysis on that, right? And uh yeah and there I think there's like blockchain analoges of this where there are people who like there's a strong case for a uh like an application development paradigm that just says like dump everything on chain just in case. And uh but the risks of that in a privacy context are dump everything by encrypting it or just um well encrypting or not encrypting right like uh there is such a thing as like social media protocols where everything is just a transaction right but the the risk of doing that in a privacy context like there's two risks right one is if it's encrypted then uh there's you know in the future when quantum computers come there might be a big oops and everything gets decrypted and then two is uh like this metadata issue, right?

Like basically if you put everything on chain then like that still reveals a lot of metadata. You have to worry about where transactions got sent from, when transactions got sent like and uh like often just keeping data local is just is like actually a very tried and true way of uh of protecting privacy. But uh if you're going to keep data local then like that that does require like actually this like extra tooling, right? And I think one of the things we should actually yeah like start thinking about doing is just like actually building that tooling. Um and so maybe one of the cases of this is in for I'm like starting this now is that uh good general purpose tools often come from having like a reasonably diverse set of motivating examples.

And so if you start building like tools for uh with some motivating examples then like at some point you realize like wait these things have these five things in common and then we need to combine right and so and I think like the the faster we can uh really get our feet off the off the ground running in terms of this and like really think about like even a couple of different categories of of applications interacting with each other. So private money, private identity, private voting, I would start with those three. Um and then obviously private messaging. Um then uh like we once you even have like a suite and you see how every everything starts to fit together like you know like that at the private messaging level you roll down to just sending arbitrary packets and you might do whatever at that point. Exactly.

Um I want to just do a couple other things. Um we kind of brought up uh possible doomsday scenario with AI uh tracing back who somebody is. Um how often do you use uh any LLM in a day-to-day basis? Quite often. Um, so you know, I've definitely like started playing around with uh the local LLMs more.

So um you know I have a a Q130B and uh you know like I I ask things to through Oama when I can and like uh I mean what's interesting is uh there are times when like local first is like great even from a UX perspective like the internet often breaks and like sometimes like literally loading Google translate is actually slower than like opening up a llama and saying like hey what's this word in English or what's this word in German? What are you using it for? I guess more specifically, have you uh used it to uh help improve Ethereum in some form? Oh, quite often. Yeah, I mean all kinds of things actually.

So, uh coding is one use case, right? So, um often you know it does a very good kind of first pass of writing uh certain types of code. This is Python or Solidity. Um Python, it's done I've I've done it with Python, I've done it with Sodity, I've done it with Viper. Um so works well for that.

Um obvious like research. Um so uh um like checking um different kinds of ideas like uh yeah even like talking to to it about like quadratic funding mechanisms. Yeah. Like it's far from perfect, right? Asking criticisms or helping run, you know, with scenarios like what's like an example of uh some of both, right?

because like you can do it for you can definitely use it for to like try to criticize your ideas. You can use it to just like make simulations. Um like basically the thing that all of these tools make really easy to do is uh like whenever you want like if you just want like a 100 line scripts to just like run and test something then you can just go do it and then the script appears within like 30 seconds then you can run it right. So, but you in this case, at least for research, it's still it's the onus is still on you to verify that it didn't hallucinate and uh that the conclusions are correct and uh just from your experience, how often does that happen? Um it's definitely often incorrect.

It definitely often misses important things. Um so I defin it's like it's definitely something that's supposed to like work with you and not replace you, right? So the analogy that I use is uh so uh the um do you know what year um chess AI started beating average human beings? Is that the same as the Casper mention? No, it's actually earlier earlier.

Yeah. 95 1980 roughly. Yeah. And then and then when did AI start beating the smartest human at chess? Who knows?

96. 1996 Gary Gary Kasparov versus Deep Blue. Now, when did human plus AI stop being better than pure AI? 2003. 2017.

