AI DAO Simulator w. James Pollack

Transcript

foreign ation on the Pod today we are talking about an AI Dow simulator I feel like it's the ultimate nerd snipe this is one of the coolest episodes that we've done my guest today is James who is an artist and programmer has worked for various startups and agencies building software and experiences in the past we connected because of his interest in complex systems and how agent-based modeling endows could intersect so basically at a high level this episode is about a dow governance simulator that James has built basically what if you took a hundred different uh AIS and you gave them tokens and you gave them a governance Forum to argue in what kind

of things would you learn about dial governance from that so really interesting design space that connects a lot of the recent Trends around llms and artificial intelligence with the things that we breathe every day which is Dao's web 3 decentralized governance and protocols so in this space we often talk about how we're speed running the history of coordination and of Finance we could speed run it even faster if we set forward a bunch of AIS to go argue in our governance forums and find the optimal parameters for dial coordination so that we don't have to do it ourselves so really interesting design space and uh James and I have been connected for three to six months sort of nerding out

on these ideas and James is actually building a code base that's helping to explore some of these ideas so coordination I think this is a really fun episode that combines some of the latest thinking on AI and Dows and if we are successful we will be speed running the design space of what's possible in dial coordination using AIS which wow the future is here it's just so cool so coordination without further Ado I give you James enjoy ballet Dao is an open Global Collective working to finance and democratize the governance of synthetic biology Technologies to protect the future of our planet research and progress into synthetic biology is critically constrained by the world of nation-state regulations and bureaucracy from middlemen but the

potential that synthetic biology has for the world is massive it can produce a world in total harmony with nature we can grow stuff using cells like software it's awesome and the best part is we no longer have to pull carbon out of the ground to grow and maintain civilization don't you think we should be funding this Valley Dow does and that's what they're here to do to try and make the solar Punk World of Tomorrow a reality today so join Valley Dao right now and save the planet at valleydow.bio what's up James hey how are you Kevin I'm pretty good I'm excited to get weird you ready to get weird yeah same I'm I've been looking forward to our conversation since we first met a couple months ago yeah same so

um basically uh maybe maybe a good jumping off point would be for you to tell us what you've been building this Dow simulator uh sure so I've been working on an agent-based model and an agent-based model is a computer program where you can um code a bunch of I don't want to use the word in the definition so instead of Agents I'll say like characters you can think of them like characters in a video game uh and you give them roles and rules of how they should behave and then you run uh time through the simulation so there's a bunch of time steps and every step these agents make different decisions about what they should do based on the rules that you've given

them and um the situation that I've decided to put these characters in is that they're all members of uh decentralized autonomous organization so um the Dow members are represented by different agent classes and they can create proposals vote on them invest in projects a bunch of other activities and the simulation also incorporates treasury management token pricing Revenue distribute kind of things that happen at the level of the Dao itself which is one agent and then at the level of the individual dial members who are your investors and delegators uh and Developers yeah I just think it's so cool and I

can't wait to unfur all the possibility space here so um I think the reason that you and I got connected was that in on February 2nd I wrote this post on the super modular Community Forum um about a dow AI governance simulation toolkit uh which I call dags just to try to meme it but um so basically what I wrote was we could build an AI Dow toolkit where I could spin up 100 different AIS give them all tokens and set each AI to maximize their token balances give them a governance Forum to argue in and let them vote on Snapshot and tally and I want to run like 10 000 different AI dials with slightly different starting conditions and like parameters for the agents like they're

friendly or adversarial and see how those dowels evolve differently so basically we're doing a parameter sweep on governance Frameworks and then you can start to like introduce different rules to governance or like maybe you introduce quadratic funding to the Dow and you see how that affects the agent's behavior and um and like once you've done a like this this sweep of different setups you could find Optimal governance setups across the design space of governance and and like to me that's the exciting idea here is that um instead of my friends having to argue in the governance Forum over what's what in their Dow what if we could have a bunch of AIS simulate that and then find the optimal governance setups and for

