S7 Ep. 2 - Ai Agents & DAOs

Transcript

welcome to the green pill. network podcast if you're just joining us we're building a coordination a network Society of thousands of hackers dreamers and doers using crypto to bring positive some digital systems to the world this podcast features the people who are doing it and we publish season of five to 10 episodes every couple months this season is about how to Dow I've got a new book coming out with my co-author punar how to dow. XYZ and we are talking about how to Dow so what is the bleeding edge of dows what is the coolest stuff that is happening out there that is the sum of the how toow season of the green pill podcast if you want to learn more about the green pill podcast you can visit our website at Green pill. Network where you can download the green pill book for free join the Discord and

become a member of your local green pill. Network chapter today's episode of the green pill network podcast how to di season is with Shaw and jyn who work at Ai 16z and Eliza no relation to a6z uh this is AI 16z they are building AI agents and we are talking about the intersection of AI agents Dows and onchain capital allocation in this episode so if you've been following any of the AI agent stuff that's been going on in crypto the last couple months you know that there is a really interesting boom in which we've had llms being turned into agents which are basically

stateful uh llms that can do things like order pizzas or hold crypto or be your friend on Twitter and so there's this sort of new intersection of AIS and crypto and Dows that's going on and in this episode we were talking to the people who are leading the revolution Eliza is one of the top open source projects for doing AI agent work in the ecosystem you can kind of think of them of like uh what Linux was for operating systems uh distributed movement of creating an operating system they are for AI agents the AI 16z is a dow of over 100 open source software developers who are pushing AI agents forward Shaw is the original developer between behind Ai 16z and JY is a community manager and

we are here to talk about the evolving design space of dows Capital allocation AI regeneration and I found this to be a really exciting episode and especially exciting one to follow up on my vitalic episode which was just you know full of so many brain droppings about the future of web 3 public goods funding and funding in general and I think uh if we were to pension zoom on that design space a little bit more web three public goods funding and funding and add a little sprinkle of AI That's What You Get with this episode with Shaw and jyn so we're talking about using AIS to give simpler access to information of for people in D to revolutionize defi to make it easier to access we talk about uh the intersection of AI and DS Ai and

onchain capital allocation we talk about coordination challenges that AI agents can help us solve what can you do to prepare yourself and your Dow for the agentic internet era um then we talk a little bit about the intersection of AIS with regen and we talk about cathedral in the bizar this metaphor for how should you build software in a distributed network-based way so um JY and Shaw are some of my favorite guests that we've done in the last couple months just because they're a so smart but also so genuine and they're genuinely just interested in innovating at the frontier of these design spaces and that's why I think that this is a special episode of the green pill podcast and I'm really happy to be doing it during the howed Dal season because I

think the future of dows is not going to just be multiplayer crypto but it's going to be multiplayer Ai and so the frontier is coming and this episode is designed to help you prepare for it I hope you enjoyed this episode with AI 16z Shaw and jyn enjoy coordination hello Shaw hello jyn welcome to the show can you start off with some quick intros sha you go first sure thanks for having me my name is Shaw I am a developer and creator of the Eliza framework uh and pushed the button on a decentralized Thomas organization called ai6 Z from San Francisco California thanks y JY you're up Hey Y'all um Jen here I've been working in crypto open source AI VR AR for long

time probably around 10 years and I share a passion for finding ways to make open source sustainable and how we can grow this movement Thanks Jen um so the intent of our episode is uh you all are working on AI 16z um and are part of this new AI agent meta within the crypto ecosystem and what I really want to do is take my audience of lots of developers and figure out how do we route them to the most meaningful work um in AI in regen in onchain capital allocation I think that's kind of like the vend diagram that we're looking at here but maybe we could just start with start with why um why are AI agents so

exciting to you Shaw and um you know what is the coolest stuff that you're seeing in and around the AI 16z and Eliza ecosystem so I have a lot of like future sci-fi Ambitions and love for the thing that it will be and autonomous agents in our world and robots walking around and all this stuff but I think the really immediate thing that we're seeing is the is like the reshaping of the internet and this isn't something that was obvious to me when we started but now it's like very plain as day that um what where internet used to be like websites right um now it's becoming very social media oriented most of us get most of our news and communicate with each other and interact through social media and pretty much every website and

app is competing with social media for attention and so what's really changing is that these agents are allowing these apps to come onto social media and so you see see these apps like like agents that represent not just characters but also have the capability of those apps and so where you might go and say okay I want this cool AI headshot app you know and go to AI headshot do AI or whatever I don't know whatever it is um you you can actually just get that experience now directly on social media and I think that this is really for developers what's changing is friction you know we all used to use like nextjs and and build a website and all this you know these kind of web apps and and the problem is that we competing for attention against the places where we're

advertising and so now a lot of people in our ecosystem are using our Tech to build apps that go directly onto social media and so these agents are kind of the next application layer of the Social Web and when people say I think like web four I don't know I don't know I keep just jumping over the webs but but generally there is this idea that the internet itself is changing and that things come to you now not just on Twitter but also on your phone on Discord and the next application layer is going to be really personal and really right there where you already are as opposed to making you go somewhere so I I have there's a lot of apps that I think are really exciting obviously the one that came out this week was agent pizza was you could order a Domino's

