[English][Pop-x HK] Deep Funding and Public Goods Funding

Transcript

Yeah. Is this fine? Yes. So I talk uh I've heard that Indian accent is hard to understand in Hong Kong. So if you don't understand just like let me know. I'll also try talking more slowly so that you can uh follow along. Um so my name is Danch and I joined the EF around two months ago. Um I'll start with a story about how I got the job at EF. um kind of I was tweeting every day and in uh last year in Berlin, Vitalik followed my account after I tweeted about an event that he attended and I said this is the main takeaways from the event and he liked it

and he followed me and after that uh he kept engaging with my tweets and other things and I wrote a tweet saying that I want to use AI to quantify impact and uh after that he got in touch saying hey I've been thinking about how to also work on this. How can we use LLS to like measure the impact of some work and he then laid out a mechanism called deep funding which kind of is what I then got hired at the EF to work on. So the lesson is like try to tweet every day and like you know be active over there.

Um and not just heads down building but also writing and uh getting influence in the ecosystem through that. Um How do I change the next button? Yeah. Um, so maybe it helps with understanding how deep funding is different from protocol guild. Uh, protocol. How many of you have heard of protocol guild? One and two. Okay. So, not many.

Uh so this is an example of protocol guild over here where u it's a list of 190 of the core ethereum developers. So if you're working on consensus clients execution clients you get added to the protocol guild list and all the layer 2 ecosystems give some air drops to protocol guild. So like the famous thing is called the protocol guild pledge where you say that 1% of my tokens will go towards the Ethereum core developers whoever is actually working on the protocol. So they have a very simple model where you uh like if you're like launching your own token you put 1% into

the guild and then the guild gives it to contributors based on how long they've been contributing. So if I've been working for 5 years, I have a very high bait and I'll get a lot of the money. But if I've been working for 6 months, I'll get less money from the guild. So it's 190 people. On average, they get $70,000 every year from Protocol Guild. So that's kind of the like, you know, it's the newest experiment on the block. Um kind of so how is deep funding different from that? So this is where deep funding is where people can give money into an algorithm or a splits contract and that gives it to the repos.

So the biggest difference is protocol guild gives it to the developers like a basic income like if you're working on Ethereum you will get enough money that your that your needs are taken care of but in deep funding we give it to the repos. So the repo like nethermind or get they can take that money and hire new people. So protocol guild gives the money to individuals who are already working. Deep funding gives it to organizations uh who can then hire new people if they want. Uh protocol guild also gives it on the basis of time which I'm not a big fan of. All of us know the meme of like the 10x developer who in 3 months does more work than someone does in 10 years.

Uh so protocol G doesn't have that. If you're working for a long time, you get more money. So uh that's the second difference that we try to measure how important your contributions are. Um so this is kind of some of the big issues. How many of you have applied for a grant before in this room? One, two, three. Okay, only three people. So not a lot of people who are like applying for grants. Um, and I think like the biggest problem with that and the reason many of you probably haven't applied is what I call the only crying babies get fed. So only if you cry very loudly and you actually apply for the grants or you tell people to support your project only then do you

actually get money and a lot of us don't want to do that. I took part in a Gitcoin round and I did not like telling all my friends, hey please give me a dollar, please vote for me. I didn't like that feeling very much. So that's the problem called only crying babies get fed. Um, how many of you have been a juror? How many of you have to give away money? Sit on a committee to give away money. That's one. Two. Okay. So two people have sat on committees before. So something which you realize is that it's not fun to go through 500 projects.