our human dials actually set up our human dials that way so that that was so I wrote this post and that you and I got connected and and just sort of nerded out from there and and I'm just real I I guess that's that's a to give the the reader the backstory of how we got connected but B just this design space feels like it's so exciting to me and um you know I'd be curious maybe to volley back to you what what got you excited about this design space um I you know I I think I had almost the exact same realization as you which is that there's actually very little data about this space and that generating it um takes time and money and real people's lives and running experiments

on actual people in a dow can be very painful for those people and um discovering discovering whether you know um a certain kind of voting system or you know uh penalties or rewards for behaving different ways inside of the Dow or its structure itself how many members whether you have a native token these are things that are not easily changed right so if you want to test you know how your Dow would run um with its own token or not that's quite hard to do um so I also think that this opportunity to explore that design space I love that you call it a hyper parameter sweep the like technical term and uh some of the

Python I'm using called the batch running but absolutely same thing you want you want to see you know um in The Matrix of this choice and this choice that you could make how does it work when you choose a and b how does it work can you choose a and C so on and so forth um and uh I think I think that maybe a little bit unlike you I'm not so certain that we will be able to find those perfect um you know uh platonic ideals of what a dow should be I think that a lot of that will depend on what your particular you know goals are um and that's because um the agent-based models are typically um they're a source for generating

synthetic data so you know we're modeling a system it's not the real system um it's not going to necessarily have predictive properties but that is sort of um a really fun thing to work on is like where do you put the slider as far as how synthetic is our data you mentioned perhaps um letting your agents discuss things on a forum well do you model that Forum to just be as simple like somebody left a positive negative comment do you maybe build it out more fully so that the agents can talk to each other with llms or do you actually plug it into a real forum and let your agents make posts and interact with real people I think what you decide to do for

that will depend um on what your goals are but it's just exciting to be able to to change those things I've seen uh Dows try to do these and um uh people make these big decisions with you know hundreds of millions of dollars under management with almost no insight into it um so I'm just hoping you know even if a tool like this isn't entirely predictive I'm hoping that it'll at least give people some insight and help them have a better decision making sure yeah yeah totally I mean I can imagine a future in which you know say I'm on the governance or the steward council at Dow X um we could we could I'm imagining a

future in which this tool is um available and you can ingest the the parameters of the setup of the Dow um and maybe ingest The Forum posts that have been on the Dow and and that sets the starting conditions in the Dow governance framework um to be localized to whatever Dow X values and cultural norms and setup is maybe you could expl instead of sweeping the entire parameter space you could you could sort of sweep the space around it and say you know what would happen if we cut our proposal threshold in Pat in half or if we changed our culture such that people were 20 friendlier or we introduced quadratic funding or something like that um to me that's where it gets really

interesting and and and yeah you know maybe it's not completely predictive but it it is an interesting way of of simulating Dow governance changes before you just YOLO make them uh live right I think you bring up a great point that there could be kind of two different users or different times somebody might be trying to use this one is you've got to Dow it exists you're thinking about making a change in that case like yeah it's very hard to change the uh ship if it's already underway um but maybe people thinking about setting up new dials I think that's a situation where your entire hyper parameter sweep would be would be pretty interesting yeah totally um well so can you tell us take us into what you uh what you built so for the

listeners if you click into the GitHub um you can see this code that that James has put together around the Dow simulation framework and I'm looking at the readme now so so take us on a deep dive of what you build for sure um so agent-based models you can build in a variety of Frameworks um the one that I had used previously was called net logo that's a big one people learn when they are starting agent-based modelings but it's based on uh the jvm on Java so it's not as like interoperable as you'd like so for this project I use um a different agent-based modeling framework this one's based in Python it's called Mesa um which is great you want to be in

