Season 6. Ep. 9 - The Emergence of AI/LLMs in Web3 Grants: Exploring AI-PGF and Beyond W/ Code

Transcript

welcome to the green pill. network podcast in September green pill will be hosted by the cartographers a high context community of individuals and teams dedicated to advancing the web3 grants ecosystem they'll explore the web 3 grants ecosystem with topics ranging from novel funding mechanisms Grant councils and governance to emerging trends like Ai and direct to contract incentives this series seeks to bridge theory and practice by examining real world examples and featuring discussions with those who are closest to the action the next voice you hear will be Sav who contributes at getcoin and founded the cartographers sa will be hosting this entire season and I think that it's going to be a great season because Sav brings deep experience in web 3 grants

and capital allocation I'm excited for you to hear some of the perspectives from him and his guests without further Ado the next voice you'll hear is Sav enjoy hey coordination welcome back it is sa I am here again for another episode of the green pill podcast where I am a guest host I've got with me on the episode today plug plug and I have worked together in some various places um he is deep in the web three grants public goods face uh active Builder with potlock and doing all kinds of cool stuff that way so I wanted to pull him on the episode I think he's got some really interesting perspectives on AI llms and public goods and just kind of the impact of these things you know as we move forward so yeah let's kick it

off so plug I'll turn it over to you maybe give us a quick introduction let us know about the things you're working on and we'll see where the the conversation takes us yeah um hi everybody I'm plug um I'm a really big fan of green pill this is like every time I'm on a plan I download as many green pill podcast as I can and it's been a pretty big driver in terms of how I learn and how I also uh shape for the product and the protocol that uh we're building and so yeah I'm really really grateful for not only all the work that you're doing in the grand space and green pill and and just gitcoin in general for like Paving the way so yeah big shout out to y'all but a little background about uh me um basically funding uh starting building an open

funding stack called potluck which is uh the goto public goods funding platform on near um I also work at near Foundation where I had public goods focused on a lot of of interfacing with Founders and Builders and identifying infrastructure graphs and grants and then building also funding mechanisms and tooling and then one of the most recent initiatives um out as part of this funding stack and it's part of um nearest kind of overall Vision to build build a user owned AI stack is AIP GF or um AI powered public goods funding um which is a proactive grants program but also of movement um to essentially build a funding AGI or a system that can give out grants better than but there's a lot to to that and there's we're still

pretty early in development in that phase so it's just a high level overview nice nice no this is some exciting stuff and I know you and I have been talking about this quite a bit you know for anyone that's been following the podcast this season uh we've been exploring all different types of you know kind of design space spes and things that are going on in the grant space and some common themes that you know listeners have noticed probably that's coming out of that is like verifiable impact you know bringing things closer to the smart contract level automating more incentives like all these things and I think that in many ways AI llms are going to be jet fuel to that right and kind of like being able to look at different data sources rationalize that help us make better decisions make decisions without humans like there's

just a lot of interesting things that are kind of going on that way for sure so when we look at this episode and kind of some of the big things I want to make sure we touch on is like you know how do you see Ai and llms being integrated into grant making processes today and then what does you know kind of like the near future look like that way what are some of the benefits and challenges of AI assisted public goods funding the impact of this on kind of Grant proposal creation evaluation and impact assessment and then dive into like future Trends and where we see you know funding mechanisms going right so um maybe we'll start with this project AI pgf that you've been working on um maybe give us an overview of like what is that what were the origins and motivation

behind it um and kind of like the current state of that program and some of the things you're working on yeah and so I A lot of times in the the PG space and even building out the funding mechanisms we continuously like to innovate experiment and this is all fun but from a practicality perspective of like running grants and running funding rounds a lot of these mechanisms are really hard for with ux they take a lot of bandwidth for Founders and these certain mechanisms don't necessarily align with what ecosystem treasuries want for for their ecosystem and so I I started kind of looking back at the problem and what we're trying to solve and um ultimately I want to make it as easy for Founders to find funding as