17. Yeah. So, actually, yeah. And this is chess, not go. This is this is chess.

Yeah. Now, with Go, I'm actually It's a good question. I'm not even sure whether we're at we're at the points where like uh AI is better than human plus. Uh uh I think that was kind of the second version of the Deep Mind bot that did prove that, but I forgot what year that is. Like if we think about like some of these more open-ended real world tasks, then for a lot of them there's like like we're entering this like regime where like basically the combination of the two is like stronger than one or stronger than the other.

And so it's a it's like a really important skill to, you know, be able to wield it, to be able to know when it's wrong, to be able to know when you should ask it, to know when you shouldn't ask it. [Music] Do you have any frameworks for like how do you get better at that? because if anything it's going to get even more the compounding of how intelligent these models are getting is uh is astonishing and uh if I'm trusting it to just create a script or propose something I have to match its uh in a way intensity to understand if this was correct or not and uh uh I don't know how to get better at that. Um, how how uh how have you improved your skill at uh identifying which crypto Twitter accounts are high signal and which ones are total bullshitters. Uh everything that's lost me money is bad and I think everything um just social trust um people people recommend and you have a curation in some form but um but what about accounts that like uh nobody else that you know of knows?

I'm still relying on some social discovery because uh the Twitter feed does something, right? Every now and then we'll discover a gem. Yeah. Um what what about like the usually search? Yeah.

If we get to like the finer points, right? So like thinking about like, hey, this account is uh good at topic A, but sometimes it has a weakness. Why? Oh, sure. I I see.

So you kind of like have a a model of experts type of uh world where you're saying this thing is really good at just this type of problem solving, but I shouldn't trust it for everything else. Right. And like you like I think you can uh like if you interact well like if you look at different accounts on Twitter enough like you can acquire an instinct for this sort of thing, right? Sure. But I think the issue is that a lot of times the answer you're looking for should behave deterministically.

Mh. Um, but human feedback for a lot of these LLMs are well just the spectrum is so wide that you may actually not have repeatability in a lot of cases, right? For the types of requests that you're giving. Yeah, agree. I mean, these things definitely have a nonzero error rate and of course, you know, humans also have a non-zero error rate.

Um, occasionally higher. Yes, indeed. Sometimes higher, sometimes lower. Um, and uh, you know, like you wanted like you wants to be like this is why you know you want to have uh, like you know like yourself and the bot and actually be able to cross check each other, right? How do do you think there's a world where we can just uh, prompt engineer fixing and improving the L1 soon enough?

Um, I mean I think we'll get there for parts of it. Um, I think Okay, I So here's like the actual thing that I would love to see. like coding AI do is I would love to see coding AI make formal verification extremely easy. Like I would love for it to be possible to like actually Yeah, that should be actually possible though. Exact.

Yeah. Like generate a machine verifiable proof that you know a particular ZK prover of the EVM actually does prove the EVM. um or that uh you know like a particular implementation of the EVM is deterministic or the particular um like generate like a hundred um intuitive rules that the EVM should follow like you know an account should only lose money with that account's permission and then automatically verify that it follows those rules like basically I think uh there is a uh like this north star that a lot of people are not realizing that is possible which is like basically much much closer than any than we've imagined for a long time to bug free code. And uh like what I expect is that I expect there will be domains where instead we're going to take all of that capability and and we're going to use that capability to write even more code. But for for for use cases we're okay we're okay with security and for usea where where we want high security and for use cases where like simplicity of the thing and simplicity of the goals is explicitly valued like I think we are going to be able to get to the point where we can credibly say there's a Ethereum protocol implementation that has zero bugs in it and like be an awesome world.

Um so Ethereum 4.0 No, that that's what the O stands for how many bugs it has. Um, one of your talks a few days ago, you kind of talked about um why it's important to not compromise on uh the values of decentralization and kind of things that you shouldn't uh break as principles. M uh what I'm curious on is uh if if the task is to make Ethereum L1 faster and better uh what are things that should not be compromised on and I'll start with that list and and I want to know how does that reconcile with the goal? Yeah, I think uh concrete things that worry me.