Python that's where all the cool um machine learning stuff and generative things are happening so being able to be in Python and plug in um is super exciting um so I've built agent-based models before um but maybe not one quite this big or complex so um I started with just sort of a base class of like a Dao member you know what does everybody need to do um they vote they comment on proposals they hold tokens and they have a reputation um even some of these base assumptions like you're already making choices for the Dow right like there's a reputation right we're keeping track of it with points like already I can think of a ton of different ways to keep track of

reputation um it can be kind of paralyzing because you want to give people you know every option um but you do just have to make a couple arbitrary choices to get started and then hopefully leave in some levers so you can come back and tweak that stuff later um and so from the dial member I um you know thought about some other roles um of people in Dows and the things they might do so we have um like developers who um you know be brought in to like build and maintain projects um investors invest in proposals and projects um I've even modeled like passive members people who like hold tokens um and they might vote but they're not

really like getting involved in projects and stuff and um The Proposal creators who create and submit proposals Regulators ensure compliance with external regulations and requirements service service providers that's similar to a developer but it might be like an external developer or if you need something translated um and I also have validators uh arbitrators delegators um you know looking at this list I think going back if I was starting over I look at this and I think to myself like people often hold multiple roles in a dow so this is one of those like base assumptions like does everybody fall neatly into one class of like I'm an

arbitrator or I'm a developer maybe people's roles change and this model doesn't quite account for that yet or maybe people hold multiple roles um so starting with it's got about 10 different classes and all the roles or individual roles once you start running the model it'll be interesting because your developers are going to like develop they're not going to be investing in things but I think in the real world somebody can be both an investor and a developer so I see some potential for that growing out later um so if go ahead oh sorry I mean to interrupt um I was going to say like in a way um we're getting into some of the fundamental trade-offs when it comes to

modeling complex systems of which all economies and social systems are uh it's just it's really hard when you've got these agent-based systems and they're intelligent agents to make uh productive assumptions about who is what and in what ways they behave and um I think that's kind of one of the challenges here is having the right level of abstraction that's not true reductionist uh with each of these and you know I know from my experience working with say say git coin dial which you know I'm now disaffiliated from but I've watched it from the outside is that you know some of the people who are Services providers are also proposal creators are also developers or also delegators and you know you're you're almost like

overloading a super class at that point with different Behavior types um so yeah just a meta commentary about complex systems there um and a totally valid one and one that makes me rise it's important to point out that these agents sort of out of the box they're not AI if they have like rules they're responding to conditions you're saying you know if um token price is more than x do this if it's less than x do do that um so in that sense like out of the box they're not very intelligent what is great about being in this environment and actually all of the cool complex system stuff happening lately is that um

yeah your agents can actually be plugged into these systems like um generative AI for creating those behaviors you could generate you know 200 different behaviors and plug those in and have the agents pick from those or you could plug in reinforcement learning and give your agents A you know reward function and so they could actually change their behaviors on that or um you know let them be self-reflexive and actually modify their behavior while it's running this simulation is not there yet but that is why I built it in Python and so that we have the potential every time you call a step on an agent that's how it works as a Time step but

when we step the developer we don't just have to have them you know do the rules that we wrote at the beginning of the simulation if we've plugged in the ability for them to generate new strategies try them out kill the ones that don't work and ride with the ones that do yeah then stuff's gonna get weird then then we're going to see some stuff in that hyper parameter space that like we didn't even know existed because they're going to be creating new parameters as they go so um um but but level of abstraction is a huge problem with any kind of simulation um and I don't have a great answer other than like you know on the one side you have like maybe a game and then on the other side you have you know your your real data

um I do no good I was gonna say like you know what one of my sort of hunches about how web3 and AI are complementary is that I broadly think that web3 in a way is about heuristics um and the heuristic that we're all traditionally trained on is like my bank account balance I want my bank account balance to go up but in web 3 we now have like an internet native heuristic which is my token balance um and then AIS are kind of about heuristic maximalization in a way which is like how do I make the most intelligent strategy um to maximize that heuristic and so I I applied to this project sort of the Insight was um if you give an llm or an AI a token balance and you just say make number go

up either your own token balance or the token balance against let's call it ether usdc or something like that you would get all sorts of interesting behaviors but um you know just as a way of creating the multi-agent system and and moving towards like more advanced reflexive behaviors versus rule-based behaviors there's like a ladder up there in terms terms of of realism so um yeah just the insight there was heuristics versus heuristic maximization I'm I'm kind of thinking that's increasingly how Ai and web3 fit together do you think that um the Dow itself could have you know a heuristic Beyond you know token go up do you think that these organizations