There are like so many projects that are sent to you and you have to look at each project and make a decision. So something which Vitalik said, this is a funny joke. Vitalic said that people at E Global told him not to be a judge anymore. They told him, Vitalic, you're a very bad judge. We don't want you to be a judge. And the reason is because he has to go through 30 projects and that's just too much for him. He's saying if I have to just do three projects, I can do a good job. 30 projects and I just like zone out and I just start giving random numbers after that. So that's the second problem that it's all about quantity over quality. Go through 500 projects and it's not like go through three projects but have a good uh well thought

out reason for um so those are the problems that deep funding is trying to solve. Um so now we come to the main part of it and this is like a call to action. How many of you have worked and can use AI? Well, like rough show of hands, how many of you AI, machine learning, all of that? Like two people. Okay. So, not a lot. Um, but kind of what we're looking at is we have uh 15,000 edges in the Ethereum graph. So, if we look at what makes Ethereum work, it's consensus client, execution client, and chef tooling. So share tooling is things like solidity, execution client

is prism and consensus client is something like get like you know like broadly might have mixed some of that up but like those are the categories. So if we look at those three and all of their dependencies we have about 15,000 edges and 5,000 open-source repos that make up Ethereum. Uh obviously not all 5,000 repos are equal. Some are more equal than others. So we are hosting a machine learning competition for people to build an AI model which predicts how important a repo is. So we might say this repo is worth 20%, this repo is worth 10%, this is 5%. So we open it up to the community to submit their own prediction of how

important a repo is. And so this is kind of how it works that there are 34 seat nodes. So you have to a plus b + c plus d should equal to 1. Figure out how important the repo is or its dependencies. So that's level two. And level three is uh all the dependencies of a repo add to one. So if we look at brave browser, how many of you have used brave browser before? Brave is two people and three people have used brave. So brave is actually a fork of an open-source repo called chromium. So Chromium is like the main repo and Brave is like a wrapping on top of Chrome. So you don't want to give all the money to

Brave. You want to give more money to Chromium because Chromium has done the hard work. So that's kind of the part of level two that how much is the repo itself and how much is the dependency of the repo. Um so this maybe shows it much easier. Uh this is the graph. There are 10,000 edges over here, 10,000 links and all of the important repos are listed over here. Um, so basically in one line if there's one takeaway that I want you to go away with from this talk, it's that uh we are having a competition where the community has to submit models which say how thick these lines should be. So this

is called an unweighted dependency graph. So this is a dependency graph but there's no weights given every line is equal. So the competition which we are hosting where we want participation and $50,000 in prizes from Vitalic is that tell us how thick these lines are. How much weight does each have so let's assume that everybody here in this room like submits their own model. So everyone here submits their own model for how thick these lines should be. And um so now how do we decide who is the winner? How do we decide who gets the $50,000 from the talent for submitting the most accurate model? Um so this is

the jury mechanism. So this is how we decide the winners. There are some there are we have about uh 80 judges uh who do comparisons like this. Which dependency gets more credit for TU's success? Is it Pavl or is it Elyn? And they might say that Elyn deserves 256 more credit than Pabl. And they give a reason over here. So the judge only gives like five of these comparisons or 30 of these comparisons. Very few but very well thought out because we want the judge to give wellought out comparison and not just a lot of comparisons. They can't do

all 15,000 but they can do maybe 500 or thousand totally. Each judge only does like 50 of them. Um so we use their comparisons and their uh judgment to select which AI model is the most aligned with their thinking and that AI model then gets to give money to all of the Ethereum repos. So uh so this is like an example of what the end output of deep funding might look like. This competition will be on until June 10th.

So if any of you want to try this out and get your hands dirty with machine learning, this is like a great challenge to get involved with where um this is just based on GitHub stars. So do not take any of these numbers seriously. But kind of the way to interpret this is if I put money into one public key uh it gets distributed by 15.3% into remix 13.7% into go ethereum. So because the prompt which vital gave was I want to have just one public key. I give money to that public key and it gets distributed to all of the important repos in Ethereum. He does not want to get into the debate of who deserves more

money. Am I giving too much? He does not want to get involved in that. He just wants to give money into one public key and get distributed to all of the repos in Ethereum. So that's kind of the end output of what a model submission would look like. Give weights and we see which submissions correspond to the jewelry mechanism there. Um this is like a really cool graph. I'm reaching the end of this presentation but this is a graph that Vitalik made to show how deep funding would look if we looked at ideological ideas of uh Ethereum ideological uh kind of things. So on the first level over here you have Ethereum.

