[REDACTED] 2024 | Democratizing Defi With Muti Chain Agents

Captions as YouTube publishes them, usually machine-generated, so product and people names can be misheard.

1. 0:05

   so uh super excited today to talk about the future of defi agents uh defi is incredibly complex you borderline need a background of Finance to make any money on it and not get front runed by hedge funds and so we think that uh AI agents are going to really democratize access to defi so we have the perfect guest today and uh just to kick things off we're going to just go quickly in order introduce yourselves tell us a little

2. 0:28

   bit about the project you're working on and uh why you started the project start with zero okay uh I'm zero uh I'm the co-founder of the DAT tra and the tra is a decentralized trading boat and uh um yeah uh help you to get the the profits yeah great it's great yeah hey uh I'm David uh co-founder of Valerie we co-contributor to Olas allas um very much subscribes to what Cameron just said we want to give everyone

3. 0:59

   autonom AI agent so that people can Outsource goals to agents they own um and have them deal with the complexities of things like Defi and other use cases Lucas I'm one of the co-founders of the company called Almanac we've started Almanac because first we saw originally we saw how inefficient defi was and how it could use some data driven intelligence how those could improve the decisions being made within

4. 1:32

   Defi and then at one point we've realized that this is a perfect input for agents which at that time became a vehicle an idea for a vehicle that can help us navigate proud defi because dii in in our opinion grew too complex over time it became fragmented and became chaotic and at this point uh we think that perhaps defi was never meant for humans it was meant for agents all along totally and um I think this is how you know people this is why people get into

5. 2:08

   cryptos honestly to make money and so starting with zero he tell us just a little bit more about Delta trade and the three strategies that you currently operate and you know what about them are sort of well positioned for agents to do yeah uh currently dat tra have complet the uh three three VT uh we call that GRE swing and DC and DC is very uh simple is is uh let uh uh let the user to uh fix the interverse to uh seal and

6. 2:37

   buy your uh assets yeah it's very simple and but the grid and swing is very uh complex um yeah um um it's I I I I kind to explain it in one line yeah yeah who here just as raise the hands like knows what dollar cost averaging is like when you're trading okay it's a it's a good amount but like the vast majority of people don't really understand this and so out of these three strategies which

7. 3:00

   is the most popular on the platform yeah um yeah us is very very complex to understand what is grid but the grid and swings really can help user to get more and more profits from the uh from the market especially B market uh uh maybe you you can choose say our website website dat. a yeah um um we we hope that um um let let the user to use dat Tre for a EAS way so uh what we want to

8. 3:32

   do in the future is integrate the AI agent let the AI agent to uh to let the user to use the AI power strategy so user just want to uh uh just want to say the high apy and choose choose API and choose Prov your asss and then one click to create your BS then waiting for the profit awesome and uh David you know Olas is there's a lot going on in the Olas ecosystem right now you're building across many chains uh you have a pretty amazing prediction Market product uh can

9. 4:05

   you tell everyone a bit more about the OLS framework and exactly what is you know being most adopted today yeah um so what's the goal we want to have you operate an agent that you can Outsource one day uh arbitr complex goals to and then that um attain that for you today they very like specific domains in which the agents are act uh active in so the most dominant use case is prediction markets you can try this

10. 4:34

    out yourself today um on ol. network operate you can download an app called Pearl and then run your agent locally which participates on your behalf in prediction markets 247 um we want to bring that same con uh concept to defi uh and in a in a cross-chain context where you give the agent higher level goals um Express the natural language and then leave it uh to the agents over long

11. 5:04

    periods of time to execute those strategies why do we why do we think that's important because to what Lucas was saying there's very few people who will bother engaging with the complexity of Def Defi and if you really want to grow um end user adoption I think you know user owned agent is One path the other one is to basically provide some very centralized um Robin Hood like like experience and so we're very bullish on on the agent uh

12. 5:36

    piece and like there's now daily active agents are going up every day um yeah I would leave it at that that now and uh Lucas your current Target customers a lot of Institutions is that accurate and so but you guys are going to slowly start opening up more to to retail can you tell us a bit more about kind of how the institutions are approaching one onchain trading at all and sort of what some of their uh I

13. 6:01

    guess prerequisites are to start participating yeah so the reason we started with with professionals and with institutions is that a lot of the um a lot of the Alx value proposition comes from the optimization we're offering the the fact that you can not only create an agent with us it's about the um the we call it strategy optimization Suite we offer and in that piece we allow the um we Fork essentially the the blockchain uh we create a testing environment where we then use