you know have clear enough missions or strong enough missions that it is something you could embody in the actual you know sort of rules for example like the dial could be set to maximize the token balances or it could be set to you know try to redistribute the most income to its members and not care about having its own balance be high or it could just try to survive um I can imagine a bunch of different sort of goals for the Dow itself yeah yeah what does biology say in the network State book is that all of these Network states have their one commandment which is the thing that they think is wrong with the world that their Network's trying to change and the toy example he gives is like people think that we should be eating keto and so

like there'll be like a keto Network State um you know the one that I'm more familiar with is is get coin and their their mission is to fund what matters so basically to create coordination tools that fund what matters and and you know you kind of get into like AI alignment here because the token balance will be expressed programmatically on chain but then the mission can only be expressed socially and it's almost like memetic social reflexive Dynamics versus token economic reflexive Dynamics how do you create a strange attractor that aligns both of those things as opposed to creating opposed like ones that won't hold together um so so yeah uh I think that you want to have you you probably have dials that have a one commandment that's strong and

you have dials that have no one commandment or a weak one commandment and the Agents walk differently uh based off of that yeah um that makes sense to me um yeah I could imagine if forget client for example um you would want it to try to maximize its number of projects and then you'd build in some sort of rating for the impactfulness of projects and have it try to you know maximize that as well um I think that's that's gonna be super cool once you get that up and running I think most people are trying to maximize for coins alone and then I love the idea of being able to balance between that like maybe we're you know 80 Mission driven but 20 you know coin driven I don't think it's a total either or you

can kind of mix between those well obviously like something maybe a meta commentary is on the boundary conditions of the of the the agent based system because I know say you were trying to model maker down uh their primary product is a stable coin called die and mkr is a secondary token which you can invest in uh or you know purchase that helps create the stability of dye and and allows you to set the parameters on the system and then you've got the governance Forums on top of that and you can agent-based model all of those things um you know I suppose given an ebm and in this toolkit but in a way you can't even model the market success of that

dial without um if you have a simulation that'll simulate a dial that good why don't you just simulate the whole stock market and call it a day right yeah so um yeah feature request James can we just simulate all of web3 uh is I think the ultimate scope Creek above all of this uh absolutely with enough computing power uh I I'm sure that someone could but I can pick up on something you said right there I thought was super interesting um you mentioned maybe running an evm inside of it and that hadn't occurred to me is that like so this developer agent that we're making you actually could like set it up so they had an evm that

it had access to and maybe you give it an llm and you'd let it write code of some kind I find that very interesting the idea of actually like um having some agents that can run some solidity code because that is what people at these Dows are doing ultimately so in a large enough self-correcting system would the llm Discover code Audits and re-entrancy attacks and all of those type of things well and social engineering attacks I think would be interesting because it's going to be getting all of this other stuff from the Dow um as well yeah that's I mean let me write a note down this is what's the the analog in like um I was watching the Deep Mind

documentary about about go and how there had been sort of these human go players that have a cognitive limb like limit to how far ahead they can think but the the AIS didn't have that cognitive limit and so they could explore the possibility space and they were discovering new moves that the human players had never really ever thought of and I think that that's where it gets interesting with these AI based systems um totally I'm ready to like press go on the simulation sit back like you know take a look at the graphs or read out and just be like holy crap like what's going on here I think that's the moment that I'm shooting for certainly yeah well

um I I guess you know one of my next questions for you is like is is how do you how do you want to monetize this I mean I can I could see you very obvi like a very obvious uh business model for you with this would be going to Dao X and saying hey I could simulate your governance and I could I could simulate a bunch of things about your Dao um you know uh pay me x x number of tokens and I'll simulate your Dao and and we can make recommendations about how you should evolve your dial forward based off of what the AI Model come comes back um but it also I know this is an open source piece of code so I hope to see it on get coin grants in the future uh it feels like you're building a public good for the space if this