On the first level you have its influences like the open-source software movement which is 20.5% responsible for Ethereum. You have Bitcoin's monetary policy 19.5%. Um and then you decide how much is Bitcoin's contribution original versus its dependencies. So here Vitalic has said that Bitcoin is one/ird original and 2/3 it belongs to the dependencies like the Austrian economics, Marxist labor theory on value. All of these have helped Bitcoin grow. So kind of if I have to give an example, if I put $100 into this graph, um $20 goes to Bitcoin's monetary

policy, but only oneird of that stays with Bitcoin. So around $7 stays with Bitcoin and uh the rest goes to its dependencies. The rest $13 and the rest $13 get distributed to those dependencies with the weight shown over there. So Austrian economics gets about $4 out of that because it's one/ird responsible and marks gets about 25 cents. So if I put $100 into this graph, 25 cents to Max, uh $4 to Austrian economics, and $7 to Bitcoin's monetary policy. So that's how you interpret this. We are obviously not using this for funding, but this is like an indication of how we can fund

open-source repos with this. Um yeah before I go on to that I'll just talk about why we want this which is companies have always outco competed opensource networks because they integrate the profit center and the cost center so how many of you know what profit centers and cost centers are one okay not many so I'll give you a simple example like if you look at Twitter they have an ad division which earns revenue so any company department which makes money is called a profit center because they're earning profit for the company.

Something like community notes uh community notes uh tells us whether the tweet is true or false but it does not generate revenue for the company directly. So that's called a cost center. So in a company the profit center gives money to the cost center without any control over them. It's not like the ad division can say I don't want to give money to community boards. There's a central department the CEO and the CFO who gets to decide how much money goes to which department. Whereas in open-source networks we don't have that. If there is a profit center it gets to decide how much money goes to the cost center. So things like solidity is a cost center. Things like a are

profit centers, but there's no way for a way to support its cost centers, which is why companies have historically always outco competed open source networks. So the hope with these dependency graphs is that we can actually better integrate profit center and cost center in open source networks. Um this is another challenge that we just finished where you have to build an AI model to predict how much money each project will get in a Gitcoin round. So the Gitcoin round has not begun. People submit AI models saying this project is going to get $100. This project is going to get $500. And then when the round starts, we stop submissions and we see which AI model

best predicted the results of the round. And that's really cool because even if a project does not take part in Bitcoin, we can still guess how much money should go to that project which is what the top AI model which predicted the results. What is the top AI model saying about how much that project should get. So this is another challenge that we just completed. Um this is the deep funding challenge where you have to tell us uh oh sorry this is the past funding challenge. So we just finished this challenge. We got 21 high quality models submitted which looked at things like the like the winning model found that number of open issues is a good

predictor of how much funding a project gets. So they actually like submitted models and they told us if you go to this link over here you can actually see what type of AI models were submitted uh in this challenge. Um and this is the deep funding challenge which is going to be opened till June 10th. We really want good quality submissions. That's what the challenge depends on getting good participation from AI people and people who want to get their hands dirty with AI to actually make submissions. Um so that's open now and uh yeah this is the team that's made it all possible. Um kind of something which is cool is that we work with people who have been thinking about these issues for years.

So open-source observer has been working on GitHub uh repos and how do we represent them? So we use them to generate the graph. Uh pair has been working on the voting interface. So we use them for the jewelry. Uh drips is working on how do we get money to open source repos. So we use them for distribution. Uh pond is the platform for machine learning competitions. So we use them to host it. and Alo Capital has built a lot of different funding mechanisms. So to increase the speed of experiments, we just have to combine the existing uh projects into something which is like something which is new. So rather than building everything from

scratch at the EF, we just work with partners and combine whatever they've built into deep funding. Um and yeah, this is the last graph. If you find any of these ideas cool, take like scan this QR code and join our Telegram group. Thanks. Yeah. Was it understandable my accent or Yeah, it's pretty understandable.