possible so that they can just focus on shipping and the more traction that they get the more rewards they'll automatically get and that ecosystem treasuries can easily find people who add the most value add to their uh to their ecosystem with as many friction as possible and so it wasn't really build the most decentralized funding stack the most censorship resistent funding stack the most composable funding Stacks that that wasn't necessarily there these were kind of in the direction of what we were building but it really came down to like how do I solve this problem and like recently um especially now when um since July when you kind of put this um idea out there's been a lot of emergence uh with AI technology especially with open source Ai and the user own AI stack

where even just now we can actually use AI to automate a lot of the workflows that we're experiencing as Grant operators so I think a cost can come up to like 10 to 20% um in terms of overhead for for Grant programs for some of these programs and these are like million dollar sometimes even billion dollar programs um in terms of like the collective web three space and um it's they're just so much admin and I saw it as a huge opportunity to not necessarily have to build the most decentralized thing but to also um really tackle this problem by just reducing overhead in the time it takes um for this funding process initially and then moving towards this vision of like taking each part of the Grant workflows and then

building composable agents to eventually uh build this funding hii but it really came out of that like oh I'm really frustrated about running these Grant programs and these mechanism designed and I just really want to focus on Saving Time totally yeah you know so some experience that I have with that is like you know internal to getcoin we have this tool Checker you know that we've started to build out that basically takes a look at the eligibility criteria the answers that grantees give you know weighs those two things against each other and it provides a score that kind of tells us like how close to the eligibility and and what's being asked of a grantee does their application align to right and really this was born out of the fact that we have getcoin grants you know every quarter there's lots of applications that come into that and how

do you essentially kind of create a a way that um reviewers can streamline that process and kind of easily see which are the applications that are in Easy yes which are the ones that are in Easy know how do we give better feedback you know to grantees and ultimately improve our turnaround time right um we see all kinds of issues today with like grantees and they all have different experiences based on the the programs that they're working with and I think some of that is the fact that you know programs are overloaded and to your point there's a lot of manual effort and kind of things that go into you know running one of these programs and it can be be hard to keep up with the volume right and so in that when you you kind of talk about aipf and you know some of

the things that are built today maybe walk us through like um what do those systems look like how are they starting to drive efficiencies and where do you see you know that going in the near future yeah and and so right now like there's so many elements towards like the overall Grant workfield process there's everything from like you know writing proposals to um to to reviewing proposals to checking eligibility to posting updates to doing the KY see payouts to a voting mechanism if it's like quadratic funding to doing diligence on teams to even giving evaluators and voters incentives for providing valuable feedback there's just so many elements and agents around there and systems to address these individual problems in the funding workflows could

be like its own startup in itself and so it's really like going at all of it at once isn't necessarily the best approach and so we have in kind of if you go to AIP gf.com exlained we kind of outlin the road to the funding AGI and how it will look like all the way from demoware which is things that kind of look cool but don't actually work all the way to a system that gives out grants better than humans um and we really look at that Evolution as starting with demo Weare um and this is kind of already the case so you have a really good example uh that I like is funu goods. a which queries a lot of projects on gitcoin and kind of gives you a strategy for distribution and also gives a score for matching for similarity it's like that's really cool

it it's a really cool demo but I don't think anyone's necessarily using it in uh production um and it's really kind of siloed with a few agents um then I see kind of an evolution of like many agents like okay we have different people building different agents there's someone building AI grants you know you guys are bringing eligibility requirement uh Checkers maybe we designed an agent for Designing the best mechanism for Aloe these are kind of all independent agents that we're using as sort of a tool Maybe don't necessarily give them a wallet we just kind of use it within our workflows um and then we have a little bit of the data layer so how do we Within These agents like how do we begin to pipeline this onchain

