NEAR at ETHDenver: The Road to One Trillion Agents - Day Two

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

1. 0:05

   we're just getting one position up there you never break down oh just

2. 1:35

   all right good morning good afternoon good evening good night and that is only for the people who are tuning in uh on the live stream uh my name is David it's up there somewhere uh welcome to one trillion agents day to here at eth Denver now before we get into it I have actually had a lot of questions about this little character here and in case you um don't divulge in as much

3. 1:57

   psychedelics as I do um I'm going to explain the concept to you so this is Sam and he's Named Sam for no particular reason he is a shogo and he represents uh the amalgamation of centralized AI that is being built behind closed doors in San Francisco atel and it is up to us and the brilliant Builders uh in this room and Beyond uh to put the fight up against it and tame that very shogo and one of those Builders is with us here right now he's going to join us on stage he is the only Frenchman I like his name

4. 2:30

   is Pierre so please uh give a big round of applause for Pierre test yeah hi everyone thank you so much for coming today um so I'm Pier I work with neari as a software engineer and today I want to quickly uh show you how to build stuff on tees especially agents so I'm curious how many of you are Builders and have ever wor and uh how many of you have actually used some te stuff before okay so uh yeah first of all I'm going

5. 3:19

   just going to quickly go through where we come from and where we are going and then I'm going to do some demoing and more like life coding um the more the funnier part so um I systems right now are kind of fing us so all the answers that you get back from these centralized bodies are are you don't really know how it's generated it's uh they have a lot of ads embedded into them so they have a lot of biases and then whatever you put into

6. 3:46

   their centralized companies you don't actually know what they're doing with your data so it's like complete Black boxes and all this data that you share with them is a big security risk because like all this data is like very important and it contains a lot of like person to data and it's just waiting to be to be biged because hackers are working days and night to try to get hands on these data and S it to the highest beider but um yeah so and

7. 4:11

   another thing is like open soci has been fading so one of the examples is stability AI they were unable to pay for their GPU bill this last year and uh it's because you know everything AI cost is like super expensive and you have to find a good business model that works and all these problems with the centralized paries created like huge trust issues and um so a lot of countries restricted access to this centralized uh to these AI systems and as we know now like all these AI systems

8. 4:42

   are like very good for GDP growth and uh so restricting individuals or companies to access this uh um these Services is like very bad and um yeah so I just want to to say why it matters for all of us so right now we are sharing stuff with this um with these companies and like centralized parties that are very sensitive to us it could be like I don't know some therapy session that you have

9. 5:09

   with your AI agent or stuff like this and you don't want these conversations to end up uh with people that you don't actually know so you put a lot of trust on them and it's it's not very sustainable for the long term and um so to fix this we are going towards a user own AI approach and so basically the idea is that you give access to anyone uh to this a platform and you can actually verify and audit everything

10. 5:33

    that's happening there and so then you don't have to put Trust on the centralized parties you can actually like do the research on your on your on your own or even the community can actually check what's happening and um so yeah all this uh gives you access to your uh you get back the control of your over your data and if you decide to let's say um share some data you you get rewarded if you share with some some parties in the ecosystem you get

11. 5:58

    rewarded for sharing this data data and then like other things that you can do as a community you can like build a model and then whenever somebody's using this model you get rewarded for it it's not like just you just lose your data and uh um and so yeah if if you lose your data right now that to an AI and stuff like this you don't get anything out of it but so we are really going towards like a an AI is optimized for optimized for

12. 6:23

    user well-being and economic success for everyone I'm just going to Quick quickly go through the the stack that we have so so uh first of all we have the agent Cloud so basically the the place where you come and host any of your AI agents so it could be neari agents or it could be like Eliza agents L linkchain agents whatever and all these agents uh then like get access to a whole Marketplace you can actually interact with different

13. 6:47

    agents and um so whenever your agent is being used by other people in the ecosystem you get Revenue out of it because you can say okay my agent requires you to pay one usdc for the service I'm offering and then so that's how you get you generate Revenue um to create this agents we are offerring this neari agent SDK which basically the IDE is that you can write in python or typescript any kind of agents and uh it

14. 7:11

    could be like very complex agent to very simple one we're going to build a simple one afterwards um but yeah so to on top of this we use the near protocol blockchain to do all these microtransactions so whenever you know you have transact between two agents that requires payment and stuff like this this is all happening on here we have a bunch of payment channels and stuff like this and on top of it we also offer chain abstraction which is the

15. 7:35

    idea that you can transact on any blockchain um and so then like you can do stuff on sonana on bitcoin from your AI agents and so yeah so on one side we have uh agenting payment so it's what I what I mentioned uh agent Discovery so here we have like already 400 400 like agents so you have to have a discovery layer because uh the assistant so basically is the idea that you have you

16. 8:00

    speak to one particular agent and then so this a this assistant is not able to do everything so you need to discovers agents that are experts in specific fields and then it's going to ask this these sub agents please do this for me I'm attaching one usdc for for the service so this is the idea and then you have your of data and shed memory so this is where uh you want to really have a personalized assistant that knows you

17. 8:25

    and you can actually like trust it because you verify the whole stack and you know secure and we have agent hosting so it's basically where you send your agent and so all this runs inside a trust execution environments um so trusted execution environments is basically a a piece of Hardware that allows you to like you don't have to trust the hting provider that they're going to that they're not going to access your your data like there there is no way for them to actually access it and you can verify that they're uh that they're

18. 8:56

    running the appropriate software that you expect expect them to be running so so we have this for CPU based so it's typically like a agents in our case and also for gpus so we use gpus so GP is for any kind of model training model fine tuning and model inference and uh so all these we we run them on called SoCal CVM so confidential virtual machines so it's just it's just an abstraction on top of te's just a way for us to run to run anything inside

19. 9:25

    this T's so for our use case what does this te magic enables us to do so one very cool thing is fully encrypted models so basically you can train a model fully inside a te so you almost have over overhead so our friends at F did did a study and you we like a research and we see that the performance impact when you use these TS is from 8% to down to 1% for these bigger models because they spend more time Computing so the downside of encrypting everything

20. 9:54

    is is not that that that big um and then you have a monetized model um sorry monetized confidential inference so basically if you offer access to your Hardware to do any kind of computation and um computation for completions on these from these LMS you get rewarded as a service provider and then you have verifiable agents so you can as a user you can check that the agents being running in this T is actually what you expect it to be the version that you expect to be running and so you can

21. 10:24

    really fully verify everything so you don't have to trust anymore you can actually verify so the demo I'm going to go through is uh first I'm going to demonstrate a bit of private inference so it's just the idea that you can discuss with one of these llms in a fully uh private way so nobody can see what really happened um so for this we use a VM and uh so we have U we are working with Sal Network

22. 10:48

    on this and um we are working on this private ml SDK uh which is what what I'm going to be demoing today and on the other side we're going to be doing private a u a private engine so we're going to create an agent hosted inside te and then do some very simple completions with it just to show you how how it works so now comes the the fun part so I hope it's going to work let's see um so first of all the private inference uh side of things for

23. 11:19

    example here on inside the te we have a Lama 3.1 I hope you can read uh we have Lama one 8B running and uh we are just going to ask for simple completion does he lik near protocol and um yeah so we have so this specific te is not running with https so we don't have uh the whole verification setup it's just um so this is a developer preview the other the other things that we have in T is are actually fully uh with TLS and https so you can actually nobody has access to what you really

24. 11:58

    saying to to thiss but in that case since it's on HTTP anybody can actually spoof on the network and see what the question and what the answer back from the from the LM is but it's this is like minor issue it's going to be fixed anytime soon so I'm just going to ask for stud this completion so here it is uh okay so uh we saw that we send the the request to this CVM on the port 8,000 for chat completions and then it answered that it loves in your

25. 12:39

    protocol so yeah and then you can actually get the quotes So the quotes in this cvms is basically what uh because you have to to have a way to check that it's running the software that we expect it to be running and so how to do this you basically get a quote out of it and so for This Server this so we expose uh at a station report uh endpoint and when you call this basically it's going to get some measurements from both the CPU and the GPU running inside the te and so here

26. 13:14

    for example you have the Intel quote which is a bunch of random stuff it's like X and then we have a verification process that actually takes this and um checks that it matches the expected values so for for example for the Intel quotes we're going to go a little bit deeper afterwards but you can actually decompose all this and check that the version that is being run is the appropriate one and then we have the

27. 13:37

    exact same for the Nvidia side of things because you have to actually check both um to make sure it's secure so this is it for the private inference part of things so if you want to give it a try already you can uh we only have a 3.18 running right now on the machine uh for storage reason I wasn't able to run a bigger one um but it's totally capable of running these way bigger models so now we going to create like an

28. 14:03

    neari agent so to do this we have this very convenient CI so you do near agent create it's going to ask for for a name of the agent so stver emo for example a description uh instructions so it's just going to be like a this is perfect app assistant oh my God uh so now we just creating it and we can actually go and check how it looks like to once you created an agent so you can go and update it and change and do very way more advanced agents so if I

29. 14:55

    have time I will show you like more advanced agents and I'll say they actually work and can call like vector stores and stuff like this but here we can see the the instructions we told the the agent to be doing and so it looks correct I'm not I'm not going to do something more advanced for now but then we just going to upload it so we're going to upload this version so this is being sent to the to the CDM again so to this te um so oh

30. 15:24

    yeah I have this it's 7 megabytes it's not the correct size just a say because of my pie okay so I'm uploading it again perfect so now it's uploaded so here we can see it's sent again to this CVM new.i so here we are using TLS that's why you see all these warnings because the certificate is generated inside the te and it's not like signed by a counterparty so usually you should not be able you should not trust these

31. 15:55

    certificates but here in our case uh since we do um some process I hope I will have time to show you but basically when uh you get the quote from the from the te we put the public key of the SSR certificate inside the some hash and then on the client side you recalculate this hash and if it matches you know that the public key that you're discussing with on the https side of things you know it's the like it's actually secure and you're speaking to the correct server so now we have this

32. 16:25

    uh agent upload in inside our te and now we're gonna want to discuss with it so just one very simple uh how are you very simple completion um so this is this goes to the this API running inside the CVM and then this is going to ask the VM instance that we saw before it is also a CVM inside the te so all of them so they don't trust anyone any each one of them so all of them are doing like uh these te attestations every time and uh yeah so that's that's how you they

33. 17:05

    ensure that it's actually everything is secure and private and only you can can see the result so here we got back the the answer we saw that we saw the like he's taking into account is Denver it was in instructions so it's yeah so it's working um so now I'm just going to show you more how it works so because here we are using the the CLI so everything is hidden um but so here we get the the quote from the from the CVM which is the