One is extreme low latency, right? Compromises on decentralization. Exactly. And like I actually yeah like I've thought about it for like a while and I think extreme low latency would worry me more than extreme high scale. What is extreme low latency like 1 second block times less than that?

Yeah. So 1 second not good. I don't think so for decentralization. Yeah. And uh so here's my argument for for this right.

Basically, uh if you think about the the ping time uh from um you know like say North America to like say Southeast Asia or to Latin America like 800 milliseconds, 600 milliseconds. Exactly. And so if you like the one slot takes multiple rounds of communication, right? So you need minimum three. Exactly.

And so if you have a 1 second slot time then like basically you know someone who is an 800 millisecond hop away will not be able to participate at all. Okay. And and uh can I push back on that? Um wasn't the original plan to do sharding which forces this to be simplified and this kind of the architecture you end up with anyway? So my concern with Starlink is that like Starlink is one type of mechanism.

It's also very not censorship resistant. Um it's uh like satellites have high sorry I said sharding. Oh sharding. Oh sorry. Okay.

I I heard staring because you are creating these yes sub locations and I would actually be okay with a one second slot because everything I know will be North. Okay. Yeah. A that's a good point because like the way that Ethereum sharding like the Ethereum school of thought in terms of sharding has historic historically worked is it's worked through random sampling, right? Okay.

And like actually with data availability sampling, we're also doing um random sampling except it's like kind of inverted and it's sort of bottom up instead of top down. So So you're giving counter argument to your own point. Yeah. But no, but the random samp but the the property the random sampling has is it like it samples randomly from the whole world, right? like or from like the entire set of nodes from the entire set of data and it has to do that for security.

There's uh so there's another school of thought of sharding which has been relatively expressed less actually Juan Bennett is a big fan of it which is local sharding like you imagine you know you have the Prague shard um and then you know you have the Singapore the Europe and then Exactly. And uh like my the reason why I've personally been like less enthusiastic about this is because it does introduce a lot of complexity to because like while it's hard to 51% attack all of Ethereum, it's easy for someone to like bring in the capital to 51% attack Prague, right? Like or any one city. And uh then like basically you have to worry about like what happens if individual cities get 51% attacked and like what that means and how do you recover from that? And uh but um but yeah, can you 51% attack long enough Mhm.

and actually affect more than a few blocks in this scenario? Yeah, I don't know. I mean, so just your your your damage area is to one block, right? Right. So I think the version of this that makes sense is actually L2s, right?

And Yeah, fair enough. Yeah. And because like L2s have a lot of similarities to shards and uh like if you actually like having more deliberately localized L2s I think eventually it'll happen right like if you think about like gaming like gaming servers have been like shorted by continent for like two decades now right so uh yeah I think we are going to have like a local um you know like regionwide potentially even even citywide alters and I think that's great. So if the ETH research team, everybody on Twitter and abundance of how much you should listen to them are all in favor of reducing block times to be a second. Uh I guess what do we do about that world and uh looks like we've compromised on decentralization according to that metric.

Um is that also affecting censorship resistance? I would say yes. Um so I think like geo decentralization fits into censorship resistance very directly and uh so first of all there's like a couple dimensions of geo decentralization right one is like literally which continent you're in and two is sort of are you close to the edges of or are you close to kind of the central hub like the arteries of the network or are you further remote like basically yeah like if you have uh nodes in six continents but the nodes are only in data centers like that is also a type of go decentralization and that's also a type of vulnerability right and like obviously you know you could imagine you know the CT like podcast bro saying like hey all the users with real money are in New York what's the problem if all the servers are in New York but the reality is that like even um you know the DJs in New York really value the censorship resistance that comes from nodes being distributed ac across mult jurisdictions in many continents Right. So, it's a property that benefits everyone. I mean, sure.

Yes. Um, and I agree. However, I guess my question is people have already expressed we would like to get block times to be 1 second. Uh, is that a thing that you're going to try to veto or or it still goes through? What's kind of the outcome of this very much in progress work that seems like the the research team is already doing?