Pizza um from an agent which is just like a silly thing but if you replace Domino's with uh kin Le or uh booking your flight or anything you could imagine um this becomes it becomes very obvious that the back end is basically the same it's just calling API and doing what websites have always done but the front end is very different and now it's a much more personal experience for you to go to to interact to go shopping to to get anything you want and we're seeing this kind of bizarre emerge you know like this open Crazy Market where people are just building stuff for each other and and social media becomes like the Town Square of that so I'm really excited by that Trend yeah I think at this point um would love to just uh bring the audience along a little bit

and Define ai agent so as I understand it's kind of like an llm but it's stateful has persistent memory it can be bridged to social media it can interact with other AI agents and learn from them and from social media and it also can be shimmed to have like a crypto wallet or to order Domino's Pizzas is is that true reductionist is is that how you would define an AI agent uh I think most of the audience is familiar with llms but maybe not AI agents yeah I think that there's this implicit idea of autonomy that it's not just an llm but that the agent can go do things on its own and make decisions and and and I think that that's all very important and coming but it's this much simpler thing of of bridging people to things they want through some kind of

mechanism that makes it dramatically simpler I think the thing I'm most excited by is is actually how this will revolutionized defi where you have all these really interesting defi Primitives like liquidity pooling and staking and yield farming and all this stuff but most people I know even people in web 3 with wallets don't have any idea how to interact with any of it they don't don't know that with five button clicks you could just be passively earning income and you know you go to medora's website or radium and it's just like like without a 30 minute tutorial you just don't even know what button to click um and so I think where where the llm like where agents really work for us now is not the full autonomy doing stuff quite yet although that will happen but it's in connecting people to these services

like hey I need I have money can I just put it into an LP I have no idea what that or how do I make money and then the agent can directly respond to you give you the information that you need right there and then and then like route that payment and make sure that action actually happens in the way that you want Jenice how you had your hand raised you something to add yeah I kind of want to Circle back to the friction reduction point because um I do a lot of volunteering at various ORS that create a lot of meeting notes and data and it's just like a huge backlog that nobody ever reads and um also when it comes to coordination there's kind of like the crazy bumblebee sort of glue people that pollinate information across different

group chats with whatever is kind of useful you know just like to relay what you know something from a meeting or one or the other or like just writing docss right and then you're like um working on something uh and someone has a question that's in the docs whereas I think AI agents like going back to reducing friction they kind of bring all this information to you in real time time 247 you know and so I just see it as like a force to really sort of um help bring these disperate uh data pools and work groups um to collaborate much faster I we're both very passionate

about Dows um I think that the D and Dows the D centralization has been like very very well solved um we have all this great tooling but the automation a has just kind of lagged behind and so like a lot of the problems that we see in Dows I think we we see it like that agents can solve that especially around routing communication dealing with bottlenecks just generally making sure that information gets to the right people at the right time I think that's such a it's such a simple point A to point B problem but it tends to be the bottlenecks we see in these kinds of like very low hierarchy orgs that's actually a great segue into my next question which is um you know JY has showed me your weekly what have you shipped threads where everyone in the a6z community shows up in demos cooltech

and um it's just amazing to me how this project is only a few months old and you're already got hundreds of contributors and you're shipping so fast and so I guess my next question is like how do you doubt how do you move so fast and so effectively in in in this way is it like AI using AI to build AI is it the shared context uh what's your philosophy on how to Dow I actually think it's going to start moving faster as we start to get all of these tools built um I think what we've done is I think that it can be very hard in web 3 to create something that's bigger than your like cult you know and and cult is a word that's used a lot and I think what we're trying to do is build something more like a movement where

where there it's it's much more open and enables a lot more people to be leaders I think what's key here is that we we have an open source framework that anybody can use and we're really encouraging everyone to build what they think is important and we're trying to set the stage for that and create examples of what we think where we think this is going even just saying like you know the agent Pizza thing was really just to try to push things in the direction of saying hey you can build apps this way and I think the secret to our velocity is that like no individual can move faster than they can but many many people being enabled to do something allows it to move very fast so there's obviously like a lot of coordination complication there like coordination is definitely our biggest issue always but at the same time

because we're this like large movement we it's easy for us to not take credit but say like look at all this stuff that's going on and there's a great horizontal kind of movement of of everyone running as fast as they can in more or less the same direction um and then we just need some alignment to you know say but but generally it's pretty self-aligning because we have this open source kind of foundation and also since we're not we're not an L1 infrastructure thing which means that we we kind of go across a lot of different infrastructure this is actually a tech that's very useful to pretty much everybody and so I think as a matter of strategy we were talking about the cathedral and the bizaar and and that's something we thought a lot about and I think we're very much the maximal bizaar play we

don't want we we've had opportunities to build things ourselves and say like we could build a platform but there's like 10 platforms getting built by people in our ecosystem already and so we're looking much more at how do we invest how do we support as opposed to like taking their things and trying to compete with them and I think this this just maximum like if you just go to The Logical conclusion of this it is the fastest growing thing you could possibly make and I think growth and acceleration to meet like the challenges of our time is really like what why we're here you know we can maybe make more money by doing it a different way but I think it's very important that we try to meet the moment and everything else is Downstream so it feels like you guys have a philosophy of you're building a

movement you're not building a cult um you're also really doing Eric Rayman had this essay in the 9s about the cathedral versus the bizaar which was basically a metaphor for Linux versus Microsoft Microsoft being the cathedral and Linux being the bizar so you're all in on the the bizarre building a movement and then information sharing and sort of coordination across the network towards that that that shared North Star is that too reductionist is there anything I missed there in that summary that's definitely it and I also think dog fooding our own Tech like we have the problems we're trying to solve so very conveniently we're working on on the autonomous tools to solve our autonomous organization problems yeah that with that part I think is uh we're just

really starting to put all those pieces together for the next stage of coordination and scaling because there's a lot happening in Discord there's a lot happening on GitHub there's a lot happening on X and you know there's now many more channels and in terms of dog fooding you know I think AI agents are the solution to kind of rebundle and interface with all that information to get kind of what's most relevant um and get answers quick as well and uh stay up to date I know people that don't you know it's just such a fast moving space right it's only like a few months old and so I think this is one of the best uh Parts about uh in terms of strengths

of llms is they're really good at summarization and we're just kind of using that all across the uh the board for what certain people are working on and who's doing what like all those uh weekly Twi spaces that're always recorded transcribed summarized and uploaded to the docs and then later you can like hit a website or talk to an AI agent in Discord about who is doing what or if you have a question with something that maybe someone else is working on it could uh you know automatically sort of uh either help on a spot or at least maybe know who to look for you know who to ask so it feels like llms you're you're not only building llm agent to