grant information whether it's from gcoins Project registry or it's for karma Gap or it's from evaluation from metrics Garden like how do we take this data and pipeline it back into the these systems so that these agents are kind of fine-tuning and and adapting to um overall systems and then how exactly do we combine these agents together so that they interact with each other so that once you write your proposal it submits The Proposal on chain to the to the respective Grant programs that once the proposal is submitted know there's another agent that knows to review and give feedback of another agent that pings another person maybe another agent that actually you know checks if you have an SBT or you're already um meet the requirements for a grant and another

agent that triggers the payments after they check for kyc or whatever you can kind of see how all these independent agents that we're using in our workflow can connect together and really build an entire system and the reason why I think web 3 is a unique place to experiment with AI and grants is because in traditional foundations even in traditional Venture when it comes to funding like there is no open API for the data on project information there is no open API on their milestones and there is no programmatic payments layer um to trigger this and there isn't large of Treasury that actually have it in this programmable money so I see um this as like aipf in the context of web 3 as

one of the best places to experiment because we have a lot of open people that would actually and this is the next stage of the evolution of aipf want to Pilot this with a microgrants program and actually give an agis or a collection of Agents um a budget to give out funding based on certain um ecosystem kind of par and then eventually building to a funding AI where we're actually having million and billion dollar treasuries run by kind of the collection of these agents um and so in terms of like what is live right now I'm not planning to build anything I want you know the getcoins the other builders I want people to build on alll stack I want people um to build on top of any every kind of funding mechanism and then every

uh with kind of this user owned AI stack this open source AI stack this uh decentralized AI stack to begin to plug in in a lot of these pieces and so there is stuff internally we're building at potlock and then there's stuff we're also working with other teams to ship out in terms of these agents and we're still pretty early and one of the uh we have a lot of different gaps in terms of like the already the like that's in terms of the selection of products and tools that are available for AI developers especially open source and decentralized AI developers that have kind of that limit where we can progress right now in terms of this road map of the funding AGI so um right now there's there's been agents that are been created for estimating the time it takes

for a proposal for helping kind of diling project for even uh um you know using levels that FYI and estimating the amount of budget and whether it's um an accurate budget there uh we we've done also like Checker like eligibility agents we've done agents to write proposals um and actually cast it on chain uh but we found that like we have all these Silo agents um one of the main things that we're focused on is actually an agent playground that can connect a lot of these together and provide kind of this Dev tooling layer um for um the this multi-agent framework that's predominantly what we're working on right now is actually shipping out the Asian playground um that is compatible with a lot of these uh like um a lot of

these AIP pgf um agents that the ecosystem is building there's also we're starting to like interact more with like working with different uh evm protocols too we're working heavily with like via prize on turning kind of all of their contracts into different agents and this is like one one thing that like we've kind of struggled in with the past and especially like as like we're building a funding stack you know gitcoin is building a funding stack is that um it's really hard for us to integrate our daps into each other or our contracts into each other but with agents I see us kind of compartmentalizing a lot of the best parts of our protocols and then uh providing kind of a base layer to interact with each other so I see this as a way uh for kind of

interoperability in the uh in the pgf space to really uh kind of come outside of the evm so yeah I know I kind of rambled a lot but hopefully that gives a better view like super fascinating man I mean um definitely so much to unpacked there I mean one I could not agree with you more and that's why I'm so excited about just the web 3 grants and incentive space around you know these concepts of like you have the data that you can introspect and it's open you know there's so much potential that's unlocked through that in terms of you know really leveling up the the process of evaluation and you know measuring impact tying incentives to that in a way that is just not really possible in many

you know traditional kind of like Grant you know funding structures so it it's fascinating to see kind of the layer of agents that you know you start to build around the top of that that are context you know domain specific then they can start to you know talk to each other communicate collaborate uh Etc when you talk about this this future that you see right like do you see that playground concept that you mentioned being something that if I'm a grant program there would be you know sorts of kind of like Open Standards right like open you know systems that I would be able to opt into and so as I do my evaluations that benefits the broader Collective because that's helping to build the context and kind of like the expertise of these