Yeah. So I think uh there's a few things that could happen right. So one is uh that like a lot of the time you know there are compromises that end up making a lot of sense. So like for example one option is uh you know you could imagine some notion of like like microlocks where you have slots every 4 seconds and you have microlocks with some weaker form of consensus every 1 second and then like that could actually work. Is lock time the same as finality here?

Uh no or um the current plan is three swat finality. So uh so you get your 1second block time and 3 second finality right well in this case it would be 4 second blocks 12 second finality and 1second microb blocks right that's like one possible route and then if you do that then the one weakness of microblocks is that if one proposer disappears then uh like which only happens like 1% of the time then you have to wait the entire 4 seconds and microblock is not the same as pre-confirmation they're actually very similar um I think there the one big like variable in the design space is is basically does it just come from the builder or is there some interaction between the builder and the attesters and the interaction between builders and the attesters it does make it a little slower but it provides the benefit that like the trust doesn't just depend on the builder's reputation which is good for builder decentralization yeah so that's one um and then another type of compromise is like if we're explicitly okay with saying like hey you know L if L1 is going to scale a then like what's the point of L2 and like one of the points of L2 can be extremely low latency and so we just say like that is actually one of the value props of L2 and uh you know like if that's what you as an L2 have been building for then like great right that works um I'm just realizing there's not a lot of things we're disagreeing on so that's not that's not going well what do we have different opinions on how do you scale going to multiple events every year. Will we see a Vitalic bot? Oh. Um, what what what kind of bot are we talking about?

Like an an LLM to replace myself or are we talking about like the teleresence thing? The uh the human experience in in transmitted form. Yeah, I mean I uh I mean I would definitely love to like try the uh the the teleresence bot thing. Like I mean I remember it being really cool. Um I mean if we're going to talk to an LLM then you can do way more fun stuff than just cloning me, right?

You know, you can like uh I can't wait for you to be misqued by an LLM hallucin. Yeah, you can like 5050 um you know like test your opinions. Yeah, like me and Gregory Maxwell and like Anatoli. Um, you can uh, you know, like you can you can take like Cardic uh, but with like pro Bitcoin maximalism dialed up to like plus 99999. That's a good experiment.

That's a hackathon project. If uh, yeah, you want to see some samples. Mhm. Okay. Anything else we can disagree on?

How fast should we see a lot of these changes to the L1 that have been talked about in the last couple months? Yeah. Okay. Uh so my view is like uh generally I think we should scale L1 by uh about 10x over the next year and a bit based on delay. Why the in a bit?

Why why not be more ambitious here? Uh no year would be great. Um but based on like basically delayed execution, block level access lists, uh like networking level improvements, distributed history, storage, like the stuff that we already tal talked about. And then I I think at that point it would be a good time to like basically take a breather and like have a period to verify that we're actually doing okay on decentralization properties. possibly even, you know, like do a fork and like focus on kind of Oh, that's a controversial comic.

Hold on. Say that one more time, please. Yeah. Um, take a breather, verify that we're fine on on the decentralization properties and uh like simply do a fork to tie up loose ends in terms of Oh, you mean a hard fork, not a create a second chain. Exactly.

Yes. Okay. Yeah. Yeah. Versus soft forks.

Yeah. For Yeah. tie up loose ends on you know like account abstraction censorship resistance possibly you know like VM improvements um simplify the protocol switch to SS like stuff that doesn't kind of you know like threaten decentralization and at the same time I mean like we become more confident and then uh once zkms uh like really become mature enough that we can rely on them then like go go even even beyond that. Um, so like that's my approach. I mean, I know there are people that are kind of even more like more like brave than I am and they, you know, don't believe in the take a breather step and they believe that we should like literally commit today to doing a thousandx.