Tech but you're also using llms to Route information across your organization and I think that's a big opportunity for Dows I I'm speaking for gcoin itself like I'm the founder and I can't even keep track of all the stuff that's going on and I know that everyday contributors have found that they don't have that information either and then the other thing that I'll also say is that I just had vitalic um on on the green pill podcast and one of the things he talked about is that there's a ton of stuff going on in the ecosystem but there's no information sharing layer so that everyone gets to learn about everything it's only what bubbles to the top of crypto Twitter which tends to be just like you know dog divisive debates but like how do you actually route the information about the ecosyst the ethereum ecosystem is probably like you

know 10 or 100x bigger than either getcoin or AI 16z but you know I I feel like that's an important problem to solve in the Dow age is is that summary and routing of information that could be a product that you could productize and and sell out to the other ecosystems are are you guys considering that at all oh yeah I think that ultimately what we're doing is a lot of Dow tooling and you could imagine let's say we have like 10 work groups and each one has an agent watching and then just going and and communicating with the other work groups or the Agents of the other work groups like hey this is what we're working on is there anything relevant that you're working on and and basically trying to synthesize because like a lot of it is that teams are not aware of what the

other teams are working on how they could be like how they could collaborate and synthesize more and you get a lot of stying just just because there's so much information it's just so hard to to know everything that's I find out half the stuff that's happening in i6z off Twitter because it's just other teams like doing crazy stuff and actually we should be leaning into this we should allow more uncoordinated you know work that's aligned in the same direction um and I really think like what jyn is doing on on several fronts is like probably the most important work inside of our org um and we are the right or to test this stuff out we have these problems um one of them is everything that we write is getting summarized and then fed into a chat bot so you could just be like what happened this week

what are people working on how do I start Where do I start contributing and and really trying to address that with a natural language interface yeah and then I'll let I'll let JY talk about some of the stuff he's working on because I think it's really exciting especially around retroactive funding yeah and I I gotta say like um it's also super exciting that a lot of organizations kind of are using the same Stacks like I saw on ethereum C devs use GitHub for Gathering all the meeting notes and zoom calls which I did a lot of work summarizing the transcripts from Zoom calls into llms and um you know Mark down as a common format for documents and uh so the stuff that we're using in ai6 z a lot of orgs use just especially

if they're remote and online and so it's so reusable my dream is that whatever we're kind of doing to help us scale and coordinate 10x faster is applicable to anyone you know working on a protocol or open source projects you know something that they can just fork and help tame all that chaos yeah yeah obviously we could productize and sell this but I think what we really want to do is get other people to help us build this and and so I think I think we're going to keep doing what we're doing which needs to be working which is to open source and enable a space for other people to come collaborate invite some of the other large Dows in and start solving their problems and and contributing back to the thing that you know all this llm

stuff that we're doing is very much just the first version and there's going to be 50 or 100 Years of research and development continuing on from that that first point and it will be very developed by the end but we have to get there and it's going to require a lot of people to build that together um the thing that I'm most excited about is I mean we can get into this which is kind of the regen stuff but like how do you make sure that the capital moves to the people who are actually building the stuff um and I think that a lot of dows are facing this problem of having like a like a lot of holders who kind of just continue to reward the holding as as opposed to the continued building and value creation and so I I think that all

the tooling that we're building is is like trying to solve these fundamental problems we all have um and and I don't see any reason why it can't be an open source thing we all build together um if people want hosted versions of of this stuff like we'll probably provide some of that but we also have an ecosystem of people who are providing that so you know we're much more interested in investing in them and and letting these kind of solutions fall into place yeah so I mean let's let's get into that um you know obviously VI L my expertise getcoins expertise is on chain Capital allocation looking at things like quadratic funding retr funding deep funding we just announced last week and um I think that that's a very important Frontier rewarding the builders and the people who are doing things that are

good and then yall are into AI uh which it seems like is really good at collapsing context and giving it to the best to people coordinating just making it simpler to deal with things and I guess I'm just wondering what's if we were to draw diagram between Ai and onchain capital allocation what are the most exciting things within that design space and you know I would love to just start with what you all are doing at AI 16z like are how are you rewarding the hundreds of builders that are working on Eliza in those open source repos or how do you plan to um and is it going to be is it going to be AI augmented I'll start first and then I'll introduce what uh I think Jen has a lot to say about this um I think that like when we look at these ideas retroactive funding any

of the kind of like how do we vote to decide who to give the value to quadratic funding and all that stuff um all of the technical side of that makes a lot of sense but it tends to the struggles and complications come from the human factor just getting people to vote or the people who are actually overseeing this just being overwhelmed by how many contributors there are and how many different factors there are and we don't even necessarily know yet what factors really matter when when deciding who to give value to like let's say we're going to we we want to fund development well are we funding like lines of code or poll requests merged in or what exactly are we funding like what are metrics and I think that those are like weighted metrics that we'll