agents right I think right now I see we have hundreds of Grant programs across crypto that is the primary way that we're driving growth in our ecosystem uh but none of those programs are connected in such a way that you know um we we talk about composability and kind of like being able to build you know systems on top of each other from our Tech stack but the people process around that is very siloed and so I I view this as being something that can really unlock a lot of potential that way you know do you see it that way too where it's kind of like this be this would be a public good this would be something that people could opt into programs could opt into and then that grows the context of of kind of The Wider Collective as you know they use these

tools yeah and ultimately I I definitely want these silos to break down it's costing us a lot of money even at the near Foundation before my time we had a lot of Grant programs that we had a lot of Grant farmers and historically that actually really hurt our ecosystem and actually made us not want to interact with grants and really kind of bring it down back for just critical infrastructure and focus on kind of um getting teams Venture funding and a lot of that could have been prevented if we had information about different grantees and different projects and how they behave across different ecosystems so even if I'm building on near like it would be advantageous to integrate Every Chain in any data Pipeline and even

offchain information whether it's checking their Twitter their LinkedIn and kind of and I think a lot of this like pulling from external information now especially with decentralized compute part of this user own AI stack we can actually gather and publish onchain to build these open apis for every chain um in in a very decentralized way and so I'm really um excited uh about that and building out this data layer and uh I think it like we definitely need to collaborate and I'm I'm also very open for new chains that don't necessarily have this data layer this accountability portal or way for projects to post their updates or um for evaluation to actually build these Primitives and these data layers so that agents can um can operate on top of this

so I'm all for uh collaboration across every chain like ultimately like my goal is to make it easier for project Founders to get funding not necessarily to push a certain chain um and at the same time a lot of our clients or like people that we're building software for are these treasuries and that they're ultimately happy and that they're funding the best team so like that that has uh necessarily nothing to do with being siloed or being closed in how we operate um and there's a lot of cross collaboration across kind of all ecosystems if we're kind of building an open Tech and we're siloed I think it's very hard to make that case to traditional foundations and and other people in the space uh to integrate as well so um I do think this is like

something that is actually loow hanging fruit like as an industry of rant operators there's not many of us and I think we can kind of systematically um build out a lot of these data pipelines and and that's what I really love about what you're doing at cartography Syndicate because um it's incredible work and we plan on plugging into a lot of it nice nice no I appreciate the the shout out that way for sure we've talked a lot about kind of like the programs perspective on this but ultimately the inputs are from Builders right like when you look at what these agents are going to look at you know the connections that these agents are going to make the impact that's going to be created that is all Builder side it's the programs that offer the incentives you know all of these things that can enable that but

it's the builders that deliver on it how do you view that Builder experience changing right like today or evolving you know today we have tools like Karma which we all love you know where you can track your milestones and you know start to do that in a way that's on chain you have open source Observer which starts to surface kind of like you know those Builder metrics and there's other you know tools and things out there that um you know play into you know surfacing the the impact of Builders and you know providing that data Upstream to program to be able to make good decisions how do you view that Builder experience changing you know how am I as a builder going to interact with these agents going forward you know how's that going to like really streamline you know the way that I'm I work today and and you

know help me versus what is a very highly manual process what is a different process for each program I interact with so maybe walk us through like that Builder experience yeah I mean what I in Envision like that Builder experience is is really an easy one one click approach for setting up their initial profile in their wallets and um things to happen automatically as a result of that and so one thing that I really like like karma is great but ultimately like we need to make the experiences of reporting As Natural into the workflows that are already existing and so um like here's an example like right now um you know after I do get a getcoin grant a funding round you know I