Like I don't believe in that. Um, but uh I mean in general like I do think there's a lot of room to scale safely. That's awesome. Um maybe as the last question um what do you think Ethereum should steal from another ecosystem that is working well and is not working well in the Ethereum ecosystem? Yeah, it's uh I'm I'm thinking I'm going going through the the list because like I mean you know like a lot to be just technical.

It's a it's a very uh broad abstract question indeed. Um, I mean I think like ecosystems outside of Ethereum that I like respect for their mission like orientation. Zcash is one, Celestia is one. I think it's like actually up there in terms of having its like you know like actual cippher punk content. Um, Bitcoin also is one.

What would you steal from those? We steal them. I think uh there's like good um aspects of uh of the culture um in um a lot of those cases. Um also uh there's uh a lot of like like as we discuss the uh like scaling while remaining decentralized. I think we should talk to a lot of Bitcoin people and like really understand well like basically what is the concrete thing that they've been that that they've been wanting out of small blocks and uh ensure that uh like we actually can uh retain uh th like those types of benefits.

Okay, I think we finally found something we disagree on. Okay, my answer is we should be as loud as possible. Uhhuh. And do the best we can to talk about all the things that are happening in our ecosystem which uh we kind of rely on hope to uh to be the propagator. Okay.

Which ecosystem are we going to learn the loudness from? Which ecosystem are we learning this loudness from? uh basically every ecosystem that has mostly no product but still has surprisingly high amount of attention. Okay. Um okay.

You know what if we want to go into bravery um I think uh I mean the thing we can be careful here. Yeah. I mean the thing we can learn from Tron is uh like uh you know like actually uh like couple more like basically taking this kind of uh you know like go go approach and like actually scaling out and like actually delivering a concrete thing that benefits huge numbers of people in developing countries. Yeah. Um yeah um let's see what else what do we learn from Bitconnect?

They're good at dancing. It turns out uh stable coins are in high demand if uh if that thought is in anything. People will happily do algorithmic stable coin experiments. That's true. Um no, I think that was Teruna.

Um fair enough. Yeah, it can be both. It's not it's it's sadly not a competition, but somebody's still trying to out compete here. I know. I thought the big connect conferences like you know the like they felt production value was good.

Yeah, that's a good takeaway. Mhm. What have we learned from the US dollar? Um, turns out that uh uh that you definitely have consequences uh if you uh if you just suddenly increase the supply and change uh the max supply number. That's a good point.

Um I was going to say network effects matter, but there are definitely things that are powerful enough that can break network effects. Intimidation usually works. Turns out um especially if you get to somebody 70 years ago. Are you saying Ethereum needs a navy? I don't want that.

Navy Dow. I don't know. Can we buy secondhand Navy submarines on eBay? Probably. Probably.

Yep. I mean, while we're at it, let's Salana. Salana. Yeah. Um you know, it's like some of everything, right?

you know, the they're definitely one of the chains that's, you know, got that like attitude of just Yeah. being willing to talk about their stuff, you know. Um, Bitcoin. Yeah. Coin.

Um, oh, sorry, Bitcoin. Oh, Bitcoin. Um, yeah, we went through Bitcoin. I mean I think uh like VAT there's like strengths and weaknesses of you know the kind of uncompromising approach to Bitcoin maximalism that you know we can uh learn from the successes and the failures of um yeah we can steal making ETH money. Perfect.

Well, Metallic, thank you so much. And uh let's give a big round of applause. Couple last minute things for all of you to uh to know. There's a lot of prizes here. $150,000.

1-in blocks are giving $20,000 in prizes. Flow, Flare, Hideera, Lair Zero, Pit, Rootstock, Vlayair Yellow, and World are giving 10,000 each in prizes. And Barabor, and Protocol Labs at $5,000 in prizes. You can check out all that information on the ETH Global website. And if you're still looking for a team, there's a session happening in room C1 where you can find other team members and get ideas.

And with that, we want you to have fun, enjoy yourself, and make sure that you realize this is about experimentation and learning. So try your best and present whatever you get done by Sunday. So thank you so much for being here, and let's get hacking. [Music]