Autonomous Trading with the Mindshare Index AI Agent - powered by Shade Agents

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

1. 0:03

   GMG AGN I am coming in with some absolutely incredibly hot news and it is that there is a 115% tariff on every single AI agent transaction across the board on the NEO blockchain and I'm bringing with me MJ and Hassel to tell you all about it and to open the lid explosively on what is nothing less than a remarkable technological feet with the latest in shared agents and uh onchain autonomy. Hassel MJ, how you doing?

2. 0:37

   How's it going? Are you in too much pain? I hope not, but I would love to know a little bit more about you. So, if you could uh take the mic, give us some introductions and then we will dive into what the uh the hottest news in the world of shared agents is. Yeah, GMGN.

3. 0:52

   Thank you, Demo. Thank you for having us. Hey everyone, this is MJ from Yonder Labs and today I'm joined by my co-orker Hassel and we are going to be talking about a new AI agent that will be launching today. Hi to everyone. Nice nice to be here and I think that it's going a great day to present this for you.

4. 1:14

   AJ, if you want to start it's okay because it's okay if you want to start to talk about the agent. Yeah, 100%. Anything that distracts us from uh whatever is going on in the world right now, I think is is a huge win. Um MJ Hassel, we've already discussed it behind the scenes, so I am lucky enough to be privy uh to some of the magic that we are about to see happen today, but MJ, why don't you give

5. 1:39

   us a brief overview of what it is we're about to reveal to the world and just how exciting it is? Yeah, for sure. So let me quickly share my screen. Um so yeah so um everyone um super excited to be talking about this mindshare index agent today. Um so yeah so this is basically an agent that we're going to be launching today. Hopefully during this session in real time and hopefully we'll also see this agent make some trades uh in real time. So the plan for today is that I'm going to quickly

6. 2:14

   go through a few slides. I will talk about you know what is this agent and then I will hand it over to Hassel to talk more details about how we built it how it works and if you are watching this you know um stream uh stay stay tuned because at the very end we'll be launching this agent in real time uh hopefully everything works all right let's get started so what is this uh man

7. 2:36

   share index agent right so this is a fully autonomous AI trading agent it's powered by NIA's shade agent stack I I'll be talking about it in a minute. So the TLDDL here is that this is an agent that can decide what crypto to buy and sell based on the amount of attention that they are getting online. So how does this work from the very high level right? So basically the agent consumes

8. 3:01

   data from AITO which is a platform that aggregates you know mind share and sentiment data across multiple platforms like uh you know Twitter, news websites and so forth and then the agent will be leveraging a large language model to analyze the trends and sentiment right and it will come up with you know some uh actions to take like buy and sell some tokens. So the agent will uh you know sign and submit these transactions using near's chain signatures and near

9. 3:30

   intents which actually provide you know a seamless you know probably the easiest you know multi-chain or crosschain trading experience that enables the agent to you know buy and sell and hold assets on different blockchains. So as I mentioned uh this agent is powered by Near's shade agents.

10. 3:49

    So just in case if you're not familiar with that this is how the architecture looks like. So um you can see that there were two key components here. The first one is the worker agent. So this is basically you know an agent uh that runs know this basically runs the agents uh you know logic. It runs inside a trusted execution environment which is very important. I'll talk about it in a minute. And then this worker agent has

11. 4:13

    access to offchain APIs, compute and storage such as you know the kit API and worker agent will also be using a large language model or any sort of model. Um this is basically where the decision making process is resolved and after that the agent will come up with some actions to take right. So the second piece here which is very important is the smart contract on near. This is very

12. 4:38

    important because the smart contract can do a few things. So first of all the smart contract will be know verifying and registering the worker agents. So what we what do we mean by that is that every time a worker agent instance spins up it will first need to register itself with the smart contract. So this actually uses uh a primitive of the trusted execution environment called attestation quote. Uh so basically uh the environment here can uh produce a

13. 5:10

    report that uh show the current state uh of the environment right. So uh and this report is also you know signed by a key and attested to by the manufacturer of this environment. So in this case this could be you know Intel and Intel can say that oh you know we we recognize that this is one of the CPUs that we sold and uh um the environment seems intact right so when a smart contract receives the registration call from the

14. 5:40

    worker agent it will first verify the report right you know the signatures and stuff and after that the smart contract will also extract the code hash from this port which is the code hash of the software that is currently running inside this environment. Right?