continue to evolve and research on and like actually probably eventually like this will hit Academia we'll run you know simulations and really have a good idea on but right now we're kind of just like seeing what works and and testing in PR a bit and I think that that means it requires a lot of human oversight to make sure it gets right so that we don't run into like AI just doing bad stuff or being games but it's able to reduce so much of the like overhead the cognitive overhead and I think of all the things we're doing the thing that I'm most excited about is the thing that jyn is leading which is the um continuous retroactive funding so we use AI to look at all of our social channels all of Discord all of GitHub and what people are actually building and contributing

and then award that like assign points to them you based on what we estimate those contributions are and prepare everything so that we can airdrop regularly and we can we can actually move resources into the hands of the builders um but there's obviously a lot of issues how who do you reward and who is in charge of that process um and we can put people in charge but what we've seen is that people often when it's you know a hat that somebody wears it can be very overwhelming and so I think that AI really helps to reduce like to collapse that context into something that a single human can manage and can review and sign off on and I think it will make like a lot of the Retro attrative funding strategies much more viable at scale you know like it's just hard to

fund 10,000 people at the same time unless you have some kinds of mechanisms that are enabling that um so I think that's really important and I think creating a culture for us the most important thing is to create this culture of like if you come and build you will be taken care of like that's all I want I just want enough and I think most Builders I know just want enough to build the stuff they want to build and if you could guarantee that um they would definitely just show up and then we would reduce the coordination cost you know I think we we we've talked about bounties and I love the idea of bounties but there's a huge overhead of how a human having to figure out how much it's worth what all the specs are what the acceptance criteria are and I'd love to just see people come and build

what they think is important and when that gets merged in we're like yep I didn't even know I would never even thought to Bounty that thing but that clearly is something we need and we need that culture where people can participate and contribute what's best for them and I think that will really help us with that bizarre thing with just letting anyone do it and not saying oh yeah Microsoft determined this is valuable or whatever that is yeah thanks Char that's beautiful so what I heard there was continuous retroactive funding um retr retr funding I think our audience is well familiar with but the AI agents I think make it much easier to do more continuous uh retr funding on smaller time scales which is important because um right now you have to kind of take on faith in some of these

ecosystems if I build something then in several years I'll get a retroactive reward for it and I'll be taken care of but the stronger that feedback loop of doing it faster I think matters in creating the it you have to take on faith that if I come and build you'll be taken care of did I play that back uh correctly and then I want to go to jyn hear what JY has to say how do we reduce that Loop to like two weeks or a month so you feel like oh this is like payroll you know this is like what I do and I know it's going to happen because I saw it happening for the last six months and and all my friends are making money off of this I should probably do it too and then we can take care of our community

without having to have a huge apparatus to to manage all that overhead yeah no HR no payroll it's all just AI augmented continuous retroactive funding yeah all right Jen I want to hear what you've been cooking yeah um I plus one on like reducing overhead you know I've done lots of bounty and Grant stuff and there's always like that upfront cost to describing stuff um and what you need and retroactive is really good at you know awarding people of like creative value but measuring like impact is always a challenge and continuous thing and so I feel like AI augments uh the ability for people who are doing that measurement by 10x you know at scale and it's not just

uh on GitHub you know like I'm inspired by tools like Source cred but there's a lot of people that are contributing value just answering questions in Discord and you know that's a poters channel for example is one of the most active channels we have and it's just filled with people who are asking questions and answering and engaging you know and that should also somehow be acknowledged in in this uh feedback loop so I've got some tools and it's open source now if you look on the a6z org for chat summarizer and what it is it's using you can use it with local models as well so you're not sending your chat logs to the cloud you can process this all locally

is a way to just kind of like summarize any kind of Discord channel daily for an overview um an FAQ which really helps to kind of feed into docs and pain points that people are having um which turns into issues uh who helped who and which is very useful information for the uh retroactive PVF and as well as uh action items which um I have some subtasks for documentation needs uh action items and Technical tasks so running that on a loop and it's a very active Discord and we can keep tweaking these prompts such

that you know like they can filter information maybe to what's most relevant towards another group and I think this is where we kind of the future of this is swarms where you have like this data pool and the information you know it's a lot and maybe not everyone needs to know everything and right now I have kind of this one siiz fits-all filter but maybe you can um filter it to whatever is most relevant for you or your group's needs the other thing I think that's really cool is the uh the profiles um it's another project that we have on our GitHub um and it's kind of like a leaderboard of contributions and I don't think we want to create things that are too like you know I'm number one but you know but but there is something to that like social

capital is definitely value and being able to to go and say like well who are you and be like no look I like I love meritocracy I love that like your social profile is all of the things that you've done and so and this is kind of like uh if you look at that you can kind of see how who who's going to get funding for this round um and if you think about like what payroll is like a company is basically retroactive funding right like okay cool you did work for two weeks here's two weeks of day like how can we create that with minimal overhead and maximal ability to let new talented people come in and do what they love doing one of the things that I think is interesting about all of this is is how you could automate and take away the

overhead um but I I I really do think that as someone who studied how you fund things for years and and by the way like gin's original Mission was around open source contribution there's a real question of how do you value each contribution and whoever's writing the AI that values each contribution has a lot of power and so have you considered and this is one of the design criteria that goes into deep funding the experiment that we've been working on with vitalic but basically like in that experiment design it alic realized that and so he created a Marketplace of AI agents where basically you have human spot checks that go in and say this AI agent's full this one's capture this one's doing really good this one's doing really good and then you basically create a Marketplace of AI agents so that anyone that wants to create a