Yes. So talking about uh kind of you are in charge of AI public goods governance uh so um as I as I mentioned like I think coin public goods or say gitcoin portfol operational cost so I understood that like cocoin is really new in so we are seeing like better outcomes But do you think the operational cost is going to fall drastically over maybe next 6 months or maybe the next round? Um, so I think one thing which I do like about Gitcoin is that they've

separated their matching fund budget from their operational cost budget. So their matching fund currently is about $10 million. Uh, it used to be 15 million, it's gone down because they kept it in E, which was actually quite smart. If you look at early projects like Bangkok, Bangkok actually converted their ETH into dollars, but something like GLM at the same time did not convert and went up. So, Bitcoin, it's not like a given that Bitcoin would have held on to all the ETH and it's gone up in value massively. So, they have like a huge matching fund of $10 million which they cannot use for operational costs.

That can only be used for matching and the operational costs basically come from like VC firms. uh there are VCs who invested in Bitcoin and they use that for their operational budget. So I think the bigger issue is not the cost. Uh the bigger issue is that they've not experimented as much as they maybe could have in their um and they're not using that money as effectively as they could. That what that's what I feel is the bigger issue compared to reducing costs. And uh moving on next I think uh bitcoin or say the in the infinite garment as opposed to like sponsor or secret uh public goods. So everybody has their own definition for public goods. Actually I

think public goods are different from only like open source software because open uh open source software is actually one of the sector of coin. It's not just like public good doesn't mean open source. So like could you give us a precise definition maybe your own version of what actually public good is possible? Yeah. Uh I actually have a very mathematical definition of it which is public goods is equal to value creation minus value capture. Um so if you as a project have created a lot of value like say Facebook has created value in the

world but they've also captured all that value they've managed to get enough revenue to actually be equal to the value they've created so they are not a public good but something like Linux has created a lot of value for the world but they have not captured the value they've created so Linux is a public good so public goods funding should try to bridge the gap gap between value creation and value capture. Okay, so that actually reminds me of the uh revenue evilness, right? That that stop always mention. So I guess as a following up question I think like right now not only

Gitcoin not only EF but also like the OP breath house or other foundations are waiting or trying to value influence as much as or what's the impact and uh also as a business this is a research I know house so I strongly recommend the fellows to use paradigm excise so maybe not and not don't go that but like BIM Foundry uh truffle and balcony uh wallet kit like those are super good tax sites but also like they maybe they are open source softwares and they are not taking like revenue or like giving value back

from those open source software but what they actually are monetizing is their impact and also influence like they are amplifying what they actually did and I sometimes I think they over exagregate what they have done like they are they're claiming something they have like other dependencies that you mentioned so how do you think that that thing can be changed or say um how do you think foundation or everybody can help a lot on that yes so I think I mean I call this problem the halo effect where they are supposed to have like a good reputation and because of that people support them

in bitcoin rounds like people give them grants and because they have this like reputation and you don't actually dig into the details. So I'll give you one story from the funding which is a lot of the judges gave good marks to web3js. Um so web 3js is like a version of beam but uh and it's very old but no one uses web3js anymore. So deep funding was going on. All the judges were giving good marks to web3js. But the EF was doing its own investigation and the and the Ethereum Foundation found that no one is using web 3JS anymore. So because of that they discontinued the repo. They even wrote a tweet web 3GS that we are

retiring. And I'm looking at the spreadsheet results of the judges and I'm seeing like web3gs is so useful inside the ecosystem. And then I'm looking at the report from the EF which is saying no one is using web3 GS anymore. So I'm like how do I combine both of these? Um so that's the halo effect in action right that like it has a good reputation so people are giving it good marks without actually digging into how many people are still using it. Um so that's kind of the big issue that is not solved yet. Yeah. And uh yeah so so mult like um also like like the following question