15. 5:56

    Basically the smart contract also needs to check whether the worker agent is running a trusted copy of the software right and after that the smart contract will register this worker agent uh and worker agent will be you know be able to call other methods in the smart contract. So what are some of the other you know methods that the worker agent can call right? So the smart contract is

16. 6:19

    also integrated with Neo's chain signature service which basically gives you know the smart contract access to a set of keys that it can use to interact with other blockchains right so whenever the worker agent you know wants to take an action or make a trade it will send uh the payload to the smart contract right and this is when the smart contract can check whether these proposed you know transactions make sense right based on certain rules that are defined in the smart contract

17. 6:47

    contract and if so then a smart contract will call ch signature service and to get the payload signed and and return it back to the work agent. So this is how uh you know the flow works and if you think about it right this is actually very powerful because first of all these work agents they are running you know confidentially inside uh you know these trusted execution environments which means that nobody can extract you know

18. 7:11

    data from uh these work agents because you know in many times they're actually dealing with you know private data from the users right and secondh if you think about it um you know because of this design anyone can actually run an you know run an instance of the work agent and contribute to the system, right? Um, as long as they can prove that, you know, the work agent is running inside a trusted environment and it's running a

19. 7:37

    trusted copy of the software. And, uh, you know, if one of the, you know, worker agents, you know, dies for whatever reasons, you know, anyone can quickly spin up another instance and pick it up. And last but not the least, because uh, the system is integrated with chain signatures. So this means that this agent can actually you know interact with you know any blockchain out there right they can buy and sell and hold tokens uh you know on

20. 8:01

    any chain and a chance signature service currently supports all the ECD SA ch which includes you know bitcoin ethereum base and so forth and once it adds uh support for ed25519 uh chance it will basically support every single blockchain out there. So this is shared agents in a nutshell and we believe that this is you know the first tax stack that enables developers to build you know fully autonomous agents and yeah we we also believe that in the in the near future

21. 8:31

    uh pun intended we will see developers building all sorts of you know autonomous agents uh you know to custody user funds or uh you know to uh you know replace you know humane the loop in our defi infrastructure. So just to quickly wrap it up, uh you know why does this matter, right? Um so the mind share agent that we are launching today is a great example that shows how you can

22. 8:55

    leverage the shade agent text stack to build fully autonomous agents that are also resilient because there's no single points of failure, right? And also uh last but not the least uh this is a great example showing you how you can um you know build a fully autonomous uh agent agent uh that can proactively react to you know events or to data you know in real time right instead of just

23. 9:20

    you know passively you know responding to users inputs. So yeah so if you want to learn more about the shade agent stack here's a link that you can you can check it out. Uh and without further ado, I'm gonna hand it over to Hassel.

24. 9:36

    Well, hi again. So I want to focus in the main features of the agent and how was developed and construct all the step by step to work. Let me share my screen. Uh the first and many important thing about this is that the agent is running inside a that it secures all the execution all the agent execution and the first thing that we work is to take all this the information with FA. So using FA we initialize all the values to

25. 10:12

    register the agent to a contract. This contract is uh created with all the necessary methods to validate that this h code that is received is valid and also when you create the de you need to assign an account. This account is an ephemeral account that every time that the is running we generate some entropy to create this account and we need to found every time. So once this is done uh we

26. 10:43

    obtain near intense balance to know which trace we are going to execute and to know the w that the traits that the near intense are also we use kit to obtain the manual here I want to talk about what is mana but because I think that is something that may be confused man is a value that represents dimension shelf of a token in Twitter. For example, if today some token is really popular, the this value

27. 11:16

    is calculated according to other crypto market assets and we obtain this value. is not uh like sentiment or something like that. It's more related to the people talking about something some asset some some token because it's important to use it to calculate it this value and then the LLN use this value to suggest some traits. Once we have the trades, we create a quote. This quote it's use it for the intense but the advantage part

28. 11:52

    here is that the quote is signed in the NPC service. So how it works? We get the quote calculate a payload and then we call the sign method with a key derivation and a payload. Once we have that we ask for a sign from NPC service and then return this signature to the agent. Once we have this signature and the original quote then we create a commitment and this commitment is published to the solver box. So it's like different technologies that are

29. 12:26

    involved to create a secure environment from TE from uh using FALA intends to obtain the traits and the commitment with the solver boss and KTO to obtain this manual value. So once the intent is published we can the intent contract execute the trades and the user can see the the result in the app. So if you are prepared I can start to show you how this works.

30. 13:04

    So we had FA and we start the AJ. Let's wait just a few seconds because FA is needs to calculate all the values. By the way, just want to mention that the $10,000 here is real money. Yeah. And if all goes well, it might be a little bit more than $10,000 tomorrow or Yeah. If we double the tariffs, it might be a little bit less. Let's see. Yeah. So, the first thing that we have is obtain the letter

31. 13:54

    to the agent. Lenwood found it. There you are. The agent got the fun and start the agent execution. use the KTO API info and the asset balances that we are we got right here. Now, let's wait just a second to see the price.

32. 14:56

    or maybe we can we can talk through through what it's actually doing, right? So, it's fetching from the Kaio API and it's saying, "Hey, Kaio API, knock. What's the what's the hottest coin in the world right now? What are people talking about?" And then the Kaido API say, "Hey, hey, agent. The uh the hottest coin in the world right now is Fcoin." Oh, it's Trump.

33. 15:23

    So, what has it done now? It's It swapped five USDC for Trump, right? Yes. ask for the NPC signature to the NPC service. Then that signature go back to the agent and then we publish that in that intent to the solver bus and then we can see how this change to the new balance asset. I see the agent is very conservative only $5. Yeah, today is sad because sometimes it's like three different assets or more.