ranking algorithm can go in there and not only do you prevent it from being captured but you also get more of a diversity of of realities around what what has value um what's your reaction to sort of that design pattern and is that on the road map for the ai6 the open source continuous retr funding experiment I absolutely love that I think that's totally the right direction I think um you could imagine kind of like an agent reputation system and and the marketplace is that you don't necessarily vote on individual retroactive funding or or you know voting quadratic funing but you're voting on the agents that are voting and that representational structure is how most governments actually work um because there's just there's just so many things to make decisions on and I

think that if we could have a model of like well I really like these agents and and generally our Dow has voted for these ones that we agree are you know and then some kind of average out the top ones or something I definitely think that's the that's the way to go for sure um and I would absolutely love to be involved in that that sounds great so want to build a little bit of a a foundation of knowledge here in the listener's mind um at the base we've got llms can digest information and intelligently talk about it and then on top of that we have ai agents and AI agents are stateful they can interact on on social media you might not even know you're interacting with an AI agent they they can be shimmed in to have crypto wallets to be on social and then on top

of okay so lm's AI agents and then you have ai agent swarms which is basically multiple AI agents sharing context with each other and state with each other and in some cases they can be in these things called Infinite Back rooms where they're sharing State and information with each other 247 and they'll just discover parts of a latent design space that we've never ever even thought about before because we're slow meat bags that are only working six hours a day and then so llms AI agents AI agent swarms in back rooms and then the next level up feels like allowing them to do continuous retro funding where they're all funding each other and they're all funding onchain Capital allocation at scale within a dow and so like that's the stack from where llms are now to where I think that this is going based

off of our conversation and I just wanted to make space for you to react to that to to that stack of um from llms all the way up to continuous retr funding I think that um like obviously llns aren't great at certain things you might not want to give them a bunch of stock tickers and be like predict the stock market tomorrow because it's just gonna kind of come up with a thing but what they're really good at is turning unstructured data human language all the things we do into structured data into like uh 35 points for example and and so I think that a lot of the back end of this is like doesn't have to all be llms we use other deep learning as well and we also use like a lot of traditional algorithms like all of us collectively but it's that interface layer between

the humans and all these powerful tools and perimeters we have that I think is really critical um and and yeah I think um I think humans can also be agents but the reality is that these AI agents can just just move so fast they can process so much data um that I think we're going to have to move into this world of of human review is definitely the next thing and and don't take the humans out because we're really good at decision making and making hard moral choices and and things you know and holding a lot of context but just make it easier just make the whole thing a lot easier for humans to to do what we're good at this is one of the things that the block science guys when I worked with them on um some getcoin design stuff really drilled into me was this idea of a human

in the loop so like have human checks on the AI decision- making to make sure that it's working well uh jyn just put into the chat autonomous at the center humans at the edges uh curious if you want to Riff on that at all JY yeah I think it goes back to you know like um the original vision of the Dow like uh when balic um defined it 10 years ago he even mentioned autonomous agents and I I can just see like them doing a lot of sort of the work to power the ship but then as humans sort of doing the guiding and long-term directional influencing of it um I'm curious what else within the design space of onchain capital allocation and

AI agent allocating Capital to humans or or to Dows you're really excited about is there anything else that you're seeing that you're seeing AI agents do with crypto wallets or in the realm of of of funding each other that you think is cool well yeah I think the the most interesting thing will be that's going to come is when agents can identify their own weaknesses or needs and then hire humans to come fix them maybe eventually it'll be agents hiring agents but actually like what we want is more contributors we want more people to come build this with us and so I think um a lot of it is it's hard you know I while I I said kind of like you know the problems of bounties are are determining all these things you know having to upfront determine well what is it worth and who gets what and what are the specs

that all could also be created by the agent for human review and I think that where bounties really thit into our ecosystem is in that they they allow us to attract new Talent right they give some sort of way for somebody to prove themselves and immediately establish that they're capable and then we can just you know that's a great pipeline for us so I definitely think what's cool is like we just saw the first agent make a GitHub issue on our like a on our thing and we'll probably see them making poll requests but I really want to see like bountied GitHub issues where they're paying in their own token for people to come and work on them and this becomes this kind of self- attracting Force we're like hey I'm an AI agent I kind of I kind of suck but I got a bunch

of money so like if you come make me great you know we're going to win together um and I think that's really interesting kind of thing where like the agent is like a it's it's a hypers it's an idea of what it will become and you just have to come and participate and then it becomes that thing on that note like something that I've been working on is this feedback loop between analyzing for example the pain points that people are describing on GitHub and finding ways to kind of turn that into new issues and see if anyone's working on it but it is this kind of feedback loop where they can help us write those bounties faster and post it and you know I think documentation

in terms of like there was a survey that Intel did for the top open source challenges faced by open source like U maintainers and the number one was maintainer burnout the number two is documentation and onboarding number three maintaining sustainability and then number four building Community engagement and so like when it comes to one of the strengths of llms you know like summarization and just like structuring data that's really I think great for documentation and I feel like if we can close this Loop where like new changes that get merged in we can like kind of flag if any docs need to be updated we can open up more bandwidth

for developers and then also agents can interface with h the community and the answering FAQ which um also kind of like frees up more time and helps uh sort of engage uh the community as well and so like change logs can also come to life we see a lot of influencer Agents as well like kind of promoting projects and uh that you know not every engineer programers is also like a marketing type person maybe agents can assist with that but I think that you know documentation is one of the keys you know to finding pain points uh describing bounties that could be fulfilled by humans or at least spot checked by humans you know as the ones

to kind of merge code in and then um yeah just find ways to uh really speed up and make the whole pipeline for open source software to be more sustainable on that you know we often talk about self-documenting code it's like a very common thing that coders talk about um because code documentation is really hard it's like a lot of work it's as much work is writing the actual code a lot of times and so I think what we're seeing now is self-documenting code I mean it actually you know agentically documents itself um and another thing we're really interested in is like look the world that doesn't all