is like um how is the that a public goods actually gets the impact because I like for my personality I go through like bankless d and d and like Chinese as well. So one of the main issue is that like oh this is like a funny joke I made on my personal is that you you realize that we have the second foundation kind of uh organization and I just made a like a joke on Twitter is that okay I'm gonna just start write a Chinese newsletter for every single week like weekly newsletter and I'm get gonna get granted. So I mean it ironically

because like over the years we've been valuing like an education uh you know sliders and stuff like that but like very recently we lost a briefly and that actually affects me pretty much. So how do you think like on that general? Um yeah it's like the EF is sometimes a like very delicate position where we don't want to be seen as king makers that we are choosing this project and then all the competitors of that project is like oh EF has chosen this project what do we do now so that's why historically the EF has funded a lot of

events which has actually led to too many events in web three you know that's like the easy way to get funding from the EF historically um and and I see a lot of that changing now like now the EF is being more opinionated because we are in a different regulatory environment EF played it very smart and we got the EPF and we got a lot of other things which would have happened had we been more opinionated um so I mean specifically on the thing of like weekly uh news I mean there's a like it's kind of like complicated one because EF had reached out But then they didn't respond. So like it's a lot of murky stuff and I haven't dug fully into it. But uh it's like hard because it's

like in 2017 it was a very different grants environment. Like whoever was applying for grants was usually of a decent caliber but after 2017 it brought in a lot of people into the ecosystem and now kind of it's become a very adversarial game where like people want to try getting as much as they can from the institutions. the institution has to be very diligent and careful about who they give grants to. So in general, I think the grants environment has just become a lot more adversary and I'm hopeful that like things like AI kind of prevent the need for even applying for a grant. So you're contributing to the relevant GitHub repos and we are able to identify that and just give you money

directly without you even applying. So hopefully it can become less adversarial through that. And I think this might be my last question and forward to the stage. So also like yesterday talking to the is like he values privacy as a like very super like super public good and like most of people here what's in ZK they are in ZK. if privacy is an issue and we all deal with them at the most and how do you see the like regulation as you mentioned for the climate change a little bit because actually I I was into gaming and I joined my this

company at immediately after uh catches me to be innocent. This is the exact moment I joined my company and um so and also I know the the story behind Noctu and also Umbra that's such a stuff going in but how do we kind of like report or call to action for people to value privacy once again since the regulation climate is kind of like yeah I think like privacy has always suffered from like a mass market problem where the average consumer does not care about privacy. That's like the bad news.

But the good news is that all the top people care about privacy. So people at the top levels of society care deeply about privacy because they are under more scrutiny. But the average person does not care about privacy. So you kind of have this weird thing where people like your tools will not get mass market adoption like signal has much less adoption than WhatsApp but signal is used by Donald Trump and all so like the top people are not using it right. Um so I think privacy should generally kind of command a premium like if you have a privacy based product like charge high merchants for it because this is the

famous right so okay so yeah that that I cannot remember the exact words but he said yesterday but forget about it so basically it's like let the bigger ones has the transparency and the weaker ones have the privacy. So, but this is actually kind of against what you just mentioned what I think it's a PM like charging a high margin say a premium amount to the top level person cares about privacy. So is that

any better ways for us to like work around it? Yeah, I think it it's like depends on how pragmatic you want to be. Like the pragmatic rule like if you look at something like rail gun, rail gun is like a great privacy protocol but it's also very expensive and that's how you start off at least like when you're just launching like you are going to be expensive and you cater to those who care about privacy who basically have money. People who have money care about privacy. Those who don't have money don't care about privacy.