think 22 is coming out let's use that as an example you know after the qf rounds I post I go to you know KAG Gap and I post my latest Milestones I still had to go to Karma Gap to to do those Milestones when I went to gitcoin and I had to register here and then I had to do all this promotion on Twitter anyway for the R so how I Envision on this is essentially where we're already posting whether it's Twitter whether it's LinkedIn to actually dynamically have those um scraped and then published on chain um um to kind of give those updates um and to for grant program evaluators to to automatically kind of assess this information and then reward retroactively or proactively based on

this that's more of like what I kind of see this as rather than going at different deps and like you can easily uh just have this happen automatically on the back end and integrated within um your current workflows and so um that's what I'm hoping for and I think we can make a lot of uh progress on their like on that front um but um yeah right now it's it's pretty siloed in terms of like this dap this dap this V2 dap and so yeah like ultimately we have to make the user experience easier than than web 2 so totally you know and there's already some work being done that way where you know automating some of those connections between you know say gitcoin

and Karma where data can be pushed back and forth right but you know what you're saying is like then that's taken to another level where then agents or interest like actively watching that and proactive and being able to go look for opportunities match them Etc this gets me thinking about this concept of markets right like at what point do you see this this ability to unlock almost grante markets you know occurring right where you know I could I could Envision a world where you have these systems that are you know built together and built on top of one another kind of like sharing data back and forth in a way that you know they can reference that data you can start to see what Builder

impact is you can start to see what the landscape of opportunity is from all the different programs out there could you see a world where agents help us to create these grantee markets right where I have very specific things that need to be created and if I know what the pool of Builders are their qualifications the rates for the kind of work that they do could that be something that you know a market emerges from versus you know kind of this application process that we have today yeah and I think there's really like if you look at electri capitals developer report there really is only like a few thousand full-time developers and maybe few hundreds of like full uh full-time teams that and a lot of them aren't even necessarily available for

work so I think as an industry together we can actually catalog a lot of these teams already and like we're all used to onchain systems so I think if we it's a coordination issue to actually get this within uh uh the next quarter and so um I definitely I definitely see a a world where we have context on all the builders all the ecosystems where they're at um and we already have pieces of this like we already have like open sour Observer absorbing all of most of like the op Stacks repo and there's even stuff like ecosystems. Ms that like are like just been scraping GitHub forever um and so there is um already people who are scraping you know onchain traction metrics this is what they're doing for

like retro Round 4 and ret six and um you know pulling from forecaster data um pulling from verified contracts automatically seeing the onchain traction um you know you can already check and verify Twitter on getcoin passport you can kind of sign a stamp and link your Twitter so you have all these data points of what it means to have traction already um baked in and being looked at for Grant evaluations um and so I how I Vision it and even as a Founder like one of the biggest time kills whether or like raising grants or even traditional funding is like closing money and so like that's mainly what I want to like eliminate is like a lot of times you know one founder will F be

focused on the product and engineering and another founder will be focused on just like the growth and fundraising um and and this could take like fundraising could take anywhere from one to six months even grants in that in those processes could take um that long so I really kind of En Envision a world where like the team SP the team size become smaller and there's just less distraction and deeper work and even now like if you look at most startup Founders today like AI is integrated into the work flows so I really see like not only this Marketplace emerging but this Marketplace being of Builders who are a leaner more uh product oriented teams um that that honestly are are way more cost- effective um than what we're used to yeah totally totally no when you

look at kind of like what the near future looks like for us right like how soon do you start to see these systems in a place that we're actually interacting with them right you kind of talk about like Dem aware and I in my mind when I hear you say that it's like the blue ocean planning phase like what's possible how do we get just kind of like MVPs that show what's possible but maybe we can't actually interact with them we just kind of validate is this the concept that you know could really work how soon until we start interacting with these systems like what does that look like I mean people are already using these like I use AI for for everything like I I just went through a a serious long iterations of contract negotiations and I've never