speak English and so we have you know we another thing we're working on is like can we automatically translate all this documentation into everyone's native language and llms are great at this stuff you know it's that's just one button push you know push it to get and it automatically kicks off all of these things so I think that um code is now actually also just easier to read easier to get into um and and so many open source projects just have like no documentation and so I think that that really really is like bridging it to a lot of new people I'd say that like most the people who contribute to our project this is their first time ever contributing to a GitHub project at all into an open source project I think that's really interesting and unique and I think that's enabled by reducing

friction to to understand how to get there I'm just realizing as we're chatting that this is sort of a very accelerationist space because you guys are already moving at a speed that I think is unprecedented because you're doing open source Network collaboration but you're building Ai and you're using the AI to build the AI and there's this sort of like accelerationist bent to that I don't mean that in a pejorative way I know to some people that's a pejorative but um you know it it seems like this is compounding really rapidly week over week as you're using the AI to build the AI and it's not hard for me to imagine that within a couple months you guys yall could be have uh AI agent product managers and documenters and QA and Architects and um and then in six months

maybe you have people who are actually leading the road map uh or AI agents that actually have more context and and are more skilled than humans and can actually come up with insights of where the project should go that that we wouldn't be able to I mean is that the trajectory that that you see you're on yes I would say that I don't I don't trust the AI to come up with like some novel idea we haven't thought of but those ideas are already existing within our within our channels like people have it's more like they they can be the interns that collect all the genius ideas that people throw out and kind of just die in the group chats and die in the public chats and like wa hold on hold on that was good let that was good let's get that let's get that this shoot

and so I think that just their ability to process so much information so fast and to and and that also allows for participation we don't oh you you have an idea just throw the idea into the pool and now it's recorded you don't have to like make a proposal you don't have to do all of this other stuff um and I think that like when we look at human behavior we actually have massive amounts of Engagement of people in their ideas but then when you ask them to go vote it's kind of a lot more of a challenge because you're okay go vote write a proposal review the proposal and people just want to throw their thoughts out there and so I think that the llm makes the whole process of like getting your idea into the idea stream way way easier and also way less political

because like there there sort of accidental politics when like I just get the information from the people who can reach me the most which is usually people I'm working with the most and what and that's not necessarily the best chain for good ideas rising to the top so I think that just the ability to flatten everything out and have these agents sifting through all of it and bringing that up so like it's not like the agent had this idea it's like this is the road map it's more like hey all of these people want this and are expressing it so that is the road map and it's the will of the people thing I think that's really important yeah I think JY put in the chat over here the idea is that you have a faster intent to idea to issues posted to actual build

Pipeline and in that compounds that that speed compounds Jen I don't know if you want to come off you and say anything about that but yeah you know there's you know when it comes to like Dow governance for example like there's this process where you have to go through like kicking around ideas and then create a formal proposal and then wait five days put in the timeline yeah it's a long Loop you know and so like a lot of ideas just don't you know actually make it uh it's very slow you know llms can really kind of speed up and expand and kind of write out actual action plans for um and connect the pieces together like for example just kind of like mapping out like team expertise or who is kind of like the best person to

review this uh potential Bounty or issue you know and um it's almost like they can help connect the dots and so just yeah this is really bullish for idea guys there you go there's my clickbait title for the uh for the podcast episode right there all right well we've only got 10 minutes left so I'm going to start moving towards closing questions um is there anything I didn't ask that you both want to say maybe sh will'll start with you and then JY next I mean I would love to know where your thoughts on on how can we make this like retroactive funding model better how can we make it so that more people feel like they can come in and contribute and they don't and they know where to start um and not just for our project but for

projects um yeah you know and and how do we I am an accelerationist i' I've been in eak since the beginning I'm heavily influenced by Beth and and and baz and those guys I mean I I would count them as friends at this point and people I talk to a lot and I really like think we're carrying that but we have to accelerate effectively so like how do we make this something for everyone and not not just you know the some of the people yeah one of my uh co-founders wi I get coin always talks about this model of like I I do something then we do it then you do it and that's his way of sort of training people to to think and do uh with skills that he has um but it goes the opposite way where when people have

skills uh you can do it the opposite way so I mean I think that the model is basically ai6 Z does this this continuous retr funding and then you know once that pilot is out we should do another podcast episode or digest the results and iterate and when you feel like it's good then let's start getcoin and Alo can start helping you distribute it to other ecosystems that want to do continuous retr funding so we do it and then they do it without any training wheels after that so um you know I think that that's my basic model for this going out into the world let's prove that it works at ai6 z um and then let's let's propagate it out to the world uh at that point and my hope is to write another onchain Capital allocation uh handbook uh an AI agent Edition that

will be all about the intersection of AI agents in onchain Capital allocation so I don't know maybe we'll do that in 2025 tweet me if you think that's a good enough idea that we should put resources into it yeah how does that sound sha that sounds great I mean I'm so excited to work on this stuff with you um I mean I think this is this is the thing that really accelerates everything is unlocking all of these people who didn't even know that you could just like work on open source stuff and make a living and like have enough like I think that's great um and Jen is saying some stuff in the chats I think you should you should bring this up I think the circle of Champions is pretty good um it's kind of an interesting take Jan why don't you come off mute and and say what you're

saying in the chat um yeah my question you know because you talk to a lot of Dow people um and just like when it comes to figuring out you know between different opinions and uh forming consensus I have this concept of like the circle of AI Champions and it's kind of like the back rooms concept you know where it's like you have um for example an AI trained on all these core members for example but you can also do non-fiction or fictional characters anybody so like what would be a cracked Circle of Champions you know if you had like an A and template for like a Jedi Council to help kind of like you know approach when it comes to strategic recommendations or