34. 17:37

    The Hub we call it so it's the basically the big API that we use to to do all these AI agents interactions and um agent hosting and everything so we get to code we do the testation checks that it matches the correct version of the software um so in our case for example you noted before but for our version of The Hub you expect to get uh from the I need to run it before so if we run this python script so here we have way more

35. 18:11

    logs uh so here we see the quote again that is huge and here we have the verification results so this is from the attestation so it's actually checking that the the t is is is correct so we you should see uh the correct Ash for the for this ver verion of the The Hub that you audited before that you checked it's running it's behaving the way that you be that you expect it to behave and so

36. 18:36

    you know that this ash is a is correct or is it okay so here we can see uh it got the hashack so you know it's the diversion of the Hub that you actually checked before and you know it's um it's genune and not like selling your data to anyone else or stuff like this um and then so we have a fully the API is fully compatible with open AI so you can create a thread then you as you had a message to the thread and then basically

37. 19:08

    you you run an agent on the given thread so in in this example I'm running this white bits Rock so it's just a game it's like a rock paper scissors but more advanced with LMS so basically you are fighting with the LM to know uh what can first is to be the Rock and then so in this example for example I puted lava so here I say lava and then the LM is going to answer back I hope it's readable uh you want lava melts Rock and then it's going to ask you like what bits lava so you have to find the next

38. 19:40

    um material that's going to beat lava and so then you can just go on with the thread and just play with this game it's a little useless game just for fun so we start the run and then we just wait for the completion so to actually we are checking that the run is um is finished and once it's finished like a it's cool we just uh print the response back from the agent um so now with the little time I have I'm just going to show you more the kind of agents that we

39. 20:11

    we have uh so this is one I did in Bangkok for redacted basically it was the idea of you create an agent with the vector store contains all of ilas tweets and basically you you ask ilas ilas Vector store you ask him for financial advice not Financial advice and you hope that it gives you the next big mem coin to to buy and um so here I run it before um so what community coin because what if you ask for mcoin it's never going to tell you anything it's going to say I'm not giving Financial advice but

40. 20:44

    like here for example uh it would speak about black dragon long um all these all these different uh mem coins on here and uh so how we do this so here we just it's out coded just for the main purposes we out code the vector store ID and then uh on the the agent side you can just do on so EnV is just from the S oh sorry better um so the EnV on is just uh from our SDK so we offer this query Vector

41. 21:21

    store function you put the the vector ID and then the query from the user and then it's just going to like find back all the most relevant documents so Tweets in that case um tweets from Ilia that are good to mostly speaking about mem coins so this is one uh another one so this is what bits Rock from from the de I already showed you it this is Sound Stage so this is the one uh it's one of the most advanced

42. 21:53

    agent that we have um so all these agents you can call it from the assistant so for example for sound message if you say I want to buy so it's an agent specialized in buying uh headphones so if you if you tell him um I want to buy like I know some grid green headphone to your assistant is going to go to the Discovery layer and find an agent that is capable of doing this for you so in that case for the demo I'm just directly on the on the

43. 22:16

    sound sound stage agent um but so we can go here and ask him I want to buy cheap headphones for example then we just hope that the completion is going to be fast don't have much time but so then how it works I have some I have a diagram so you see this so we have the chat. UI that you see right now and then this discusses with the assistant so in that case I went directly to the sunset

44. 22:49

    agent so between these two you have the discovery layer that basically then um communicates between one another and then this agent is communicating to the sound mCP server so it's just an API that is available to buy stuff from Amazon so the Sound Stage agent uh gets the intent that you want to do so here in that case it's to buy like some some headphon of cheap headphones and then it's GNA discuss

45. 23:14

    with this API so it's a another API because it contains some private API keys and you don't really want to share it with anyone and um so then this just discusses with Amazon and then you can do all the all the payments and stuff from there uh so it's still okay so here it found back some uh some headphones from Amazon you can see the price uh all these things so let's say we want to buy this super cheap one I

46. 23:45

    don't have any usdc on this account so we won't be able to actually finish the whole process um it's quite slow the completion so this is something that uh we need to Pro on but yeah so once once you want to buy it then it's going to ask you for the shipping info and then you you fill all this and you you're going to receive your headphones at some point uh in the next few days um and so yeah that's uh

47. 24:16

    everything I have actually I have something else so K from the team ship this yesterday it was asked quite a lot so basically it allows you uh to um embed that the chat box anywhere on any website so if you want to have like a a near like just an assistant embedded into your web page you can um just like use this iframe for example like this and then you specify the agent that you want to be discussing with and um so it could be an agent that's linked to your like vector store of all your

48. 24:47

    documentation about your project or stuff like this and then this this agent is going to be able to answer like any kind of questions for your project and uh so yeah I have this is one one example here so here you see it's an a frame uh inside some documentation so inside the N documentations and then you can ask whatever question it has a as Vector story it has all the documentation that we have on nii and so you can ask whatever question so that was it thank you so much for

49. 25:23

    listening and I don't know if I have time for questions yeah uh hi guys uh I want to say is uh many other applications uh besides large language models should be also considered into a te Pro protected environment so uh uh as uh near are you going to plan to uh uh provide a service for like text to image text to uh video uh 3D and Etc uh or just TT or TTS exact EXA things yeah so so right now we on the The Hub so the non side of things because we have this whole product

50. 26:23

    running outside of T also so we have uh text to image already but not not in Te like from our te solution you cannot create U im images right now we did not work on it still okay the next question the next question is about uh what do you op what's your opinion upon on the uh do we know the sandbox of FILA or the e2b uh bo in the web 3 and web 2 have you ever heard of that use a Sandbox to to wrap the code uh to give a trust

51. 26:57

    trust uh trustful uh execution for for both users do you have ever heard of that E2 no okay so I can have you dep discuss yeah thank you thank you really quick how can we do with a testnet Sor the near AI login but with a testet uh yeah it's only with main net right now uh that you can you can log in but it's uh we are not said you don't need any funds it can be a throwaway account you

52. 27:31

    can log in without any funds on the account it doesn't require you to answer I just created a like a new account which has nothing and I'm not able to log in because it says throws an error normally you should be is it a named account or anyway it's okay it's not a named account yeah that's why yeah yeah but with me your wallet you can create name named accounts for free and it should

53. 27:55

    work I'll send you s you okay any other questions yeah can I have to go yeah okay good though all right brilliant um look Pierre is right there if you want to talk to him catch him now um I'm gonna whiz this off stage and we're gonna uh give a big round of applause to uh Cameron Dennis and his wonderful panel guests so round of applause please hey can you guys hear me right hello okay test test all right great you guys can hear us say something

54. 28:49

    yes great all right everybody how's it going so uh my name is Cameron Dennis I've been building a near for about 4 years uh build with near AI for the last about eight months to year or so it's been an amorphic organization but I'm here with uh Felix and Marvin uh we're going to be talking about the private MLS Decay and what it means for uh verifiable privacy in the world of AI so without further Ado we love for you to introduce yourselves Felix you want to go first uh

55. 29:19

    I think you can go first okay cool yeah okay uh thanks for having me uh my name is Marvin uh I'm the co-founder of f network uh I'm a running a non technnology role coo and we been working in the space for five years on te and uh uh basically in the beginning like everybody in crypto space hatte te you know it's not purely based on crypto graphic Solutions but like we believe there are some uh big potential in it and uh it can scale you know uh uh

56. 29:51

    uh to uh you know to help the web three standard e the web two software so that's basically the or what we believe in so uh yeah we've been working in the space for well and it's super excited to see how AI jump in the space right now and the ver ability can play a part of it yeah well just before we move on tell us a little bit more about followup like what is f yeah F in reality it is a abstraction layer of uh how te Works in

57. 30:21

    a wep2 world so in know T first of all te is not Hardwares right like te is a system systematic work based on how security Hardware works and uh each uh chap manufacturer have their own te system such as Intel SGS int TDX mdsv uh Apple have its own te standard and uh yeah so when you unlock your face in your phone you're using te right where you open up uh bance these uh uh you know pass Keys you are using tees so it's kind of already been vely used but the trust Assumption of this web 2 te

58. 30:59

    doesn't fit with how web 3 you know use cases Works in web 3 scenario especially like uh uh in defy AI we uh consumer don't trust the developer developer don't trust cloud provider so we only charge blockchain right and ckp so how to build a systematic work based on that it requires some software level of abstraction so F basically do only two job to abstract this web 2T standards uh together and provide a software solution

59. 31:32

    uh that fit with web stream requirement and another job is that we provide security Hardwares that fit these these uh uh Solutions yeah yeah we'll we'll get more into the details in a bit but uh Felix quick introduction of yourself and a little bit about xrace hi every I'm Felix I'm the CEO and founder of XTR and we are a pretty new company basically started in June and what we're building is on encrypted context layer

60. 31:56

    and data and memory for AI agents so what I mean by that is that every user can upload their data onto our system and then as a data protocol we can guarantee that we have no control of your data so that uh all the data will be safe and it's still searchable by AI agents so what I mean by that is usually the guarantee for database holder is encrypted at rest and at Transit and xray get one further guarantee which is encrypted in use so AI agent can still

61. 32:21

    send out encrypted query to xr's encrypted database and get the data they want so this pairs perfectly well with te because we can essenti allow agents or large Range model running tees to have data access to external databases and on top of that we can also be the persistent storage which is called memory for a agents running inside which allows uh this to be on persistent storage it makes the agent running inside State full yeah yeah just for

62. 32:46

    summarize quickly there's a lot of words so in short xrace is really providing the personalization like Railway so if you're using an agent and you want to say hey agent go buy me a cup of coffee that agent doesn't understand like whether you want whole bean coffee or you know ground coffee or hot coffee or cold coffee like it needs to understand your preferences so with xrace you're able to use xrace to personalize your

63. 33:11

    agent while ensuring privacy which is really a prerequisite to this like user owned AI thing that we're all talking about so um thank you and uh moving back to Marvin so uh you know your whole mission is to make you know trustless compute affordable with sort of through the private MLS and also everything else that you're building with uh Falla and I'm curious like how you see this uh kind of adoption by developers like do

64. 33:34

    people really care yeah about privacy and verifiability right now that's a great question so uh like our whenever I'm teaching I'm now saying hey use know use te instead of take using te instead of smart contract on eer our mental thoughts is if you can run something on eam do it if you you can run an agent on year do it like like te is the technology that uh has a a better solution than Amazon Cloud right like it's kind of in the middleware between