thing reaches any level of usefulness for for Dallas to be able to simulate their upgrades um man that'd be so cool if somebody helped me build some sort of module that let you plug in your actual Dow information on the top of this that would be amazing um probably just training data I mean at my level of abstraction it's just training data I know that there's probably a lot of things you'd have to change but totally um so monetization is interesting um with a project like this uh yeah the Consulting route um Dows do pay a lot of money to people to come in and do the salaries or sometimes like a million dollars to have someone come into your Dow and help you make a decision because there's a lot of money at stake so you know maybe we're

not going to offer or through the simulation be able to give you you know 100 efficacy of what you'd get from a consultant like that but if we can give you 70 at you know one percent of the cost I think that's a pretty good trade-off um so one interesting thing I kind of mentioned earlier is this slider between the gamification or not um I actually think one cool way to maybe make some money off of this would be to um turn it into a game so at every time step uh you would you know add some goals add some options for what the player can do and then you know simulation games are a pretty big genre there's a lot of people who play those

already and this would be um another cool Advanced simulation game that's kind of web 3 focused and so like you know a blockchain um focus game that's not actually on the blockchain but dealing with blockchain topics like it could be pretty successful um uh I think everybody the past couple weeks has been freaking out about this paper that came out where some researchers took chat GPT agents and they put them all in a little town and they had a um they threw a birthday party for one of the other agents and um people are like very excited about that and that seems like very table Stakes for me like oh I was I was surprised I guess is what I should say

it's like people were like that excited about these agents planning birthday parties I think like if you could give someone the ability to come in and play out their fantasy of like I'm you know the number one contributor to a Dao or like I'm a super smart Dow investor or like I'm an incredible delegator I know just who to give my tokens to to make things happen um I think that there's a way to monetize that um yeah um I I think that um you know what comes to mind for me is uh when I was a kid we used to watch this show it was called BattleBots and it was just like people who had built different robots so they would fight each other um and and I'm almost imagining like the

entertainment value of of having people fight in the Dow governance Forum uh and they're just like Battle Bots but for Dow governance I think would be all like interesting in my like emotive effect of like my friends are fighting in the Forum is is way different than like you know something I'm emotionally detached to it's like oh look at the meta of this Dow governance bat uh like bot bat bot battle that's happening in the governance Forum there's almost like an entertainment monetization or entertainment shelling point there like if people are joining in to look at agent birthday parties then I'm sure that governance Forum well yeah they'll be into like corporate rating and all the weird stuff these agents are going

to do once they get going vampire attacks on other chains like I think that would be fun to watch and um you know these models run had lists you're you know in Python you're just running a step and it's outputting some data and you can visualize that data with like Network graphs and different kinds of visualizations um but some kind of like visual on top some Sims kind of thing I think it could make money that way obviously just like what I learn and others learn who are like heavy users of a product like this I think they'll very quickly start to have um an intuition backed by you know synthetic data perhaps backed by you know real world data as well into what

works um and I think people like that will become you know valuable dial members a like this tool should let you on a personal level go into a dow set it up Model A change that you want to make and come to that dial with some evidence that says like hey we should make this change because when I run it through the models in all these different scenarios it shows us all doubling our money I think you'll get more votes for something like that so on a personal level like I I guess I could use it to just make my own Dao investing like really sharp and try to monetize it that way um but that seems not as fun as building stuff to me um yeah totally so I don't know do you

have any thoughts what do you think I mean I just think that there's this is you know just such an interesting and novel an important design space I think that there's many implications in past that uh that this could go down but I mean for me you know one of I hope that one of the results of of this podcast is that uh you know we've got 5 000 listeners between five and seven depending on the episode and I'm hopeful that like if just one percent of them check out your GitHub and they're like oh I'd like to you know hook up with James and run a simulation against my Dao um you you get more interesting I think results partnering up with the existing dials and helping them solve

their problems than just sort of building this in the abstract you know this is such a wide design space Maybe eventually it'll be generalizable to all dials but just like hooking up with existing dials and helping them with their needs I I think that'll get this to the next stage of maturity I think that's right on I have been building in the wilderness a little bit it is an abstraction right I'm not an active member of a bunch of dows um my friend who I spoke about the idea with who pointed out your Tweet to me he's uh his name is Dane Lund he was the DOW architect at Alliance Dao and now I think he does like the um governance stuff for optimism but he's not like on their board