34. 15:55

    Yeah, I think honestly is probably a good time to be risk off. So, so well done agent for uh for managing your risk tolerance there. But that was incredible guys. You saw it uh you saw it here first and you felt it here first. Um yeah, that's a good idea actually. Maybe if we have like a DGEN slider and we can just set it up to 100 and it can just all in on Fcoin with the thesis that hot

35. 16:20

    air rises or something like that. Um, but anyway, that was that was fabulous. So, that you're gonna keep this running and it's going to continue to monitor Kaio uh as we sleep, as we do our daily lives and it's going to continue making those trades, right? Yeah. We configurate the agent to run daily. So, tomorrow maybe to 11:00 a.m. we are going to obtain new trade decisions.

36. 16:46

    with you my share val my share values to the assets for that day and to obtain new reps maybe change strong for anything else the asset that we are set at this moment are in the repo if you want to check it is uh available in the general labs u GitHub but the one that we have let me show you at this moment is USDC near it BTC and well track and XRP but the the developers can add uh the one that are

37. 17:23

    allowed with intense so that's the only limit that the tokens that you can assign here is the one that intense allowed yeah you can also tweak the prompt to basically uh let the agent to you know discover tokens to trade by itself instead of providing like a list of tokens so the good news here is that everything that we just showed is open sourced. So if you want to fine-tune it or tweak it, if you want to add like a DJ mode, you know, uh it's definitely doable.

38. 17:54

    Everything's just out there. Yeah, the the agent is uh configurate to use only the the values that are in the white list, but you can ask for new tokens. Maybe not quick intense but you can obtain new new balances or new trades for the tokens if you configure the prompt that execute the the agent. This is the prompt.

39. 18:19

    Maybe we can read that prompt out but let us say analyze only the following tokens in the white list asset map for each suggested trade. Consider the exact amount must be less than the users balance. trade the token and there's the format. I wonder how flexible is that? So, could I fork this repo, edit it and say um trade only these tokens but trade it with complete aggression. If you have slight conviction, I want you to go all in and if you have slight conviction the opposite way, I want you to go all out.

40. 18:56

    How flexible is that from uh from somebody who wants to develop their own version of the mind share agent? Sure. And you also can change the the ln that we are using at this moment. We configure it to use uh JAMA. But with your AI, you have the opportunity to use different uh agent sorry different LLN models. And this this will be maybe a different rate not not only one or maybe if you want to personalize also NA had the opportunity to develop your own LM.

41. 19:31

    So you can try it right here. So I have a question then. If if if we fought this like 10 times and we had them all running different LLMs, could we have a training battle of LLMs? Yes. There we go. That's that's the that's the that's the next thing on your bucket list, guys. And I just want to say before we wrap it up that this agent has already outraded me for the entire year. It's what? $5 to

42. 20:01

    Trump and it's worth $51 right now. So, good job. Uh uh autonomous shared agent. Um yeah, guys, fabulous as always. Is there anything else you want to let the people know? Jay, do you want to add something? Yes, want to um mention again that yeah, the repost have been published. Um I think it's included in the blog post um where we announced the launch of this agent.

43. 20:28

    feel free to fork it, tweak it and yeah, we're also looking forward to be sharing the performance of the agents, you know, in a day, a week, a month, maybe even a year and see how it goes. Yeah. And I also want to remark that this project use different technologies uh different new technologies. So it's important to get involved in all the stuff that near have for example using

44. 20:55

    aa to the agent near your intense to execute the trades fala to attend the execution environment and I think it's a cool if a cool stuff that we involve all these technologies and the developers can start to work with this I think that they can do different agents with different pron to obtaining different threats.

45. 21:23

    I think that's absolutely the world's your oyster. If you're if you're a developer, if you're a builder, if you're a coder, or even if you're just a vibe coder, right? Maybe one day uh they're they're going to be able to spoke this and and and play along with it and uh and see what you guys can build. I can't wait. I'm super excited.

46. 21:40

    Um nia.aishared if you want to learn more about the shared agent stack. If you want to learn more about the mind share agent that has just been so beautifully demoed by Asel and MJ there, you can go to neo.orgblog, look for the post that's entitled introducing the mind share index AI agent. Also, if you scroll down our Twitter, Facebook, LinkedIn, it will be there as well. Uh, Hassel MJ, anything else you want to tell the people? Do you want to give them some warm words? Do you want to tell them uh what the best type of tea is? Do you

47. 22:12

    want to tell them how you get to sleep at night? You want to tell them what type of car you drive, the name of your pet dog, your pet cat, anything you want. Uh let the people know before we wrap it up. Yeah, I just want to say that yeah, we are always here to help. Uh I believe if you go to that neato.ai/ Shady page, there's also a link to our telegram group. If you have if you have any

48. 22:32

    questions about how to develop, you know, uh you know, using the the text stack, feel free to, you know, to shoot your question there. Always happy to help. Yeah, me happy to help. Please don't hesitate to reach me if you have some question or something to worry or related to it.

49. 22:51

    Okay, brilliant. You heard it here first. Neo.ai/shared. Join the Telegram group. Talk to MJ. Talk to Hassel. And uh until next time, God bless. Bye-bye. Bye. Cheers. Thank you.