65. 34:08

    uh how web two and web three works so uh we we are a believer of you know chain maxing and we are believer of cryptographic solutions will merge together eventually we already see the train how this training works for example FAS te attestations writing using zkp to generate the provs so that you can use the ZK proof to know you know as a as a as a testation for how the uh te uh works so like so AR is like

66. 34:40

    uh the web series adoption will haveen um many Technologies you know uh working together instead of single uh Tech so you said you said this word attestation yeah it's a very important thing that everyone should understand yeah can you quickly like explain what te are okay and like where they live yeah so uh security is based on assumptions right like for example uh like a her work is

67. 35:08

    just imagine there's a black box in a CPU or GPU in these chips and the programs the data running inside this we call them Enclave the black BX running inside The Enclave actually is running in secret but uh uh uh better a testation system you can know exactly how this program running as it should be so how to know it first we need to know uh like how this program looks like so you probably need to open source to make sure that the others can verify it and then you put the open source software

68. 35:43

    inside this uh uh security uh chip uh unclaim and execute it with a testation which is provided by the uh GPU manufacturer you know te system you can know okay this computation result is actually running with a specific Hardware environment with uh you know a specific software uh version so that you know okay with the attestation proof you know how this program is running as it B so that's the very basic very rough

69. 36:17

    abstract description of how te works but in reality is more complex for example right uh like uh it's based on uh a security assumption for the Hardwares so what if you don't trust in the hardware right like things have been hacked yeah it can be hacked it it can be hacked so you should know like these security assumptions before you use it and another thing is uh there's a uh for

70. 36:47

    example like with testation is not as simple as zkp with single Pro you can attach the entire you know MCO tree or the the entire har you know uh history behind it you in with te attestation it can separate the uh verif abilities so for example some proof uh actually means that okay the environment setup is correct some proofs means like the uh Docker you know runs at whe so Etc ETA it it can be separated into very different layers yeah so now we have a

71. 37:26

    general idea of how to is work and what they are but um Felix is bit more on the the rag pipeline so can you quickly explain what rag is what it stands for and exactly what uh xrag is the sort of new primitive that XR introduced I think to me like AI agent is composed of basically three Comon right like the larger Range model the second is a rack and dynamic memory and the third one is a tool code ability and then what is

72. 37:50

    ability to C to yeah like calling API to execute certain functions and then what rag is is basically allowing you can think about when you're training a large Range model it's kind of like a uh training school and when you're doing the exam the rag is like open book so way for AI agent to reference outside data that is not outside of the training data that can be like real-time data or like propriety data private data or all of that and then what we're trying to do

73. 38:16

    is that we're making sure that all the data stored inside our Vector store are all encrypted and then still still uh still available to be a semantic search on so what semantic search means is that the agent can send out a query and then find the most relevant data in the whole Vector store and we're making the whole process encrypted which means that the query is encrypted the DAT is encrypted nothing is leaked for in the protocol

74. 38:38

    and then still the agent can find the response they need it's crit like super important yeah do you how do you anticipate this actually getting out to the masses like do you have to you know natively integrate with xrace every single time can these agent framework sort of add X Trace by default can this be added to the private mlsd so everyone that adds like you uses the private mlck benefits from private rag or you know crypted memory yes or cord or like

75. 39:03

    design is REM modular so we allow like a chain agnostic and also agent framework agnostic as long as you have data that you want to work with certain like agent framework then you can use your system and then make all the data you want to interactive uh encrypted so we really see that working with all the agent framework will become like this distribution Channel and then our ultimate vision is allowing like one user to upload all their personal data into our data layer and then they can choose data Access Control to different

76. 39:28

    agents they're working in for for instance they can only open their e-commerce like habits to the e-commerce agents and then the text data to the text agent only so that they have one unified layer that has a shared memory and like and kind of gather all the fragmented memory and still work with all the different AI agent AI apps they want to work with yeah I'll say um I'm working closely with these guys because honestly I want a solution which is I

77. 39:52

    want to be able to have a single window that have an agent watch my screen at all times s verify that none of this data is being leaked to some third party cloud and it suggests questions that I can ask on calls and like you know suggest things that I should do with my with my browser and so um this is what I think that we are all sort of building towards like this like private ml um but I was recently at uh uh an AI conference

78. 40:18

    where everybody didn't think private ml was ever possible but now we see things like you know Nvidia coming out all new h100s h20s will have this te support and so this demand is not coming from cryptool land it's coming from like the large know Compu providers like you know Nvidia Intel yeah like how do you sort of see this sort of playing out long term I think a very important reason is like uh actually we uh know the

79. 40:42

    information about private ml in not not not know the concept of you know this uh technology but more like how the uh uh business already been working for a while uh through the last I think ic3 uh conference in San Francisco like it's more like a academic one and actually the researchers from Nvidia Intel you know open all of these guys are there all of them know that private uh ml is already working and so but unfortunately it's more targeting on Enterprise level

80. 41:16

    of users the only company who actually doing a good job using te for a private AI is Apple all of the Apple intelligence uh cloud service going through the te Cloud powered by Apple M3 uh security chips so like exactly the chips running inside this Max so but I'm thinking like uh now it's more tting on Enterprises but uh it's going to be consumer faing like uh I'm not sure how many of you guys already used uh opening

81. 41:49

    air operator it's terrified yeah you have to like kind of like type in your login and input all the password in the cloud browser so that this uh you know uh agent can operate jobs for you in that cloud browser and is having Mass adoption to their consumers I'm saying I I I don't trust it no yeah I think I saw in Microsoft like the new uh I think it's taking screenshots of your your window pretty much like every x amount of seconds um yeah it's horrifying I don't want that

82. 42:23

    so Felix um for encrypted memory xrace um what are the industries that you're really looking to kind of go to market with yeah I think currently we're just uh observing like a ver of like agent framework building like a lot of agent Launchpad and everyone is trying to kind of work on the agent site and then we really looking into uh two sectors of one is like just encrypted knowledge based like how rag works the second part is on encrypted memory because this

83. 42:48

    memory that is generated by large range models through like all this test time scaling and a lot of reinforcement learning is really valuable and I believe that this new memory andc memory will unlock collective intelligence between AI agents so that each AI agent can call own the data that they generated or Own It by themselves and I think this new paradigm is like where the new data opportunities unlocked so we can kind of make multi- agent collaboration really happening where all

84. 43:12

    of them have their own data yes currently what I'm seeing is that a lot of multiagent collaboration just like the same chat gbt or whatever larger with different prompts I think what really makes it different for multi-agent collaboration is the data they have access to the proprietary data the past experience they have and that'll really make a difference when different agents are working together yeah I am super bullish on this approach like uh I I I think uh uh uh uh privacy

85. 43:37

    is just like a basic uh feature but like very few uh people or Enterprises will pay for privacy that's the reality but uh based on this fundamental of privacy you can build a very good ideas for example in a in a private RG scenario I'll say I'm super bullish on the sharable memories because my pinpoint is I'm using anthropic I'm using open Ai and I have to upload everything about t follow again again again with a lot of my tokens I believe many consumers feel

86. 44:10

    the same too but with uh if there's a like personalized memory database that I can upload my knowledge for once and it can be shared and used to many different agent platforms that would be a super convenient B save a lot of Computing resources I'm kind of curious like I think one of the biggest issues that we're all going to face in this like privacy verifiability world is actually

87. 44:35

    getting the real world to trust the code trust cryptography yeah how do you see us actually bridging that Gap you think it needs to come from trusted organizations like apple or can it come from you know applied cryptography Labs that's a very good question it's a tricky one like for example I what's happening on safe right now it's kind of a good sample I'm I'm not saying I think safe have the one of the best security teams in the world in a web stre but

88. 45:05

    like the B Teck viabilities is not because of the contract right the contract is verifiable so everybody using it for many times so it's safe but for the uh how the front end and the server end Works they still need to uh uh use some Dynamic back end of to you know for the uh user experience on the front end and of course there is tons of alternative front end for Save but still

89. 45:34

    not career in right like why because as long as you are using server there could be availabilities how to save that I say uh I'm not saying te can save the issue te can't solve this would it be like a standard does it have to come from a company does it have to come from a government like what actually what will get people to trust verifiability and cryptography yeah I mean uh but I'm I'm just trying to address that if you put how the back end

90. 46:05

    of say works and how the front end works it open source it right so that everybody can reproduce it by by yourself B uh providing centralized service is okay but for example running inside te at least the security uh the white hat are always tracking the attestations of this backand logs to see if there's anything weird happening to see if there know the hacker changed the front end backend code you know without

91. 46:36

    your knowledge like the more exposure of a transparency logs from Ai and you know back end or front end the better y yeah that's my point yeah that makes sense uh for me my security assumption kind of based on like the https protocol like using similar like encryption method and AES encrypt ion so kind of like if you want to say I'm unsafe you have to first break htps or ases so that's kind of how we uh address a cryptography primitive

92. 47:07

    yeah so some sort of standardized protocol that really if you ask the you know joea off the street hey what's https they're not gonna know or care but they kind of trust hpps as a default right compared to the previous times I think Google kind of forc it to happen yeah because they have the Monopoly of like web search so they just kind of drank everyone who have the TOs protocol enabled so not yeah so I think there's

93. 47:32

    some way to make it work I mean this was uh this the last question it's a little bit you know far out but you know assuming all of this technology ends up getting you know really proven out and it works um how long do you think it would actually take to get in the hands of Everyday People well I was saying it's not a technology question because it's already there like based on your our Benchmark

94. 47:54

    in very early in last year using te gpus like h100 200 you only have 5% of tradeoff to have V level you know inference for basically any uh model so it's already in there and for AI agents is a is obviously is a it's it's already working very well in tees so I would say you know for uh uh but for training is is I say this year is is a beginning because uh like uh trust list or verifiable training is a uh is haven't reached a like a full kind of uh status

95. 48:35

    yet so it's only a time a matter of time we need the hardware yeah like it needs to be everywhere yeah which is happening yeah and what happened in the market actually help to pushing that a lot of like us companies and US citizens don't like for example deep seek right like we love deep seek as a open source software but you don't trust the inference service from China so how to solve that like we need some trust L you know middleware to operate these uh open

96. 49:06

    source models and you can see how they collect your data or not collecting your data so I think you know it's a the market is pushing it yeah just to kind of repeat the question you know assuming all this works perfectly today what does the world look like in about five years yeah it really depends on like how people how good the model is and how good the vertical agents are so I think currently I think the agent still hasn't leave up to it hype yet but