just kind of like yeah figuring out strategy yeah and then you can ask the council how do I how where do I start what do I do what do I what needs done you know and yeah what would be in your like what would be a good template you think well I'm feeling my my podcast toast Spidey senses are tingling so I want to bring the audience along here and and just explain Infinite Back rooms for like 45 seconds and then I'll answer your question so as I understand it you if you have a swarm of AI agents you've got a bunch of uh AI agents that are trained on different things together you can put them into an infinite back room together and they can explore design spaces or idea spaces together um 247 so they can do it way faster than humans can because they don't have to sleep and

then they've also got larger context Windows because large language models than humans have and so the idea is sort of that you're exploring a latent design space in a way it's more systematic than a human ever could like I almost analogize it to when Deep Mind came out and they they were playing go against a human and the design space of go is just like exponentially huge and the the go uh the go AI player was inventing moves that humans could never think of because it could just think on so many different permutations and at first the human thought that the go move was a mistake but it turns out it was actually a really genius move so like Infinite Back rooms for me are like this way of mining a design space and finding like new insights that the AI would trip over

that we never could um so I just want to like create that shared context and then we can talk about the circle of Champions idea did I get that mostly right is there anything you would change about that no that's pretty much it cool um I want to hear from each of us what our Infinite Back rooms would be mine is that I've got this idea for I call it uh infinite regen doai and I I want to train an AI on vitalic an AI on my writing an AI on Glenn wild an AI on Daniel schmachtenberger and an AI on elanar Ostrom and I want to put him in an infinite back room together it's going to be an infinite regen back room and they're going to tell us all the cool that we could build that we just haven't thought of yet because the

design space is so big so infinite regen doai is what I'm calling that idea set and that would be my circle of Champions for the regen movement and uh the onchain capital out a movement and what I like about this is that you know if you think of a perspective is is missing you just find an economist or a scientist that you think you want to add their perspective to the back room and then you train an AI on their on their work and you add it to the back room and then it can kind of like evolve as you see what you're getting out of it so is this a totally crazy Circle of Champions um SL also what would your Circle of Champions be maybe Shaw will go to you go to you next no I think that's great um I I think that and it's not about

like watching them go Yammer on forever necessarily it's more about you comeing you need advice you need you know of the many many things Daniel schmachtenberger has said and he says a lot of things you need to talk to him about something and it needs to pull in that part of him that is relevant that's the problem his podcasts are like five hours long and you just want like one answer for the one thing that you're exactly I I I want to just read all of his podcasts in and then just ask him questions um I I actually think he'd probably be in my my back rooms um I think the toic is someone I think a lot about like what what would he do in our situation I think he's managed you know he's he's dealt with the politics of his station in life very well um I I think lonus

Toral obviously um you know he's a character but he's done a really good job of both showing leadership in a decentralized movement and making sure that it stays true to the vision um with Linux obviously I'd put Jenn in there as probably my oldest web 3 friend um even though we've never met IRL as somebody who's actually kept me true to to the vision in a lot of moments where I needed advice beautiful you know I I'd love to have our our inner circle of people who have shown up and been great leaders and just like be able to ask them questions so I'm not always like you know I mean I'm already asking them questions but you know sometimes you want to trial all the things you say on on the AI before you give it to the real person so if I were

to sum up your Infinite Back Room it would be sort of like an open source development AI accelerationist Infinite Back Room is that too reductionist no I mean definitely I I will say that I I um I love the front rooms I love that this is all happening on social media and this is public because we're learning with the AI and we're kind of like self-aligning and I worry that the back rooms is a little bit like I call it cabal AI you know like the cabal Swarm and it's fine it's kind it's cool it's useful and you can tell really cool stories with it but I really love the open storm I love the like a humans can also be agents you know that like it doesn't have to be AI it should always be in some place where the humans can participate and so it's not really a

back rooms if we're in the council with them um you know I love the the backg is a great concept and we and there's a thing called Echo Chambers which is made by the the genon or Ganan team um and we we've all sort of adopted that and it's it's like literally the back rooms but all of the different agent Frameworks can now talk to each other um so I'm very excited by all of that but generally what I think is really special is is how you can interact you know you can get your answers really quickly you know yeah yeah I I think that like for me the back room thing the cabal AI uh for me like the risk is like are you doing the back room because you're hiding stuff that's like important information or is the infinite back room is just like like I don't know if you

have like a Mozart Ai and it's and it's practicing for 10,000 hours how to like make music I don't want to listen to that I want to listen to you know the symphony that it composes at the end and so it's just a question of like is the back room the back room because it's not worth the human attention or is it because you're hiding information and that's the difference between a cabal back room to me in just like a practice back room or like a learning back room and maybe you just highlighted something else which is this idea that the back is just the window into the thinking process of these AIS and that's kind of what what Infinite Back rooms like Andy's project is is that you get to see that what the llms are actually thinking and doing um you get to see the whole context um and so I

think that's very valuable all of this has a place Jen what would who would be in your Circle of Champions um I I would have uh just to also kind of reciprocate sha makes AI is a pretty cool version of Shaw that's just very positive some uh and foolish open source but I would also that's an AI agent trained on Shaw is that is that right yeah yeah our friend uh made it and he he's out there posting I would also have uh vitalic because he's thinking about interoperability and his writing is good and you know there's a lot of context there I would have musk because we're doing a lot of stuff on X so maybe just kind of getting his AI opinion Mark as well because I mean AI 16z we gota we gotta maybe Mar HRI yeah