been more satisfied with the outcome of the negotiations and AI helped me in every part of the way like in terms of Engineering in terms of design like AI has been I there is not one workflow or process or department or contributor that does not use AI for their everyday work um already um so it's here um it's just a matter of how is the way that we're interfacing it and how it's pipelining and integrated into our system will change but there is still a lot of limitations like in terms of the infrastructure um there is still limitations on standards for multi-agent Frameworks and how you connect things together there's still limitations on um evaluation Frameworks or how um or how you kind of determine whether the agent

did something correctly or like observability like how do you know the steps um or like why the agent did what it did and then this is resulting in a lot of hallucination and things that might not necessarily work for Enterprise or at scale um and so it it's still like is an assistant um at at some sorts but I see a lot like there's been so much Innovations new foundational models happening every other couple of weeks I see the systems today but actually building uh a lot of these Primitives so they function autonomously with kind of like a safe and secure way they interface with smart contracts and wallets um I I see that being at least um a couple years till we see like an actual like something that is production

ready but we're in web 3 and we don't necessarily uh we have a little bit lower standards of what we're willing to uh experiment with so and we can always experiment uh with less and run a lot of pilot programs so I don't see us being replaced by completely by AI in the next couple years to run these funding programs but I do see it like literally you guys getcoins using it like I I think every Grand operator uses AI in some capacity totally yeah so you touch on a really important point there which is like the speed that the the like the speed at which this progress is happening with these models right something that I personally seen at

gitcoin is like how do you design around that so like let's use Checker as an example when that was built we were tapping into I think it was like a you know open AI 3 version 35 you know the the previous version for Checker and at the time it was great but now we have 40 we have you know Claude Sonet right like there's there's these new models and then you see llama progressing you know so much quicker so like there's all these models that are progressing so quickly how do you build like what are the design considerations around building these tools building these agents to make sure it's done in a way that can benefit from that rapid pace of innovation right are these agents ones

that can like upgrade over time is there a refactoring that has to happen as new models come out like how do you think about that yeah I think we definitely have to be um adaptive towards changing our stack uh in any capacity but um honestly a lot of a lot of these llms uh like you can easily replace llama with open Ai and they even have uh Frameworks um and tools that that aggregate a lot of these um in the decentralized AI space I think hyperbolic they they just they just raised around 7 million um they're they're one of the uh AI projects in our our past cohort at near um and and they they have llama and they have a bunch of people kind of using

them as inference and kind of aggregating other llm Solutions so I think there'll be a bunch of providers to make it as easy as possible to choose uh the most performant framework um and um this will kind of be integrated into multi-agent Frameworks but at the same time like it is rapidly changing and not only are there like new foundational models but a lot of these foundational models are better at one thing than um something else and this is this has just been emerging every month so I honest like my personal opinion about everything and even when it comes to AI even when it comes to blockchains like the tech that we're using right now that we think is King made won't be the tech that we're using in five years um and it's probably gonna be something else so we have to just be in that founder mode

of constantly shipping and constantly iterating on new tech um and really keeping up to date and and that's anyone who is in the AI game like most of their time is spent on research papers like I am more of like application Level product crypto person um and so there's so much there's so much happening um and it's it's faster than crypto and and everyone knows about just the information fatigue that um cry crypto people have of just all the new Tex St so and this is kind of an order of magnitude um for AI uh but the lucky thing about AI is um it's AI so a lot of this information this tooling is integrated into a lot of the like in in terms of lot the AI tools so it's I think the it's easier to learn

now than it ever was before and I kind of seen this too I I've taken like a break from engineering but like because of AI lately more now more than ever I've been coding every day and and normally where I would have a bug and it would frustrate me it's been pretty streamlined to to ship whole MVPs to like literally go from an idea to be like hey AI give me an interface okay put it into another uh another llm and be like hey evaluate this interface whether it's legit okay now write a smart contract from that okay now evaluate this for reentrance to text okay now create um a boiler plate next toap okay maybe take that into cursor maybe I'll take that into v0 I'll show that to my designer uh my designer might take that directly put it in there or he