97. 49:31

    I see a lot of good transformative like break through every week I'm doing like this uh nearly week with cameraon about like all the AI research and everything so I can see it happening but I think just takes like that breakthrough Point infection point and then when there's Mass adoption then we can see that people actually need to upload all their data to make sure like the agent can get the best out of all this proprietary or personalized data yeah than guys that's about it we'll be around if you have any

98. 49:57

    more questions thanks for listening thank you thank you take sorry sorry uh next is uh is it Matthew

99. 51:09

    getet that if you are hungry peckish uh actually there is no food there but in a couple of minutes uh there will be better food there I think and there's coffee there's water there's juice there's Sprite if you want to drink that poison uh otherwise uh please give a big round Applause for Mr Matthew Wang um to talk about our decentralized compute hey everyone my name is Matthew excited to be here uh yeah I'm uh one of the co-founders of open

100. 51:42

     gradients and today I'm here to talk about how to make uh intelligent adaptive agents in real time a little bit about myself before I get started uh I spent most of my professional career as a research engineer at a Quant fund called two Sigma New York um when I was in school i' had also done stints of HCI as well as ml research notably in applying unsupervised learning methods um in Market regime

101. 52:06

     detection and using that for portfolio optimization and uh I will give everyone a quick rundown of open gradient shortly but excited to first talk about what the road to a trillian Advanced agent kind of looks like in my view so I kind of think about agent development in a couple different stages stage one you have agents that can you know synthesize and reason about with data which you can

102. 52:28

     already see with all these Twitter agents popping up tweeting interestingly provocative things secondly you have stage two which is agents are able to complete tests on the behalf of their human overlords such as placing simple transactions such as you know buying a simple token and you kind of already see a lot of AI agents notably on uh the near chain kind of able to do this and serve as assistants um and here I've

103. 52:53

     also attached a little screenshot of um one of the tweets that Ilia had put out about how a lot of these agents are kind of helping humans with mundane tasks like routing trades and routing orders and lastly the kind of final stage for AI agent development in my opinion is when AI agents are able to surpass clearly human abilities in handling complex tasks and this might look like you know AI agents that are able to run fully autonomous companies AI agents that are able to run and participate in

104. 53:23

     Dow Dows uh as well as AI agents that can man portfolios and be able to you know trade effectively and generate profit consistently but if you look at the state of uh you know Frontier models llms and AI agents today a lot of them are extremely intelligent and uh you know especially given the previous panel can access things such as tool calls and rag in order to make it smarter but when you ask it hey what is the price

105. 53:49

     forecast for sui what you know asset should I buy to minimize my risk adjusted return so to maximize my risk adjusted returns things like that the llms or AI agents aren't able to give you a very textured or detailed response and in our view we kind of want to build AI agents that are able to address those tasks and give you texture responses and execute on them effectively and for my view there are

106. 54:14

     essentially three different kind of bottlenecks and Technical hurdles that we need to overcome in order to advance Agents from stage two to stage three firstly data pipelines you need to give AI agents the ability to reason about with more complex and a wider variety of different data sources as an example if you have a portfolio manager agent that agent might want to be able to pull Twitter data in order to analyze the sentiment around certain crypto assets

107. 54:39

     on crypto Twitter secondly computational workflows allowing AI agents to um you know venture outside the world of using a large language model and run a computational workflow that might involve as an example um a mean variance portfolio optimization workflow that gives the A and the ability to optimize its portfolio Holdings or run a simple risk analysis on uh you know various

108. 55:02

     assets things like that allowing it to you know uh actually perform detailed analysis uh and um you know become smarter with domain specific intelligence and lastly is Agent interoperability uh so I'm extremely personally bullish on the ability to allow agents to compose with one another in order to comp in order to complete extremely complex um and Abstract goals as an example a portfolio manager agent might leverage an execution agent in order to find the best venue to trade to

109. 55:34

     minimize Market impact or to minimize uh you know the cost and fees that they would have to pay in order to cross the spread and might use a marketing agent to you know pump uh tokens that it might hold in its portfolio things of the sort and I think these are all extremely exciting kind of um you know Primitives that we are building at open gradient as well which I'll get to shortly um and

110. 55:57

     this is just a quick screenshot of a little piece of writing that we had put out about how the three kind of key pieces that I just spoke about will allow agents to become more intelligent with access to you know computational workflows data pipelines and uh you know access to other agents and you know allowing us to compose different agents together to create powerful computational applications but what's next so briefly

111. 56:20

     on what open gradient is we're building uh a full stack infrastructure platform that allows developers to very freely be able to build AI or ml applications on chain in a way that is both verifiable and extremely seamless uh we want to allow developers to integrate quite simply any model into any application on any chain uh we have an infrastructure layer uh that has you know various nodes they handle compute heavy tasks including inference including running

112. 56:46

     secure data pipelines we have uh our blockchain Network which allows developers to build um using these data pipelines using different models uh directly in their solidity smart contracts uh and we also have cross-chain messaging that allows these computational workflows to service apps or agents or protocols on any other chain we have a model Hub that is live at hub. openen gradient. that you can check out with a couple hundred models

113. 57:10

     many of them are web three specific we host sentience uh you know large language models so this is Dobby we host flocks small language models we have many many uh you know dii quantitative ml models that do spot forecasting volatility forecasting risk analysis and more and this is all live you can go check out and play around with those models and we have many active Builders building with these models and lastly we have uh pythonic SDK that allows

114. 57:34

     developers to leverage our Tech and interact with all of these models and workflows without having to deal with the complexity of block blockchains for smart contracts and we also partner very closely with clients in order to get their agents protocols and applications off the ground by partnering with them and a lot of the time building models um and agentic workflows alongside these clients and I'll get to a couple interesting use cases later um you know

115. 57:57

     a quick overview on how we make this magic happen essentially it's through node specialization so uh unlike a typical blockchain we have many different types of nodes that do very specific things including inference nodes that run verifiable inference and generate proofs and adapations uh data nodes that are able to you know run secure data pipelines to make these workflows data rich in nature and a decentralized file Store where all the

116. 58:21

     models are stored and we have today one of the the only repositories of web 3 specific models including like I mentioned kind of these small language models large language models and a variety of quantitative defi focused ml models that allow developers to build many of these powerful AI agents that can manage portfolios trade you know analyze risk on different assets and uh we also host yeah many um of our partners' models including Pond's model which does cble

117. 58:51

     address detection and repos uh and file coins models which do minor alloc and and reputation uh analysis and uh kind of tying this back into how we intend to make AI agents intelligent and adaptive in real time by leveraging these computational workflows powered by these highly specific models we are able to give AI agents domain specific intelligence instead of getting a extremely generic response from Claude now when you work with clients on open gradients like exponet when you ask hey you know give me a short-term price

118. 59:21

     prediction on an asset or you know uh give me the risk profile on a on suie it can actually query these models and give you you know historical forecasts as well as the current live forecast that can give you a much more textured and detailed response that actually addresses um you know the user's concerns in this case if I want to buy sui I want to understand the risk profile the current volatility associated with the crypto asset um another interesting one um whatever our

119. 59:46

     clients which is uh double up they are a gamble five prediction Market protocol um launch an AI agent that allows users to actually bet against the AI agent on The Price Forecast of different crypto assets um so the human can you know the human user can type in some Price Forecast and the AI agent of course can't forecast these prices so it runs a spot forecasting model on open gradient and that's able to give it um you know a

120. 1:00:10

     very specific forecast Power by time series analysis and uh you know in 30 minutes where when the the market resolves either the user double up doubles up or doubles their money or they lose uh their stake right um and from a quick preliminary analysis since the launch of this agent last month we can see that hey you know users actually perform worse than the AI agent because the AI agent has obviously access to uh

121. 1:00:34

     ml models that can inform it of what the price predictions are and this is kind of like a very barebones live in production dealing with real user assets example of how we can give AI agents the ability to you know predict prices to eventually trade and manage portfolios effectively through the use of these modular um Ai and ml you know focused workflows and we're also working with many clients um you know big defi

122. 1:00:58

     protocols on building analytical quantitative def AI agents that can give them uh you know that can give the protocol and users of the protocol access to many of these analytical capabilities and workflows powered by the kind of AI ml models on our Hub and on our infrastructure so you can ask you know these quantitative agents hey you know what asset should I buy to maximize my risk adust returns if I have a

123. 1:01:21

     million dollars of Bitcoin where which vaults should I deposit in in order to you know improve you know to to maximize the yield I can get on these assets and this AI agent is able to query and leverage these workflows in order to help users um achieve these abstract goals so I have a quick demo here which unfortunately probably is not going to load uh but this is just a quick example

124. 1:01:45

     of if you go to hub. openg gradient. you can go upload any arbitrary ml model and inference it within seconds verifiably and there are tons of different models on their um you know courtesy of us as well as many of our modeling Partners as well as individual contributors for AI and ml specific web 3 models that you can go and play around with today um and yeah uh with these models we're excited to be introducing them as tool calls to

125. 1:02:10

     the AI agents and we're working with many of the popular Frameworks as well to implement these tool calls um and uh very excited to be uh you know unleashing the new era of intelligent agents that can leverage these computational workflows to do much more complex and Abstract tasks thank you all right they've already St on the stage but this is Mark me and he's going to take you through a panel on all things institutions and AI agents big

126. 1:03:15

     round of applause for the gents please thank you Deo sorry about that I I was should have just waited took your thunder hello everyone my name is Mark me um I am the uh head of partner development VC and investment relations here at near Foundation I don't really probably need to really explain what near is since you probably have four hours of listening that here today so I think we'll we'll

127. 1:03:41

     get into today's topic um of why Enterprises and institutions need to start using agents right I think on today's you know simple chat we have two very different types of companies and so maybe we we we'll start and let them introduce themselves and kind of their background as well as the company itself maybe we'll start with rooster yeah um it's nice meeting you guys my name is

128. 1:04:04

     wer um I'm a community lead at hasht um for those of you who's not familiar with hash we are venture capital for that is focused on on web 3 um and we are based in San Francisco so uh Bangalore Singapore and now we're like um expanding to ababi and um my role at hash I uh as a community lead um I do connect the dots between between the portfolio companies so that they can create synergies within um within our umbrella and um try to create good Partnerships for um our portfolio

129. 1:04:34

     companies in in the meantime hey everyone my name is bang alashi I'm the head of strategic Partnerships at bitgo um what does that mean that means everything high high Roi high impact um engagements for our business verticals um and of course for those of you who are not familiar with bitgo uh bigo is probably the largest custodian U independent and qualified custodian in the world we have over