14. 6:39

    agent-based modeling to mimic behavior of users and actors on chain and this is a technology that's well a version of been has has been used in try for for some time exactly by hedge funds by professionals so people and institutions that were familiar with this they were our first Target for obviously reasons they understand the value position and they understood that within crypto

15. 7:02

    within defi we don't have a tooling like this yet that would allow to optimize your strategies that will allow to improve that decision um decision intelligence so this was this was a low hanging fruit and we also we also learned from them we also saw how they want to how they are uh seeing the future of um let's say asset onchain Asset Management in crypto uh so for example the way almanak works is that we

16. 7:28

    are uh non-custodial software right so ultimately it's the the customer the institutional customer that retains the full custody of their asset and has to sign their own transaction precisely this is important and this is something that was a bit of Novel so we were I would say at the at the discovery stage with all the institutions but we we tailored the product towards that direction and I think that um well

17. 7:52

    retail or like everyday users in that regard it will follow Sim similar trajectory because of trust so instead of giving all your net worth to One agent overnight you will be gradually giving it permissions to do certain things for example execute a strategy you figured out yourself also with the help of an of of AI because I think this is also important is that ultimately uh we want users to formulate their own objectives high level strategies right they ultimately won't have to you know

18. 8:26

    know all the details but they need to understand the concept of risk yeah but you know people are really bad at knowing what they want generally and so like when we want to go talk to you know make some money and work with somebody to make money like a like financial adviser it's generally a conversation yeah so kind how do you see the interface between you know these these users who really aren't that well aware of different strategies and these agents that are either you know executing many

19. 8:51

    different strategies it kind of feels like it needs to be a conversation so how do you actually give them that that interface where it kind of leads them to the thing that they need I think this is the easy part honestly because it has been solved like for us like the the llm be um adoption and like the experience like the popularity it's it's it's crazy even across web too like how many of the

20. 9:17

    apps across pretty much every vertical has its own llm now like this becomes something that users get get accustomed to and I think over time it will be easier and easier for businesses to adopt a version of it to improve that interaction with the customer I think that would be that would be the experience you think we might need a different modality though like messaging like chat is is difficult for many yeah um so yeah David what do you I guess see the the sort of future of AI agents and

21. 9:42

    how users actually interact with them it's already multimodal I I think there's uh examples of autonomous AI agents that launch meme coins and have like a Persona which be might be streaming on Twitch like this project do that um I think coming back to your um uh Point um like chat will I I I doubt chat will go away but I I do think we need like as as a group of technologists we're very captured in this idea of interactive experiences because that's what we all grew up with like mobile apps where you

22. 10:20

    click around do things the beauty with agents is that you can abstract a lot of this so whilst chat is important like being able to to maybe describe a strategy I don't think the like large group of users will spend days and weeks interacting with the financial advisor they'll be okay like giving it some sort of high level goals um and then check back in with it maybe after a couple weeks see how it's done maybe do some adjustments and that's why like user

23. 10:50

    ownness be on a financial domain or be when you one day have your personal health adviser who is like trained on your needs mhm your data is is such a different paradigma of a product so chat will be important I'm sure one day Bo like um obviously in the financial context graphs right like it it's sort of generating graphs around things it's been doing rather than those being hardcoded from the get-go so but you know Alpha is only is the whole point of alpha is kind of that other people don't

24. 11:22

    have it and so I feel like these AI agents could very well democratize it too well and then the alpha sort of goes away over time I'm curious if you think that this is actually true and if not uh what are some examples that uh of strategies that almost won't lose Alpha over time so today it's not true and to one day maybe it might be true but today it's certainly not true so we have for instance in all us these um liquidity pools which are on various different

25. 11:55

    chains um you'd think that they'll always be perfectly in sync because of Arbitrage and like people assume Arbitrage but it doesn't like actually someone needs to do it and that actually applies for a lot of like inefficiencies across Defi and so obviously if you're adding let's say an order of magnitude more users we're adding an order of magnitudes more liquidity and it's not

26. 12:19

    automatic that all inefficiencies are immediately gone away I think there's also a lot of like local contexts like not just in defi but like generally speaking like if you think about AI adoption like there's like local context to a lot of decision making which some of it will go away but I think some of it will will be retained and also like Alpha exists on different chains in different regards and so think the

27. 12:43

    agents aren't going to care what blockchain it's using it's just going to go where users and liquidity are and so zero I know you're relatively multi-chain um do you see any sort of difference between uh the users on the salon implementation versus the near implementation yeah uh B is the Mar chain and yeah you know currently new chain has the chain signatures and or chain obstructions and we have many