might uh use uh HTML design to bring it to figma and he might export it through locy and then to a next GS app and then we take it back to cursor and then we're able to iterate and and go on that um and so like from every literally every part of the process from like requirements Gathering to all the way um to to the code like is like AI is is part of that process and so I think the learning and the new technology will be also integrated into a lot of this AI developer tooling is what I'm hoping uh but um yeah I think I'm excited to to learn a lot of these new tech stacks and I think in general for people who are designing um just any new system uh making it backwards compatible is a good uh design decision but if it really does

provide value like Builders will chew glass and we we see this constantly with um the crypto ecosystem so totally yeah you know it makes me think of this word we use a lot you know kind of like the public goods getcoin side of things plurality right like a plurality you know of mechanisms and you have like a plurality of options here you know one thing that I see happening is like today it is primarily clawed and it is primarily GPT 40 that I'm interacting with right but I would love a world where it's like I'm interacting with like maybe one interface but I'm getting the benefit of many different models on the back end right giving me um you know responses back and so when I look at Checker you know using that example

again really what Checker provides today is kind of like a frontend intake filter right that says does this meet the basic criteria of the application you know what's a score that I can assess with that where I'd love to see that going right and some we've talked about internally is like what if there's like a modular way that you can switch out the models or have multiple models and you create like an AI Grant Council of sorts right where it's like One reviewer could be the you know Claude one or anthropic One reviewer could be open AI another one could use you know the Llama models and those things can be kind of like interchanged over time and then you have some sort of layer on top of that

to review over time you know the the accuracy of those reviewers the impact on the grants that you know they provided Etc right and so maybe talk a little bit about you know your vision that way like you're obviously out there kind of like working in this space a lot do you view this as something that the models get abstracted further back at some point and we get to a place that we're interacting with many of them and like that's producing even better results for us or like how do you see that yeah I really love the uh kind of AI delegate example and and and if just kind of a case study um on top of that and if you kind of want to see like the early case studies you AIP gf.com exlore kind of anyone not even like a grantee

anyone who's building in this space kind of compiled those resources but I see um like so what blockfi did is they actually have an active uh AI delegate voting in the enf St right now um it's it's live um and so um I see um and this is something we've talked about because um on the near perspective we're actually uh switching into like a house of stake so completely kind of stake based governance mechanism for a lot um of of our ecosystem and a lot of the the protocol rewards will be directed towards um the house of stake Treasury and one of the things that we've talked about and we've had this issue especially with our last governance experience is a lot of the subject matter experts who are um who who are the only ones that can actually evaluate

these proposals are either too busy shipping or building and it comes to like where they're not even giving a quality uh feedback so a lot of times I think that even now like AI trained on its current data from a subject matter expert um may be more useful than a subject matter expert who's not entirely uh passionate about the work he's doing um and so yeah I I definitely see people like interacting with different agents but I also see agents like I'm an agent you're an agent um uh we don't necessarily know what backend is powering each of us but I see a lot of people that we're interacting with uh being agents and um there to be a point where we might not necessarily know

whether it's a human or whether it's a robot and so that's also something that could be a whole startup idea and I know they're working on uh fingerprinting and things like that but yeah did that answer your question yeah totally man no so I guess like as we we look to kind of like close out the episode here right I mean we've talked about a lot here you know I think the biggest takeaway is like we're so early you know um in terms of like this design space and what's happening and you don't see the future anytime soon as like replacing people but it it kind of gets to like you know there was an article I read some time back about chess players you know and how it was really when you paired a human with you know Watson or whatever

the the IBM system was at that time that it became much more powerful than just Watson or the human could be on its own right so I think there's a lot of fears around you know is AI going to replace us and like I don't know like maybe at some point right but I I see in the the near future a world where it really kind of enables you know and and it enables in a way that if you're highly specialized in a certain area you can go very deep and it supports you in that right so you know you give the example of I'm someone with high context on a very you know Niche specialization and my time is Super valuable and I'm usually building towards that specialization but I'm also needed to review others and kind of like grow that