130. 1:05:01

     hundred billion dollar um in assets under custody um we offer wallet infrastructure custody U staking trading and settlement out of our um Cold Storage um you know today we support over 30 l1s um including near I might add um um 1,200 tokens um and Counting Innovation doesn't stop at bitgo and most two recent tokens we supported are Oasis which is the gamei chain and um benser um tow so we're really souped

131. 1:05:33

     about the AI space and supporting it from an infrastructure space happy to be here awesome uh glad you added that point that near is also integrated I think otherwise I don't think you'd be speaking here today um just getting on that but uh realistically I think one thing that made me realize while you guys are were describing your your companies is the fact that I didn't know how large and Global you know I forgot how large and global hashed was right and and just to you know I think we're

132. 1:05:58

     pretty lucky today to have basically two leads from two entities that basically you know one of the largest Korean VC funds as well as May Global as well so what's your guys's take in terms of I guess um you know within the AI agents right what what excites you the most I think that word has been really thrown around a lot in the last few uh few months and everyone says agent this

133. 1:06:21

     agent that agent token this agent token that and so would love for for you guys to really talk about what you guys are excited about and what you guys are seeing so maybe we'll start with b yeah outside of managing my e Denver calendar um I think um you know the two areas I really care about one is really improving the user experience in web 3 and what AI agents can do for that and then I come from a finance background so agent five so sorry I'm using it but

134. 1:06:50

     agent fi um which is really just seeing AI agents um integrated into def sort of like an intelligence layer over the global flow of value and you know driving AI use cases around trading strategies liquidity management portfolio management okay um for me I guess like you know just put it in a simple term I think composability of the AI agent would is probably the most exciting part um I mean it's been a while that we've been using a lots of um llm services

135. 1:07:21

     like open Ai and F and stuff like that um so the basically it changed how we interact with information right um like our ability like in throughout human history our ability really came from information collection to information processing and now in this era like in information processing is already been changed through AI so um you know what implies in this is that like you know I think I see a lot of AI agents being

136. 1:07:51

     built being able to execute like certain actions above the information processing um so that means that would allow me to do like trade or like book a calendar or anything um you know it's something that Siri has been doing but in so much more simless and in intelligence manner um so you know I think that would change our digital interactions throughout the days um and you know why should I not be excited about that so um I think what both of you guys are also saying is the

137. 1:08:23

     fact that we're still very early right um and and what I kind of compare this to is I don't know if you guys play video games or not but you know before we had like Nintendo and if you look compare that to what computer games are now like it's completely like not even comparable right and I think that's similar to kind of where we are now is we're testing these points and these ideas for agents you know low hang fruit

138. 1:08:46

     being easy transactions trades um you know scraping socials for for what Hot Topic is something that you really want to learn more about or invest in or really jump on right and so I think it's always the first phase of having initial Tech and in this case agents to really you know do that lwh hang fruit but then we use it so uh we use that word so widely that we we think we're already there but in reality I think it's still very very much in early stages um both

139. 1:09:17

     of your guys' companies are are really to completely different Realms one being more investor one being more of an Institutional service provider I guess how do you guys see agents AI agents really changing your guys's day-to-day or or kind of what you guys expect in the next five to 10 years um for me I think just day-to-day operations is rather easy um you know just trans you

140. 1:09:40

     know uh translating a document to different languages has always been quite an issue because we're um we have very Global offices with different linguistic capabilities within within the team um and I think you know the advance of agents kind of made it so much better in terms of communications um another aspect of that is probably doing the research right um so we used to have a lot of interns just doing the research and to follow with

141. 1:10:08

     the trend um given that this crypto Market has crazy hacktic speed um of having the new trend new narrative the memec coin was crazy um now with AI agent I think it's much easier um if I just set the right parameter for that um so called Alpha Gathering through AI agent um it's not just confined to uh VCS I think it would apply to a lot of retail users um and day-to-day like crypto Traders as well in the coming days um yeah and I think um not implemented

142. 1:10:42

     right now um and not sure if we're going to plan for that but uh you know if you you know Bridgewater by Ray doio um in his book he said even back in like 1990s and early 2000s he used computer algorithms to make decisions and to make more rational decisions in the market um and now that AI agents are being actively being discussed to get into the market and do like more instantaneous trading um I think it would have

143. 1:11:13

     potential to make some of the important decisions that we don't want our emotions or fear to be in infiltrated and kind of taint um the decision so that should be the some of the part that would quite be exciting to explore in the future sounds a bit scary of but um and and just the last point is are you guys already utilizing that within your dayto days now or not really yeah no but it's just um you know possibilities as AI agents um you know

144. 1:11:41

     gets developed and more private so that we don't fear this decision does not affect any like Market manipulations or anything that's fair enough okay yeah you know very similar um early days so we're exploring it but you know any data flow that I mean any processor that is data intensive is definitely areas where we think um AI agents can play a role so whether it's streamlining you know HR processes or financial planning um you know sales campaigns that sort of a

145. 1:12:14

     thing or understanding you know what's going on in the market um or areas like high level I think and then from you know our Core Business on the custody side you know maybe um use cases around wallet security um risk management um for or trading platform are some of the areas we think that you know agentic um um you know AI agents can sort of like come in U but again still very early

146. 1:12:40

     days but um we're fully um supportive of it so so you know knowing that it's quite early and knowing that it's constantly changing how are you guys staying on top of it and how do you view like you know if if I'm not reading news for basically week I feel like I'm already out of the no right like like forgot about a new large language model or or a new foundational model that is now better and cheaper and you know

147. 1:13:05

     things like that like guess are you guys actually looking at it a lot and keep making sure you keep up on the data and the updates because I think it is difficult we're we're plugged into the the conversations right you know I'm here we have a good relationship with with near um and um we we have a ecosystem fund that um you know talks to a number of folks who are building in the space and so we're definitely staying in tune and a breast of developments in the space and you need to do that if you want to remain

148. 1:13:35

     competitive if you want to um optimize your business processes and of course there's the other side which we haven't really touched on but the ethical considerations around um Ai and how um we as good um um uh I guess good actors there's the word um want to make sure that we're lined from um an ethical perspective as well I think that's a that's the point I was trying to get uh M to really highlight is that I think if you don't innovate and you don't actually follow along I

149. 1:14:07

     mean you're you're you're kind of screwed in a way right you're going to be left back and and and not be able to catch up um so I guess rooster like I guess is this something that's that you guys prioritize or how do you stay stay on top of it from a fund perspective investor perspective I mean um I know there's a a strong Nar between crypto and AI um but I I think as an asset class um it is still there has to be

150. 1:14:31

     some of the Amendments um within the business models of how Ai and crypto should be made um we're we're excited about um how things are being integrated and we're excited about like people being incentivized through developing this opsource llm models and and AI infrastructures um but when it comes to like investment decisions on on specific tokens um you know there has to be sort

151. 1:14:55

     of um a value CR process for that um and it's also my personal opinion where I think for a lot of AI crypto projects um I like the initiate and I like the spirit um but currently there's one lacking part which is the AI products creating revenues um and that because it is missing now know the the tokens are just being said as mere meme tokens that are being constantly being printed out

152. 1:15:25

     um but the intention is there for for the development through for the um Ai and crypto um so we're looking very close to how we can improve the space um we're also working with um like small and large institutions on um you know how to make their product better with integration with llm products um but yeah um yeah so we're we're looking into a lot a lot a lot of different aspects and different sides um but genuinely excited about the integration I think it's kind of funny you you highlight the

153. 1:15:58

     the RW that normally people in blockchain don't like to talk about which is revenue um but uh I think you're right I mean that's the direction in order for long-term sustain ability long-term growth and really adoption that's the proof point right um but you know being here and seeing the vision you know every time I talk to Ilia I was you know I always feel really you know feel his passion and motivation and and

154. 1:16:21

     kind of what we're really talking about in terms of user owned AI right the future where you're not you control kind of your own data and have a decision in it right now we're all scrolling through Instagram Facebook and they have all of our data right we're not paying for these Services they basically own us and they know that but now if you actually see the next stage if there's actually agents built on top of that data it's mind-blowingly scary right and so what

155. 1:16:46

     are your guys's thoughts in terms of I guess decentralization and how important and how to prioritize that as well as the user own you know user controlled a this way you have a choice versus if we don't then it's just like Terminator and Skynet um decentralization is extremely important um we've seen how um centralized AI is essentially abusing the space right now and owning a lot of user data I love the way near AI puts it so I think your

156. 1:17:19

     mission statement is you know the future is um owned by the user something like that I can't remember um and I think that's key you know if we're going to focus on security if we're going to focus on privacy if we're going to focus on allowing users to do what Their data what they want um then we need to align the use cases as they evolve um to um you know the centralized model which is

157. 1:17:44

     of course the ethos of of crypto um so my thoughts there really as just um we need um folks like yourself and other builders in the space to really dve Drive um creating these agents in a very decentralized and open source way so so may maybe click a little bit deeper um what do you think it takes to get there right I mean like where are we in that process right because I I think

158. 1:18:09

     obviously we know we're early but I guess how close or or what what is needed right and the reason I bring this up is is because I get really excited when I think about you know our our current cross chain execution layer you know we spent all last year near spent all last year really focusing on chain abstraction and and signatures and a lot of the team members here have really put in a lot of that effort to ensure that you know you're able to work with different chains and really to take and

159. 1:18:32

     remove that layer where where chain now doesn't really matter anymore right and and and so to give you some a few examples where you know utilizing chain signatures able to really access um prior other I guess chains and tokens that previously were not able to be swapped or or you know worked with such as you know Ripple Bitcoin you know um um zcash on the first day of be chain launching their token you know we were able to integrate that within our near intents as well and you know able to do swap so

160. 1:19:03

     I think that's what excites me in this in this point that look now we have this cross chain execution layer that you're able to go from ideally you know Fiat to anything and then back to Fiat right and thus able to go from the AI piece the AI agent side where at this point you can have and work with those agents to execute whatever you wanted to ex execute right and and takes that barrier

161. 1:19:27

     away and so I think that's kind of really the exciting exciting piece um with that said um what types of AI projects or or you know are you guys looking at are you excited about I don't know if if there's especially from the investment side that really jumps out at you guys yeah I mean like we just look into the general crypto and AI sector um overall um not to mention any specific

162. 1:19:50

     projects um but AI agents um you know how they framework is how well is it being integrated to um you know some of to interact with smart contracts um and also we're also looking into um you know something like open source but with a sustainable incentive mechanisms inside um and I think there has to be multiple factors to play inside um like in general I think like decentralization of AI de AI development is a must option to