um and he this was about a year ago now we talked about it he gave me a lot of the insight into the Dow side of things then I took it and I've just been building and yes I'm totally at the point now where I'm like but what does this mean in terms of Dao it's like what should I add even like the steps that the Dao is doing when I say like it's you know redistributing to tokens or it's doing BuyBacks or it's staking some of that stuff like I haven't sort of like an intellectual understanding but I don't have any of the ground level like oh yeah we we implemented staking here's what it was like and that's the sort of stuff I would love if your community brought that to us and they're like oh yeah this step where you're planning on

doing staking like here's these other few things you didn't even know about um that would be amazing yeah for sure and and you know take that and you know they're staking and then there's protocol upgrades and then there's I want to introduce quadratic funding to this community and see how it changes things I think there's a lot of opportunity for people who are building Dow tooling to come here and hook in and say hey in 100 different simulations our tool increases the coordination carrying capacity and the token price in the Dow 97 of the time or something like that like there's that's biotics for that right there's nothing to tell you whether a decision you're making you know and I don't know if we'll get to good or bad totally but just like what

any of the implications of it are and I have the sense that that is valuable that's why I thought about the idea that's why when I heard um you post about it I wanted to get in touch I'm you know I'm not a Dao uh uh owner or anything member I guess that would be better um actively right now so I don't actually know and yeah there's a certain point where like I need to talk to users and that point is now so that's exciting when you say monetizing it one uh okay I don't know if it's a good or bad idea but what if like some part of the model was like on chain and like you know I guess I'm thinking more of like if you were going to do kind of like a

pay-per-use thing or something it might maybe it would be interesting to like you could deploy this model on your own blockchain or something like it could uh it depends how much information people want to reveal about their own stuff maybe do you have to use ZK so people aren't revealing you know private company info but um yeah as far as like would I want to run it as a software as a service or a game it's totally unclear software as a service seems hard and like on chain to me but that was just something I'd an idea that came through my head I thought I should throw out there while we're talking totally yeah yeah I mean I guess there's the whole you know you open source the code so that people can develop on top

of it and the network of that is valuable but then you sell a SAS uh tool that's uh linked to it but you know to me the hard part of this will probably just I've got a passing experience running AI agents is just that they consume a lot of compute power and so if you can abstract away the need to spin up a bunch of cloud servers behind a SAS offering I think people will pay for that but um you know it depends on what we're talking about here are we going to simulate the whole web 3 ecosystem or just a dial uh within that and depends on whether or not my feature creep of trying to get you to simulate all of web3 uh mix it into the the code base or not I think we're gonna have to at least make the next agent I think we'll make

as a forum like I find it very interesting that most of the communication that happens in these Dows is Forum based you know text-based sort of threaded conversations um and so yeah I don't know if we'll need to model all of everything but we probably will need a fairly robust model of a chat Forum which I find kind of interesting and um and I have no idea what will happen once you let a bunch of bots Bots chat with each other there but it'll be fun and I like the idea that like um we'll be able to like pull the headlines of their posts or whatever you know you'll see that they're getting you know 200 replies on a post like we should change to quadratic funding or like who

knows how the Bots are going to write maybe it'll be like quadratic funding a bunch of emojis like I'm hoping they're weird and cool yeah I'm sure they'll be weird uh we'll see if they're cool if they learn learn coolness but um yeah you know you get into some questions about how would you model a social space for a digital like you know that just having been involved in Dallas I think that the governance forum is where the formal governance happens but um I learned this thing about organizations oh man like 15 years ago one of my mentors taught me that like hey the real culture of this is not within the meetings it's in the hallways and like in the happy hours and what people actually say to each other when they're not afraid of uh reprisals from the boss and that makes me think oh well the

Dallas can fight in the governance for him but what are they Discord dming to each other and what are they tweeting about uh and so simulating those social spaces is uh it seems complicated but with llms it actually feels newly newly easier yeah the past even when we first started talking a couple months ago things hadn't tipped quite as far as they have now that it does seem more approachable to model um the sort of softer social spaces like you're yeah I'm like well I guess I gotta add a Twitter simulation and like a little bit of a Discord um yeah I think that there's a lot of um there's a lot of possibilities here and I'm curious to see how the uh how the project evolves and so there's big