Simplification, but that's how it So maybe we can have a subsidy model for I think we are open to this right now to add. Okay. Yeah. So I talked about like the halo effect. I talk about one other issue in giving money which is actually very like difficult to solve. It's um what I call like the pass out value. So it basically is like if you if you have a car that you want to sell, you get more money if you sell the car with each of the parts. So I take out the engine, I sell the engine, I take out the tires, I take out the and I sell those. That gets you more money than then if you sold the entire car,

which doesn't make sense, right? Like technically you should get like the same. And we see the same phenomena in public good projects and funding where if something like protocol guild applies for money, it gets less money. But if like protocol guild divides into 30 different projects, each of which applies, they actually get more money. So whatever. Okay. So, so like pantas whatever whatever whatever and the sigma of panther's eye actually equals more than itself. Yeah. So that's kind of because all of the distributions are peanut butter spread. I think it's like this very like kind of egalitarian

element that oh yeah we should be funding all projects and all whereas kind of the market doesn't work that way. market believes in power law distribution that like 5% of your projects are going to deliver 80% of your value. So I think we need to move away from this thinking of like being equal and including all and just like identifying your winners and just like putting a lot of money behind it which is very opposed to whatever the eos questions. So like I'm wondering what happens when there's a difference between how much impact the project has and how much

money the project needs because it can be a very very small project where by one person it's like used by everyone superactful but it doesn't really need that much money very in such cases like how do you allocate resources? Yeah, that's a great question and maybe we can actually make it a real example which is beam. Beam basically was started by a Canadian developer who single-handedly overtook web3s but it's like literally one or two guys who are making so they don't actually require a lot of money. Um so the instructions we gave the jury was uh specificity like is this repo specific to Ethereum or does

it used by everyone? Second is additionality. How well can they use the money which we give them? And the third is replaceability which is uh if this report did not exist, how difficult would your life have been? So specificity, additionality and replaceability is like the larger framework and uh something which Vitalik said is that if you make something which half of Ethereum depends on, you deserve $5 million. it doesn't matter whether you spent only $50,000 or 100,000 to make it. Um, so I guess that's where I would come from that if you are like we should move away from cost plus pricing.

So that's like the biggest issue with all contracts in the government. So like Boeing and others would say that okay to get someone to the moon this is our cost and this is our profit on top of that and then what they end up doing is they end up increasing their costs and the cost of that is boned by the taxpayers or is born by us whereas Elon Musk basically changed the cost plus pricing to be like this is like a flat rate and we're going to try making it as cheap as possible so that our margins improve which was a better system. So I think we need to move away from cost plus pricing and try to just look at and I actually fell into the same trap like you know like they don't need much money and like it's not being used well and then

Vitalic kind of corrected my thinking here by saying that look if half of Ethereum depends on it it's okay to give them $5 million and we shouldn't try looking at the cost as a variable at all. Yeah. Something I would add here is that we give them good person score if they're kind of updated in that. Yeah. Yeah. And uh and also I think in a traditional business there's something called good that is like your reputation and uh for giving like maybe that 5 million I actually encourage the atmosphere of our space as the garden.

So, so yes, any other question? Okay, cool. Just like talking about like the infinite garden like something I advocated for like internally. So we've always followed the principle of subtraction that by not involving ourselves like to give an example ENS um we made the decision at the EF to not make uh kneeling an internal part of the protocol and we actually let ENS do that and because of that we now have a strong pillar like ENS has as much money as the EI which is like amazing right and they have like revenue coming in and it's like great that they have that so subtraction has helped a lot over there but subtraction was useful when the

infinite garden did not have plants. So if we were doing everything there's no space for any tree to grow but I think at the juncture we are in now the infinite garden has a lot of saplings we have a lot of baby plants and we need to somehow make sure that all of these plants work together so that the whole is greater than the sum of the parts. So in philosophy we call this uh synthesis. Synthesis is the whole greater than the sum of the parts. An example is a car where the wheel, the engine, the steering wheel, all of that is not very useful individually, but when you put them together, the car gets locomotion and it moves. So the role of the EF like

what I've been saying in the infinite garden the role of the EF is to make sure that the whole all of the different elements in the garden are greater than each of the individual elements. That's kind of what I have allocated for at least. Uh we already Okay, thank you so much.