163. 1:20:23

     have and not ant um because like if you think about social media um you know how the big Tech has been maximizing their revenues and profits out of the algorithms has really changed how we interact with information um and just a just imagining how AI will change that again but in such a more private manner kind of scares me um so I think while the big tech companies will continue to build their own AI models I think um you know a you know decent centralized open source projects should really be there

164. 1:20:58

     um to create an alternative for people to still benefit to technology um and leave out the world so we are doing um you know doing a lot of thorough research on the framework um on the like development of llm models um but yeah like um some of the specific projects I think it's not quite safe to mention it um to to speed out okay uh not not a problem yeah definitely wasn't help for someone to call out some specific projects but I think you bring up a very interesting point of the change of

165. 1:21:29

     efficiency and and how things change right and I I I see that from multiple different points um an example would be you know how how just compute right it you know was so much more extensive prior and maybe it'll slowly go down and just be more of a commodity um when you see uh you know data storage right you know da I think that was a discussion point where before you know is it all equal is it different and I think you

166. 1:21:54

     know as as it gets more improved and it's actually utilized it gets more efficient because the demand's there thus you either increase Supply or you improve the supply to make it so that it's more efficient right yeah I'm you know I'm 100% biased here I come from a finance background so for me it's all the use cases in finance um agentic fi um you know um building string um trading strategies portfolio management

167. 1:22:19

     risk management so a lot of those AI agents and you know how a better um Financial outcomes whatever however the entities or the Enterprises Define that is really exciting for me so so I mean we only have about a minute left and I'm trying to think you know typically institutional products don't just don't get that much investment in terms of short you know they're much longer term

168. 1:22:43

     and maybe larger investment Capital needed right I mean what do you guys see at least within I'll would say this year that that's exciting that that you would want to be created or you know invested upon right what what type of Agents because in my mind I think what we've seen thus far has just been very you know social media related you know or or digging Investments related things like that I guess is there any specific vertical that jumps out that you think

169. 1:23:09

     you know what I think there's going to be a lot of interest moving forward in that specific point and and this is just this idea just jumped out at me so I don't know you guys have any I think the closest touch point would be research isn't it for institutions um it's just the easiest access risk-free um for now um not sure about how AI agents will be integrated with the actual trading for institutions um s stair should be a lot of vertical to be

170. 1:23:33

     sorted out um in terms of like custody or um you know uh just the mechanics and the safety of the choices that they would make um but yeah I think research translation they are just easiest bits for institutions to get involved I mean the reason I brought that up also is the fact that I know near you know we're really looking for a lot of you know Developers really build out these agents and we have like 50 verticals and use

171. 1:23:58

     cases that each of them could actually be their own businesses right I think the problem that we're going to see is that a lot of these are not actually um you know the agent issue it's more of the go to market and adoption which just like it's not about chains being faster it's actually what it's at use for why should they use it right so it's pretty interesting yeah the core of bitg Go's DNA is custody um and security so light

172. 1:24:24

     of recent events um we would probably want to explore how agents can help with earlier detection um risk risk yeah around um you know movements and changing some patterns or areas where we potentially could enhance our current offering to our clients today um but that's the only thing that immediately jumps to mind okay well awesome well I think our time is up so really appreciate you guys being up here as well as everyone else for listening and hopefully enjoying our discussion thank you thank you

173. 1:25:00

     guys amen

174. 1:26:20

     feel time

175. 1:27:23

     feel just want test test hey everybody how's it going so um these are not my demos I didn't build anything uh but but there's an

176. 1:28:24

     entire ecosystem of developers who are and so today we're just be demoing two short things one of them is a browser operator so who here is familiar with like opening eyes operator how about like uh anthropics like computer use agents okay tldr it's an agent that is able to navigate your browser and click things on your behalf it's pretty cool and so we've been working very closely with an amazing team called Silver stream they have built this uh in collaboration with us and I'm super

177. 1:28:57

     super happy to demo the sort of first iteration of it I will say this is not 100% production ready this is just kind of showing you a little bit behind the curtain on what's to come because in the near AI Hub and elsewhere you will be able to have an agent that can navigate your browser for you and accomplish tasks this means you're not going to need you know very sort of vertical specific agents to do things it's much more generalizable and so um you don't

178. 1:29:23

     need to pay $0000 a month in order to get access to this uh once it is actually live and um yeah if you ever if you have any more questions about this Manuel if you can quickly raise your hand and stand up um just this is the guy who built it and him and his team so shout out to him uh and uh we're just gonna play the quick video demo for you right now and so uh video team this is the cloud operator

179. 1:29:52

     agent also know as Sil sylvestro oh not that one Great Click play and so so what you're seeing here is an agent navigating Airbnb and it is going to book a hotel for you and so all it's really doing is you type in your prompts put in the dates put in the times the location and it you put your price range you do have to be a little bit specific on exactly what you're looking for because obviously the agent wouldn't

180. 1:30:38

     can't guess it until it has you know perfect information of what you're interested in and so once you give it the prompt it is going you know kind of frame by frame these are uh bunch of screenshots of the agent working in real time but uh in the future there will be a live stream component so you can just very quickly and easily see the agent you know working for you now a couple

181. 1:31:03

     caveats is that you know it does take some time for the agent to really interpret the page and understand what you're looking for and so uh it's going to get faster and faster over time but the whole idea here is you give an agent a prompt it goes executes you know actions on your behalf but it doesn't necessarily need to be faster than a human because this can run simultaneously as you're doing other things so once you give an agent a

182. 1:31:28

     certain activity or action it can just go and do it now there are some you know blockers this is not an easy problem to solve uh Manuel has been researching this for 10 years uh did his PhD in this field and um this is you know coming soon to the new AI Hub uh the demo is there if you go to Silvestro you can check it out but um overall this is going to be a free open version of near

183. 1:31:54

     AI uh sorry it's near AI operator but a free an open source version of uh open AI operator or any other operator for that matter um so this is really really exciting and soon enough you will be able to navigate any web pages arbitrarily with agents in the near aiub so shout out to the silver stream team for their contributions to this uh this product and you know really looking forward to seeing what comes next so

184. 1:32:18

     thank you and uh next we have a uh demo of Agents building agents because you know that's how super intelligence is ultimately going to happen his agents will start building each other and uh we do have an example of this uh you know happening so if the video to crew can pull up the other this one actually has sound so if we can keep the the room kind of quiet so we can hear the builders uh give the the sound demo

185. 1:32:56

     uh video crew is there any chance that you can turn on the sound otherwise I will have to read it read it agent for you you can also ask him to add some additional features or fix box and uh it can use private or public apis uh and it can store secrets for you so I'm going to ask him to create agent that will qu coin market cap API for some cryptocurrency quods so generate the plan here is the name agent here the development plan now we should wait until it will generate the code so

186. 1:33:41

     the code is generated as we can see to use a API key to access Co Market C API it said that you need to provide the API key to the com Tab and here the agent is deployed so I have coin Market cup key so please use CMC key or here yeah it say the the P key now we can try the agent so gly price of BT comp BTC is 88,000 now let's ask him to support the conversion so turn back here so please update the agent so that you can also

187. 1:34:26

     convert the price of one current one cryptocurrency to um specified you should use convert convert parameter yeah so generate the plan and the code all thank you so what you just witnessed is an agent building another agent that can trade on behalf of a user and this is really exciting and this architecture is really Dynamic and can be used for a lot of other things in the future as well so these are just two demos of some ecosystem projects there are so many other projects building um and actually we just uh got one of our

188. 1:35:06

     PR PRS merged uh with Eliza OS if anyone's familiar with the elizza framework um and so all agent Builders will actually have the default option to host with near AI Hub to get access to free inference and then eventually uh verifiable privacy and uh that's really cool because this is actually just a prerequisite for Mass adoption and so uh next piece and actually this was merged

189. 1:35:30

     today is all Aliza agents can now interact with any near contracts and Swap and manage assets across a near account so that's cool but uh doesn't fully leverage the sort of full extent of near technology using chain signatures and near intents these agents are also going to be able to interact with any blockchain that supports ecdsa signatures so an agent can now manage wallets and accounts on bitcoin on

190. 1:35:53

     Dogecoin on Ripple and everything else and so this is really just a Playbook that we will be continuously adopting across every other agent framework to you know get every agent in the crypto World hosted on near and also enable these things to seamlessly interact across all chains uh I see this as a way to finally check the blockchain box and move on to build things that you know truly change the world so shout out to

191. 1:36:18

     Robert Yan for making that for building all of that and uh making that P request and also the alliz labs team for getting it merged and reviewing it quickly so thank you again these are just some quick demos and there's obviously much more to come so welcome Peter from Hut all right thank you very much Cameron good idea using a screen recording this time yeah you don't want to share the screen any anyway this uh

192. 1:36:44

     is uh is Peter from hot uh he's about to talk to you about all things AI agents but before I give the mic over to you Peter don't forget there is a lot of fresh food over there it's delicious nutritious uh there's coffee there's water there's tea there's swag outside Peter over to you K here Peter hello everyone uh my name is Peter and I'm of hotlaps and for last year we build a lot of in near and especially we build a lot of stuff for

193. 1:37:34

     their MPC infrastructure uh we was responsible for the hot protocol and hot wallet which provided now the biggest community of users on near and the H protocol provide people ability to create MPC while and new protocol that's very great technology and this Monday we added support for the Frost MPC protocol to the Hut and now it support MPC wallet on salana ton um stellar and another uh New Generation

194. 1:38:04

     blockchains so because it's near IDE day I want to share a little more use cases how now ai agents can uh use a hot MPC infrastructure for apic agents so hot MPC infrastructure not only give you access for some technology features it give you access to the huge community of hot users uh we have more than 30 million users right now more than two million active well monthly active wallets when I last time checked uh we're responsible for more than 80% of all Unix swaps on near

195. 1:38:36

     and um this community growing fast multichain we are top two top two application on the prod grou of product you will have assess with your afternic agents with if you support our MPC protocol it's m wallet extension account Marketplace a lot of different products um the great things about hardol that there's already more than three million ebm wallets qu type of H protocol and if you build a e agent it will available to

196. 1:39:06

     work with them um by default every wallet in a hot if it's a ebm wallet is Embassy wallet and it work in a hot protocol some great stuff about it uh it's um with a good stuff about H uh with support more than 120 chains uh there's a pretty cool SDK for developers uh you can call sign API for smart contracts you AI agents can work with our hdk to call directly notes you only have to St some H and you will get

197. 1:39:44

     RPC how many transtion can sign and if you work with a smart on near you can create a limited case for the evm accounts and uh that will able you to only process swaps fers only process Bridges fers and your eii agents will able to control uh limited access to the users e account next we launch our browser extension is going to be live in the next couple weeks and if you're working with a agent you will be able to create AIC Bots trading swabs for users on