Ambitions of of being able to simulate lots of different things in web 3 before we actually put our tokens and our time into them but I think starting small and hopefully you pick up a couple interesting relationships with Dows from this episode is is where I see this going and then we'd love to have you back I mean if this if this continues going then would love to have you back in Q3 or Q4 and just see how our simulation has evolved as a category that that you're creating so um creating a category by stomach dropped a little bit when you say that but like it is cool like um I really appreciate you starting to accumulate uh gravity towards this idea

it's an idea I had had before but it was kind of floating once you grounded it and and now there's code now we have this discussion there's something for people to look at I'm super excited about um where it's gonna go I had a couple quick questions for you maybe just around um what are like the features you would want to see in a dial simulator or do you have like specific questions you would want to answer or do you just imagine like the coolness of getting there and playing around with it um yeah what what can I do for you so next time we meet uh it's yeah something that you'd like well um I'm no longer super involved in any dials since I've disaffiliated from

Bitcoin dial leadership but um I mean I do think that the the the problem space that I see around Dallas is what are the initial starting conditions that we want to configure our Dow and there's like proposal thresholds so if you needed at least x amount of tokens to vote uh or sorry x amount of tokens to submit a proposal why amount of tokens um in in order for a proposal to pass you know there's just basic questions about what are the starting conditions of my Dow that people have that I think if you can uh inform them on then I think that you add a lot of value um I know a lot of people in the Dow tooling space that are saying if you integrate my tool X it will help your

Dao why in some way and if you could prove that out in a simulation I think that that gets really interesting um and then there's just like the pure uh there's the pure entertainment aspect of of having the BattleBots and having them fight in the forums and see how the meta that evolves and and um so I think that there's those are some interesting threads threads for you to pull yeah thanks um that is the plan right now is to to pursue it I do work uh full time as a senior technical artist right now I'm Contracting with meta um I also work on on-chain games with some friends of mine we have a company we're building on-chain games we should talk about that separately at some point but that's been

a real wild ride um as well but I am um going so my interest in this came from a class that I took from the Santa Fe Institute uh they're this cool complexity Think Tank based in Santa Fe and I took their agent-based modeling course um they're the ones they have a tutorial on this python Mesa language so that's how I got exposed like oh I can build this stuff in Python but I'm actually going out there for a conference on collective intelligence in June and I'm going to talk about not like officially talk about the Dow simulator but be able to discuss that with people who get agent-based modeling and get this complexity stuff and I'm just really

excited to um to to bring that there and get feedback and sort of um a lot of people are are very focused um organizations are something people have studied for a long time but there's very little research into these digital versions of them yet and I think like the Dow the a um came across as autonomous and I think that put a lot of people off in the beginning um from the Dao it seemed like this thing that was totally sort of separate from people that it was just all code but um now I think the important part is the decentralized aspect of that if you think about this as like a decentralized

company you look at what happened during the pandemic how things have spread out how everybody's work from home now I think it's like an inevitability that most of this stuff will call it five years 10 20 years most organizations will be decentralized and so you know we have a chance to learn about this stuff now and sort of get it out of the gate on a good foot with some tools where we can like perhaps avoid shooting ourselves in the foot as badly as we have with some other uh public goods I guess yeah I feel that um and you know it's a great point about the prescience of of this opportunity in this moment but um I'm also just in it for the nerd