discipline and I I can't do both well um and more than likely if I'm a builder I'd rather be building than reviewing someone else's right but there could be a world where an agent could be trained right um and you could even potentially create markets for that I could see right where it's like if I have very specific domain expertise and I have a highly trained agent that knows everything I've done over time the reviews I've given my research it has that context that I fine-tuned myself that agent could could ultimately be more effective than me in terms of those types of tasks that you know I'm not excited about but it could help others with right and so you see this world of like many different agents those agents

you know talking to each other and collaborating you know that being something that's kind of like enabling um you know growth in ecosystems and so um maybe to like close out right the the episode here give us more of like the long-term Vision right like if if you had to look when we say longterm let's say five years which that is long-term in crypto perspective right you know five years down the road what would be an ideal state for you you know in terms of like the work that you're doing today um where do you think that could go if if you know we perform well uh yeah I'm trying to honestly replace my job like I don't I don't like Grant operating uh I don't like certain admin aspects I I look at a world where if I'm very successful um I'll just you know I'll

have no the problem will be solved and I will handle it and uh and I can just worry about Vibes and wholesome activities and uh yeah that's that's that's what I'm trying to do I think work is really overrated and we're going to really kind of redefine what it means to do things in society and how we actually uh give back and a lot of acceptance on things that we valued as valuable work as honestly a lot of busy work and really uh spend spending more time on what matters in the world and actually making direct impact um but yeah that's yeah I'm actually trying to replace my job like

um and uh they say AI won't replace your job but um I'm not going to lie like because of AI we've gotten a little bit leaner and and there there there have been people that have been replaced already so I know that's not like oh yeah they're taking our jobs but um yeah our jobs are kind of whack dude I love that answer like I was not expecting you to take it there but like I love that answer and it honestly reminds me a little bit of a previous episode we did that will you know probably air before this one where I met with uh Skelly from Wonderland and then Alp from inverter right and that episode was was really meant to talk about novel funding mechanisms and what the design space is there but near the end of that episode

it evolved into something I didn't expect but it was like one of my favorite parts of this season right where both of those guys really talked about the instate you just said we're like does this actually get us back to a place that we're we're contributing towards the common good again right I think that there is this this has happened in our world in my opinion over the last few decades where like everyone's just trying to make it and it's getting harder and harder to do that and in in a world not too long ago there was more of like a focus on a common good right we're kind of like Humanity as a whole and I think that as consumerism and money and like you know all these things took Center Stage we've

lost track of that over time right and what I just heard you say is like maybe what happens here is this wave of inovation this enablement this thing these things that AI is really starting to drive for us maybe like a best case scenario would be it can actually create more time for us to get back to a common good and solve some of the real problems in this world that I think are have persisted for a long time you know and I'm not really sure if we've made a lot of progress towards those um I talked in that episode too that conversation reminded me of the vitalik Tweet right and so there's a tweet that viik put out back in 2016 2017 I think that was like it's when the market cap of crypto hit half a

trillion right and he basically said have we earned it and he listed out all of these things that he saw at the time as the potential for what we could really accomplish and go back and look at that tweet like those things are all still very much true the market cap is now 4X that sometimes less than that depending on the day right but we haven't made any progress really on many of those things other than TBL you know and it's like suspect if that's even a real metric we want to pay attention to sometimes you know and so I love that answer man I I agree I think that um we need to look at the common good you know how we get back to that and maybe the

best case scenario out of all of this is you know AI lets us get to that so the plug it was amazing having you on the show um everyone will be able to look in the show notes and see the info to how to find your profile and all the stuff you're working on um definitely stay tuned uh there's going to be much more coming down the pipe uh between plug and myself and some of the work we'll be doing at gitcoin uh and on top of Alo so I'm excited to see that happen so thanks for joining today man thanks for having me I really enjoyed it [Music] he