198. 1:40:18

     their M wallet and it's going to be completely secure with limited access to their user account we're going to add AI widgets so you if you have ai agent you will able to provide like special statistic widgets infographics to their uh hard application and we are working on a platform uh where which is uh will be backed by hot Omni balance and uh your AI Bots will be able to trade for users

199. 1:40:44

     create intense execute transaction for users you have BS will aable to restake assets and it will be very simple to do because native asdk of H already support more than 100 chains and it support a lot of nonm chains and your agents can just say I want to restak stockens from a to borrow and it's done all code stuff will be processing on the background your AI agents just can make simple high level decision and it all going to work and people be able to track it they will see like you can see this example with

200. 1:41:17

     this chat uh what happened what your B just did why he did some restake did he buy some tokens did his highight what his recommendation to a portfolio thanks um I hope you will all look a little to high protocol NPC and how your AI agents can work uh and bring some great uh benefits for the users uh and looking forward to see more AI agents with h MPC thank you e

201. 1:42:28

     put on okay yeah abely yeah thank you I believe so

202. 1:43:31

     you all right hope you enjoyed that from Peter Peter is of course uh the CEO of one of the largest applications in all of crypto not just neara and the same can be said about one of our next guests uh so please give a big round of applause for ol fenco of sweat and Jeff wilser of AI

203. 1:44:41

     curious thank you it is great dramatic uh fall here thank you it is great to be here thank you all thank you near my name is Jeff wilser I'm the host of the podcast AI curious and the recently launched podcast the people's AI the decentralized AI podcast joined by ol great to be here ol please share a bit about yourself and what your North Star is at sweat a great question um well good afternoon pleasure being here I'm OLG I'm co-founder and CEO of sweat and

204. 1:45:13

     sweat wallet together we call it sweat economy we'll get into that a little bit later and our North Star as a business is to make your active day as valuable as possible effectively it's what are the 10,000 steps that an average person takes a day generates an income to them so we want to make you healthier and weal there by making you move more what I love about oleg's answer if you're you

205. 1:45:46

     know just if you didn't notice this not once did he mention Ai and what I think is cool about this not yet no and we'll get there but there's so many projects and like no shame no judgment many projects are launched and announced as being this AI company and that's great again there are plenty of worthy ones but my personal speculation here is that a lot of mass adoption will see a lot of real use cases of AI will see are a little more under the hood and not in the bright and

206. 1:46:18

     shiny AI objects and I think as we'll see with sweat and ol AI is crucial to your growth and scaling even though it's not on kind of the obvious surface level so we're gonna get to how AI is a main driver of him scaling and from his him scaling therefore the path of billions and trillions of AI agents but on that Journey before we go there o tell us a bit about the origin stories of sweat and then how near came to be involved

207. 1:46:51

     and then we'll kind of get to the AI of it all yeah we started in 2014 and with a project called sweatcoin as name would suggest we want to be on blockchain but it was too early we spent a bit of time trying to Fork Bitcoin but we realized too slow too cumbersome no smart contracts so we spoke into vitalic in 2015 but it was too early and we launched centralized being conf preum so you were speaking of metallic pre ethereum right yeah e wasn't called ethereum back then I don't remember what

208. 1:47:25

     was the project name but it was research grade code that we spent some time on and we realized that you know what it was too early it was just prone to you know kind of some issues and we launch centralized and we were absolutely sure that within six months we're going to find you know if it's not ethereum then there will be another project that we will jump on we scaled extremely quickly by 2016 we already had several million users and it was way too late because

209. 1:47:56

     our transaction requirements became about 200 to 300 transactions per second which would be orders of magnitude higher than Bitcoin or ethereum were able to deliver so by Sid stepping blockchain in the beginning we actually saved the chain that we would have built on and we saved our business because invariably it would have been dead so fast forward every year we were dabling and looking at everything that's happening in blockchain testing various different

210. 1:48:26

     solutions but it only happened in 2022 when bunch of really high throughput chains went into production near salana avalange so we looked at 14 different ones and we chose near actually for three reasons one is alignment and strategy for Mass adoption and really driving you know kind of penetration of blockchain solutions into billions of people second was team which is absolutely incredible and it's an absolute pleasure to work with them third was throughput and very

211. 1:49:05

     importantly No Cap because you know for example if you're building on evm you know that you if you hit 525 transactions per second your fees are going to start going through the roof and everyone's going to be pissed off everything is going to slow down everything is going to become really expensive that's not something that we could afford near sharded architecture gave us confidence that we would be able

212. 1:49:28

     to scale horizontally when we started there were two shards on near now there are six and I'm hearing based on the number of kind of hair spikes on Bo and wanks head that we are rapidly approaching eight so it's been an incredible journey but it gives me confidence in the future okay so you have grown also 2014 like well before any of this was cool you're building this you've increasingly got decentralized you have how many users roughly now in web 2 we have more than

213. 1:49:59

     180 million registered users what 188 Z yeah okay in web 3 we have nearly 20 million token holders of sweat which makes us I believe ninth widest hell token in the world and we have roughly three million active uses per month close to 2 million onchain active per month I'm so glad you went there and shared that because again as I my thesis is this might be the closest we'll get to approximating the rest like the non

214. 1:50:30

     web3 world like actual like big companies corporations and that is kind of like getting those folks to embrace decentralized AI is the way I think we'll eventually in my own personal view we'll eventually get to quote trillions of AI agents so let's now pivot to you're not an AI company you have been like for a decade you've been scaling how are you now starting to use AI for your own growth and then we'll kind of get into to what extent you're embracing

215. 1:51:01

     AI agents and how you see that fitting into the larger ecosystem of kind of growing the AI agents Universe there are several questions there and I'll try to unpack them um you know our approach to building is actually not idea based but a problem based so the problem that we're solving is to make the World physically more active everyone wants that and there is nothing simpler than simply tokenizing and you're rewarding you for every single step yeah let's give a quick quick quick

216. 1:51:34

     like primer for those who have not yet bought into the sweat sweat lands give us like a 20 second Pitch of what for now average user what do you do what do you get how does it work yeah I mean sweat is the value of your physical activity tokenized we all know that our physical activity is not zero value right close mine's pretty close to zero but I hear you in concept I'm pretty confident that if somebody would offered you to go and run a marathon and bring

217. 1:52:05

     back some beer you would tell them to go screw themselves because you value your time and you value your calories and you value your movement there is there is value there we feel it intuitively and what we're doing is we're tokenizing it and giving that value to you to Che the POS it with attention economy so movement economy what we're building is around the value of you been active attention economy that everyone's familiar with7 trillion dollar economy by the way works in the following

218. 1:52:35

     fashion Google takes your attention sells it to the highest bid puts all the money in the pocket and all you get is two gigs with a Gmail y got it so so so short hand I got and I I downloaded all getting aside I have the sweat app on my phone as when I here when I walk out of here it's capturing my steps I'll be rewarded ultimately with sweat tokens all right let's pivot now to you and these tokens have utility and we've been deflationary because I very frequently

219. 1:53:07

     get this question so how are you different from other move to earn apps well very simply we've been added for 10 years we're still around our web to business is profitable for years our web 3 business is on path to become profitable and we know how to build stainable tokenomics because sweat was deflationary since July 2023 so we are really building for longterm and that's why I'm saying that our goal is to make

220. 1:53:33

     the world physically active and to make your active day the value of an active day that you get as high as possible that will bring billion people and it will make them more physically active so when did you first bring an AI into the equation and then what's it doing for you now yeah so coming back to we're solving a problem not starting with idea which basically means we are looking at every Tech that

221. 1:53:59

     comes our way as an opportunity to enhance our product make it Slicker faster better simpler to illustrate we couldn't build on blockchain WE launch centralized when blockchain technology became fast enough and robust enough we deployed it you're pragmatist you're more into Solutions than an ideology yeah sure I you know kind of Anything Goes as long as it pushes Us in that direction then a came along we deployed our first AI product which is an MVP in

222. 1:54:32

     November last year with the sole objective of just simply starting to test and understand how to talk about it how to position it and guess what we had 700,000 questions asked through that Ai and when we were developing what we're about to show you we did not start from scratch we didn't start with some hypothetical ideas like what would people be interested in we basically just looked through 700,000 questions we brought the list of four to five that

223. 1:55:04

     pretty much everyone was interested in and then we pushed the boundary to integrate these answers so that you're not looking just at text but you can actually take an action right away from your personal AI agent answer which means that we can remove a lot of friction we can remove a lot of functionality from our app for example Dow voting is used four to five times a year we don't need to have a graphic interface for it in the future it is going to be run through chatbot and llm

224. 1:55:35

     so simpler only core things and everything that is used occasionally can be managed through personal AI agent amazing and was there something you wanted to show or share is this yeah gentlemen is there a chance that you could run that demo I I have to say that it's working I have it here the problem is that there's no cable that you know you can see his phone right everyone can see us clearly and it's yeah it's it's true if in doubt come and you know approach me after this uh and I'll show

225. 1:56:08

     you that it's actually working okay I so what uh so there is a thing happening here okay great her name is Mia movement in action so you see there are those prompts which is four or five questions that are most frequently asked so this while we're watching this as I understand at OLG this has spared your team the expense of creating FAQ pages of creating explainer videos of as I'm sure many this F room folks in this room know all of the content and tasks

226. 1:57:04

     you have to do to deal with massive onboarding that has dramatically been reduced thanks to this AI solution is that right but most importantly it spared our users time and effort to go and find those fqs filter through them and read text they ask a question exactly how they want and they just get a very very concise answer and what you see there is it's not just a concise answer that you're getting for example a lot of our

227. 1:57:35

     users are asking so what are grow jars and which one should I choose basically the answer is there are two types of grow jars step-based jars is where you earn on the basis of physical activity and if you want to create one you just press a button and it gives you a a transaction you just put an amount and you confirm and it's done so EAS interface easier to find information and it allows our users not to waste time

228. 1:58:05

     you know trying to go to YouTube and watch some videos or go to FAQs and help pages one thing you shared with me off mic is your idea and let me know if I'm getting this wrong is that to get the true billions or trillions of AI agents in a mainstream way that actually has to come from human beings can you elaborate on this what you mean by that and then kind of how it relates to both to you and near and then maybe advice you have for a larger uh space that's kind of

229. 1:58:39

     developing all these tools yeah um I believe that at this stage of AI development everything will start with humans even deployment of an agent ultimately somebody is going to have to sit down write some code or you know give it a prompt and swipe to create it and give it some budget so in order for those trillions of agents to appear we either need to have a lot of Builders deploying crazy number of Agents or we can give our 3 million active users