Snape of uh of of experimenting with how AI endows fit together in digital native organizations it's it's just a fun design space to explore and and this has got all the things it's got agents it's got llms you've got procedural content generation large language models there's like so much fun stuff in here you throw it all together in the python soup and um and I think we're gonna we're gonna see some really interesting things so yeah if you're out there and listening and you have a a Dao and you'd like to model it or help me um figure that out I would love for you to reach out uh yeah make some comments absolutely so click the link in the show

notes for the GitHub repo and get in touch with James is there anything I didn't ask that you want to tell us James before we wrap um on your on the prep sheet you had a question a couple questions one was about a period of time um that I found interesting like a history kind of thing Silicon Valley history maybe and I had an experience recently where I kind of got plugged into that that I just thought might be kind of interesting to share I make um interactive art and I got reached out to buy an older interactive artist his name is Larry Cuba I went to grad school at UC Santa Cruz he is affiliated with

them and he um on the original Star Wars in 1977 made the visuals for the Death Star that you see when they're like showing it to the rebel pilots and he did that at the time all real time he made this like crazy thing where he could like turn the knobs and it would like change like definitely kind of like over engineered for what they ended up using in the movie but it's just really incredible and um he's been making procedural art since then instead of going like the VFX route he kind of stayed the more um you know uh Fine Art route and he is working on some new like procedural content systems and uh needed some help and it was really cool to plug into an older

generation of programmer and engineer and to just realize like here's a person like a very talented programmer encoder they just don't know the latest tricks or they don't know the syntax for blender um and so he had been blocked for a couple of gears on making stuff and so I've been able to work with him get him unblocked get him building stuff again um which is for me as an artist and software engineer I found very satisfying and it plugs me into this history in an interesting way which is that Larry Cuba um one of his mentors was this guy John Whitney senior who's like one of the first people to ever do animation on a computer so like now all

of a sudden I'm like two steps away from the first computer animation and I find that like very exciting so if your readers have a chance look him up Larry Cuba he's got super cool art um that's maybe like a little different than the Unix Bell lab stuff but he was so creative that they like asked him to do this stuff on this huge movie and then um uh I I just think like taking it instead of the VFX Direction in the Fine Art Direction was a really interesting choice um and it's led him to some very different places than people you know he could have gone and started ilm I guess or gone gone there after it so ilm being industrial Light and Magic industrial Light and Magic yeah which which sort of

came out of that a lot of people ended up going and working on that um so so yeah that was just a neat experience getting plugged into the history of things and um it happened very quickly you know I went from having no connection to like the history of computer arts like being a couple steps away from the beginning and I'll just encourage people be available post about your skills and what you do like uh he reached out to me because on my LinkedIn I said I do some Python and blender and another professor from there saw and and those things don't happen you know luck luck comes in groups that are prepared I guess so um you know we had a discussion we had a discussion about uh um before the call started about whether to

I was like a little nervous about like should I open sources code should I not and actually like as we've gone through this conversation I think I made the right decision like you know the prep is there it's out in the open now the light is on let's hope it brings the right fireflies you know yeah well if you're listening to this this is the uh This Is Us dropping the story of of James open sourcing the code five minutes before recording so that I could I could send people to it in the show notes so um you yellowed you gave it to the universe and I hope that the Universe Universe reciprocates with code contributions and uh memetic love and hopefully people hire you to simulate your Dao as well but uh well thank you

and um yeah and plus one on I mean I think we're all standing on the shoulders of giants here uh I I think about the for forefathers of Open Source software and how they beat Microsoft in the 90s to make every web server open source and you know it could have gone so differently we could all be using like clippy phones and clippy coins instead of Bitcoin and stuff like that so we're all standing on the shoulders of giants and I take inspiration from the lineage of people that have inspired me as it seems like you do as well and well maybe in maybe in another generation people people be like well I got connected to James the first author of a dow AI framework so that's all we can hope we're gonna make me blush um yeah and I have to say like I'm a PR I'm

a product of public goods I'm I'm the first one for my family to go to college um my undergrad was private but I went to public graduate school I learned you know I learned everything I know from the internet from I've never met my dad so like how to tie a tie um to how to code so I I'm a big believer um in information and its desire to be free and and all the cool things people can do with that um so thanks for the opportunity to come and um tell people about what I'm working on and hopefully uh draw some new folks in yeah well thanks for returning the meme of AI Dow simulators uh from just meme into actual code excited but I mean you put it out there and sometimes sometimes the world just

makes it happen it's cool yeah well that's a great place to end it James thanks for thanks for joining and keep up the great work peace thank you Kevin please move [Music]