28. 13:06

    product have released or want want to release so we G Tre also will use the chain techn to support M chain that is a user to uh like to use our om Omi bridge to bridge your asset and by one click into the near chance and at the same time can create our data Trace B like this say or grid and uh when the user close click the close and and the asset will uh return back to the to origin so you're saying I will be able to DCA out of my Bitcoin position using one of your

29. 13:38

    agents yeah right that's awesome yeah that's fantastic um yeah I mean David I know you're also you have a couple of different multi-chain implementations uh do you see any sort of different user Behavior or you know where the liquidity is or Alpha is across these different chains yeah I mean we're still relatively speaking at the beginning of this journey but all other is across sh protocol uh we're on uh eight evm chains

30. 14:02

    Solana and we're coming to near and it's super exciting to be able to for instance have agents like have native access to bitcoin um for defi um I think again like I would be careful to project uh too far with the sort of all the alpha will go away just when we come quickly back to this one it's also liquidity so if you for instance take the prediction markets product we have um like end users put up like fund their agent um to trade on their behalf and so

31. 14:42

    they provided that liquidity and then that generates these new markets I could very much foresee a world where we have entire like markets that humans don't even look at maybe not even know about where agents trade with each other creating entirely new Financial perimeters and like um you know um Me Def agct with each other trading agents aren't new like hedge funds they're not new building an ml forever but I think the generative part of it is not really I mean occasionally you can you know have people hard code

32. 15:13

    scripts to kind of deploy new kind of things and have like Bots do that but in the sense that we really have like almost like entire Virtual Worlds which maybe for months humans don't even know about but agents are actually leveraging I think that's a real possibility I mean it will happen it's they're they're not new but what we're I think what we we are yet to see is how they will leverage the potential of of interoperability and

33. 15:44

    composability of defi because you know regardless of how how smart or fast an agent in trafi can be it is still operating in a highly silos environment it's technically silos not to mention like you know barriers Etc which prevent the uh the driving actors behind an agent from doing something because of like you know constraints Etc in in defi we we don't have that so which again reinforces this pieces that this this is just a perfect environment for them to

34. 16:16

    for them to roam I very much agree with David here I think this these these uh fears that Alpha will be gone is they we're we're very far away from that there's always enough Alpha it it exists yes there there's there's there's enough Alpha and I think once we once it becomes easier and easier over time and I think with all those Technologies it will happen to include you know agent agent agents toolings into your

35. 16:40

    portfolio management the amount of opportunities that will come out with that will dwarf your anxiety over losing like a bit of alpha here and there so we are at the at the really at the beginning of of the journey and agents will be the vehicles that will carry that so it's an extremely deciding time um yeah so I'm really thinking about like what does it actually look like or

36. 17:04

    how what is it going to take for users to actually get these user these agents in their hands and I do think that wallets are the sort of interface of crypto and I think that uh wallets all wallets will have some sort of like llm tab or you know chat function where you're going to be able to call all these agents and interact and discover but the my concern is around the security of these agents maybe calling contracts that haven't been audited yet

37. 17:28

    or you know there needs some level of Guard railing and so um how do you think agents but on the flip side agents can also provide an amazing uh opportunity to get users to better understand risk um so I guess where do you see this sort of like intersection of Agents providing uh users a better understanding of risk and also uh uh like giving them access to things that they didn't know before yeah I mean maybe I jump yeah uh

38. 17:58

    I I think crucially like humans don't have the capacity to do even more of that right so if you're not talking about adding agents to The Mex if you then also assume that the human now has to also like educate themselves about what all the agents do it it's not feasible because the agent is even faster than what you would have done yourself in Defi and like generally speaking defi today speaks to Teeny total addressable

39. 18:23

    Market whereas what we want to do with the agent basically should speak to everyone so I think there's the um huge need for agents to take on these roles as well right um for agents to be risk managing agents for agents to provide signal to other agents and that's sort of really at the core of how we think about it with multi-agent systems and allas that makes sense I would add to that that in in case an agent is uh

40. 18:49

    managing something more than your own money and it's responsible for something more than yourself there is a perfectly feasible concept that we're going to apply here and that's just underwriting an agent simple as that staking against an agent so put your money where your mouth is where you put your money where where your agent is so it's a great note to end on uh put your money where your agent is uh we have about 10 seconds left is there anything else you guys want to add no no great well thank you for

41. 19:21

    joining yeah check out their websites uh they're all on Twitter and have a beautiful day thank you