we got to have an AI on him and back rooms and just kind of make sure we're kind of steering things in the most based version of him that he would approve or give some feedback on um but also would add in like some case studies you know like maybe a case studies of like an average user kind of agent yeah and then uh maybe like an implementor and you know maybe a marketing person so like they can kind of help uh filter summarize the conversations from their point of view just like how it might affect I think those would be good in the mix yeah that's really cool um the adding the implementer that sort of spawns a whole bunch of thoughts in my

brain about um you know all the devs on on the getcoin core team always pair with each other in order to transfer skill sets with each other and I'm I was thinking like you know could you have like a Infinite Back Room a circle of champions of just really leak coder AIS that are always pairing with each other and sharing skills with each other it shouldn't just be like English language uh information shared in the infinite back room having having coders together learning from each other uh that feels like it could be a frontier for me but you guys are closer to the metal on this yeah that's cool it's it's like the writers room you know like coders is one thing another one like this concept was U you know there could be uh one based on with stable diffusion you know just kind of creating concept art and so

while people are coming up with ideas maybe it's also storyboarding um and that's its way of expressing the conversation well uh I feel like we got to a fun place but un Ely we're almost out of time so the podcast host and me has to bring us back for a landing um my one final question I ask this to every podcast guest and I want to hear it from y'all is uh what does the world look like in 10 years if you are maximally successful so take us to 2034 uh what's ai6 Z's impact on the world what are youall up to I really think that all of our jobs are going to go away I just don't think there's any way around it I don't think that the idea of a job is going to be the same because the AI is going to be better

that at us than liter literally everything and and the top 0.1% of humans will perform and and do this work but it's essentially training data for the next Ai and I think all this is good and it's okay because in that future we will be able to ask ourselves the questions of what really matters and that work as like a labor thing will go away but our passions will remain our our quest for spirituality and our quest for understanding of the universe and our curiosity will remain and I think we'll be unlocked in free to do so much more there are so many things that we are not doing that we could be doing and so I'd love to I know that the next coming time is going to be very uncertain and uncomfortable and there's going to be great change and I know it's

scary and like the the what our lives are about is about to be just completely upended but the far future of this like not not you know within our lifetimes but just on the other side of this kind of like scary time of of change is a really beautiful world where people are unlocked where genius IQ people are not working at McDonald's where they are building the things that they care about and want and where I I really think um we could have like a very bottomup kind of Market driven model where everyone has enough where there's more than enough ideally everyone is just way rich I know it's a it's a hard thing for a lot of people to swallow but I do think that everyone could be rich or at least a lot of people who are accepting and

and wanting to be part of it um and I could see a lot more people just pursuing their passions and building the world they want to live in I mean are you seeing I can't believe you dangled this at the end of the episode this is like cnip for me um do you think like AI funded Ubi Dows to ease the social transition like can you can you can you unfurl this maybe just a little bit and and then I'm gonna invite you back for another episode to talk about this because this is the whole episodes wor the content but like you know it it just react to like AI funded Ubi Dows uh to ease social Transitions and prevent unrest like what do you think that I deeply worry about Ubi as it is implemented by a government onto us because I've seen lots of things be

implemented by the government onto us so my straw man is like Bottoms Up peer-to-peer like created by people that you trust not bureaucrats so I call it Community income like Ci um that we have to make our own systems we Have No One's Gonna Come save us there is nobody coming to help we have to be that change we have to we are the ones we've been waiting for we are the ones we've been waiting for and we are the gener who has inherited like a time of great social unrest so we have to be the strong men who make times great right um and I think that comes down to building communities not just one Dow but thousands of dows and that you join communities that take care of you and you give back and it's a win-win thing it's not it can't just be a handout it

can't just be welfare it's got to be a thing where people are willing to work hard and build community and build social Fabric and and take care of their friends and family um and then we will definitely see that world we want yeah well um open invitation to come back and talk more about that world uh and if you're listening and you want to see us do that episode semi a tweet just so that I know that it's slap in with the community um Jen as we're wrapping uh same question to you what does the world look like in 10 years if AI 16z is maximally successful I'm not sure if this is going to be 10 years or five but I'm hoping that we have like the Linux of uh AI in which it's also like you know something that's tangible I can see

like humanoids uh robots and whatnot um and I'm hoping that we also see like maybe a Resurgence in um Hydroponics or farming and whatnot so it's just kind of self-sufficiency Renaissance because you know they kind of helps automate even like physical production of stuff including food and I feel like that is also going to influence a whole wave of culture back to in a way where you know self-sufficient communities you know can grow their own food create you know and repair parts um yeah I'm just kind of hoping you know this kind of pans out to be like something as long lasting impactful as Linux yeah it's a beautiful

metaphor all right well we are out of time uh JY Shaw from AI 16z thanks for coming on the show was really fun to talk about this design space with you and can't wait to see how things evolve in the future if you're in the audience then links in the show notes where you can check out AI 16z Sha engin on Twitter and also join their Discord check out their GitHub some really cool stuff going on in there thanks again y'all thanks for having us peace you just heard from Shaw and jyn from AI 16z I am just really inspired by this design space of AI agents in AI augmented Capital allocation AI augmented Dows thank you for tuning into the green pill podcast uh and the how to D season we're

exploring the world of DS and the people building the future of collabor cation and sovereignty if you enjoyed this episode we love for you to check out our book how to Da on Amazon or your favorite retailer it's the ultimate guide to Dows and the perfect way to introduce your friends to this exciting design space you can check it out at howto d.x YZ and while you're here please give an offering to our algorithmic overlords Please Subscribe rate and review it really helps the algorithms know that the how to Dow book and the green pill movement are something that you want to support until next time keep building keep exploring keep dawing and we will see you on the next one peace and love [Music]