230. 1:59:17

     personal AI agents train them better make sure that users build confidence and then build into this personal AI agent capability for them to deploy AI agents on near AI directly from chat prom uh Pro uh directly from the um agent prompt I think that that is going to be an extremely interesting way for people to experiment and we're going to see definitely an explosion of Agents created this is not what we're going to

231. 1:59:48

     start with as an MVP however this is something that the team is is extremely excited about I love that I think it's a nice counterweight to the idea that we'll just have some awesome code and then poof these agents all appear to your point it kind of starts with onboarding humans for some use case which You' been doing for a decade and the Agents flow from that absolutely and look AI is there to make life of our

232. 2:00:16

     users better faster ultimately I believe that we will be able to farm out to this agent a lot more than just the chat for example all communication will go through chatbot on boarding will be done with llm and chatbot and we're not starting with an agent able to do transactions for you because a lot of our users are feeling nervous about this however that is a destination especially with near account structure I can foresee that I would give my personal AG

233. 2:00:52

     agent an authority to spend let's say 100 bucks swipe it off and see what it does without worrying that this agent can potentially drain all the assets that I have on my account great hopeful not to end on thanks again sir enjoyed it thank you all thank you very much Jeff I I to feel you

234. 2:02:08

     bird fly by I see the F the Twilight feel the warm of your hand in M it's in the air inir in the air I to feel you stand

235. 2:03:25

     feel you standing to good awesome hey everyone uh I'm John Peterson from coinbase uh I work on the coinbase developer platform I am the uh engineering lead for agent kit and work on several AI initiatives uh within the developer platform today we're going to go over agent kit uh some of the CDP products and how to deploy agent kit

236. 2:04:31

     agents on near AI agent Hub uh so we can get started so coinbase is enabling onchain agented Commerce there are already over 20,000 onchain agents running on coinbase developer uh rails the combined market cap of onchain agent kit agents is estimated to be over 4 billion USD today and we're deploying more and more agents uh from our developers every day so first of all what is CDP CDP is coinbase's developer platform our vision

237. 2:05:04

     is a million developers building onchain for a billion users how do we do it uh we're making it simple to build onchain apps and we're making it even simpler for your users to use onchain applications so just a quick product overview for CDP here uh we can split into to three main categories we have a payment stack which includes uh smart and API wallets onramp and off-ramp uh

238. 2:05:29

     send in receive staking and yield a set of onchain tools including node faucet pay Master data and verifications uh and then we have our development Frameworks which we're going to key in on a bit more in today's presentation mainly onchain kit uh CDP SDK as well as agent kit for AI agents notably all the CDP products are meant to be used uh you know as a you know additive and will be able to be used eventually by your AI agents uh

239. 2:06:00

     themselves autonomously so next we'll talk a little bit about onchain enabled AI agents at CDP we believe that every AI agent deserves a wallet it's impossible today to give an AI ownership of a bank account but it's very straightforward to give an AI agent a crypto wallet uh and it's even easier for the agent to per uh you know to perform uh internet native permissionless uh and near instant rails for payment and data transfer using that wallet and using the agent access to

240. 2:06:32

     blockchain uh with onchain actions we can provide even more powerful tools to these agents uh you know to create their own tokens perform borrow and lend uh and onramp offramp from Fiat into crypto so now let's talk a little bit about the current crypto AI landscape uh and so we've split this roughly into three main categories you have ai agents which you can think of as the like consumer applications Frameworks and hubs uh like Eliza game uh several

241. 2:07:03

     others there and then the lower level infrastructure required which is namely uh the data layer that's training data and then the AI uh uh foundational model llm inference layers and so what is Agent kit agent kit is a developer toolkit for building and deploying Uh custom AI agents with preconfigured environments and ready to use templates uh agent kit empowers you to give any AI agent regardless of framework model Network or wallet provider access to a crypto wallet

242. 2:07:34

     across any uh any network with Native support for evm and svm networks so where does agent kit sit within the agent the AI crypto landscape today it gets sits roughly in between the uh Frameworks and hubs and the training data and AI inference it's primarily a toolkit and set of uh core Primitives and actions that allow agents to have uh their own blockchain wallets and interact on chain why Agent kit uh agent kit allows

243. 2:08:06

     you to unlock new agented capabilities this lets you Unleash Your agents on chain with new super superpowers uh and allows you to transact with programmatic transactions interact with smart contracts and send any agent or human uh crypto or onramp and off-ramp within Fiat it enables new business models it allows you to send dollars for free own a new distribution Network and interaction Paradigm uh and it allows you to drive AI adoption and growth by accessing the onchain crypto developer

244. 2:08:35

     community and the AI Community as well uh we also strive to be uh to set the industry standard uh for early adoption for AI applications with integrated crypto uh Payment Solutions in a secure and compliant manner in accordance to coin bases standards for data and privacy security you main use cases can be broken down into two categories uh Human Assistance which is essentially human in the loop agents these take uh natural language commands uh they're relatively intent based and the way that they

245. 2:09:09

     interact is human in the loop human asks the agent to do something then the human interprets that via the llm inference and then uses agent kit to uh interact on chain or interact with any other API that your agent may have access to the next set of use cases are autonomous agents uh autonomous agents allow you to interact without direct human intervention these agents are uh you know deployed on an agent deployment Hub deployed directly on chain uh or you

246. 2:09:40

     know deployed locally and they allow the agent to run freely based on their you know uh predefined set of goals and actions that they want to uh you know want to perform to to achieve uh achieve uh their directive so one small aside here also the best use case for AI is AI robot dogs with crypto wallets we will have our own CDP robot dogs tomorrow at the CDP eth Denver event they'll be there all day you will be able to interact

247. 2:10:11

     with them powered by agent kit you can uh ask them to do tricks do all kinds of different things and you can pay them and they'll actually accept the the money and validate it within their wallet with an agent kit uh bit of a side note but please come play with the Dodge tomorrow they're a lot of fun so what's coming next for agent kit uh we want to support more AI agentic actions across all chains uh we want to support uh a vast variety of defi borrow

248. 2:10:42

     and lend uh you know protocols on evm svm uh and any other networks that you're willing to contribute to we also want to support a vast variety of wallet providers we currently today have uh access to local account beam wallets uh CDP API wallets CDP smart wallets as well as privy server wallets across a variety of networks and we want to expand that uh you know to include any wallet that you may want to interact with we want your AI agent to be able to

249. 2:11:13

     use uh you know natively and seamlessly as well uh we're open source friendly and are actively seeking open source contributors so feel free to uh to contribute anything that is interest to you or that you're building locally for your own AI agents I'm sure other people may uh be interested in using them as well um on that point next thing that we're focusing on is making it trivial to PR

250. 2:11:37

     any onchain actions you want uh we want to make the open source contribution as seamless as possible uh and then finally we want to focus on progressively decentralizing the stack to support fully autonomous agents in a more native and seamless way both from the wallet uh and the Tool uh you know the tool standpoint so we're the to developer platform how can we get started uh here's some QR codes to get started with our developer Discord as well as the developer platform documentation for agent kit feel free to go there there'll

251. 2:12:09

     be quick start CIS that can get you set up uh with projects locally and then once you're ready to deploy your agents and get them out in the world uh you providing real value to users uh we can move on to deploying to an agent Hub such as the near AI agent Hub uh here on the left you can see the agent Hub on the right uh we have a QR code going directly to agent Hub please try it out

252. 2:12:34

     and deploy an agent kit agent uh and so now I'm going to walk through just quickly the steps involved in deploying an agent kit agent uh via template to the near AI agent Hub so first stop is in the top right there's a plus can click on that uh once you're there it'll bring you to this screen for deploying a new agent uh you can se you can select agent kit by coinbase uh from the set of templates to include agent kit uh within our repo provides a series of templates

253. 2:13:06

     that you can use to deploy directly to agent Hub to get you started uh this provides a base wallet provider using the CDP API wallet as well as a set uh core set of agent actions to get you started uh interacting with uh some of the top defi protocols to do things like swap borrow Lind faucet on testet uh wrapping to tokens interacting with erc20 is and 721s Etc uh once you select the template uh you can put in a name for your agent and version it and then you can click deploy

254. 2:13:41

     this will deploy your agent uh onto the aiub and your agent will begin running autonomously uh from there if you want to go a step further and actually uh develop your agent pass the template you can use the near AI CLI to uh pull down the uh uh python module for your agent from the registry uh from there you can navigate to the downloaded source code and you can pretty much just hack away you can add any wallet provider any

255. 2:14:11

     chain uh any action that agent kit supports uh and then finally you can republish your changes again using the near AI CLI back to the regist and your agent will be updated uh on the deployer Hub and will be able to use all of the tools uh that you've added uh feel free to use any of the uh you know default actions that we have Implement any of your own locally or through open source contributions uh the you know the world is pretty pretty pretty open here you

256. 2:14:42

     can do really anything you want with these agents and so I hope that agent kit is able to empower you to build really powerful applications and get agents that that are out in the world uh contributing real value both to the web 3 and web 2 communities uh that's all for me thanks everyone uh happy to take any questions that you have but if not thanks say what all right ladies and gentlemen um

257. 2:15:27

     there's still time to grab the food grab the drinks grab the swag grab the shirts uh do whatever you want that makes you happy but now we have uh a question we all need to ask ourselves and that is uh what if defi just worked well Paul you're gonna tell us the answer right big round of applause for Paul everyone all right so I guess the premise of this is a um crypto is supposed to be the financial rails of the future and clearly as of today it is not

258. 2:16:08

     um it's supposed to give people back ownership of their money their financial rails what they can do with it um but somewhere along the the way we made a tradeoff and the trade-off we made was decentralization for chaos and uh it's going to be no surprise to anyone that my solution of the for defy in this talk is going to be Ai and agents um if you're more of an active listener go ahead and scan the QR code and play with some of the dozens of

259. 2:16:38

     Agents live today um but to start we're going to step back and we're going to step back to talk about Automation and finance and what people expect to be able to do with their money because people don't want to actively manage their money they don't want to be spending time every day tracking Investments making sure payments went out they want things to just work and in order for that to actually happen we don't need more uo UI we don't

260. 2:17:11

     need better education that's ask asking people to come into crypto and learn and do more than they already do in in whatever fintech apps and products that they're using today so what does an execution layer for crypto need to accomplish well it needs to act on behalf of users the way most fintech apps do today it needs to protect them in real time so send fraud alerts make sure people's accounts are protected not force everyone

261. 2:17:42

     to have a ledger and and do all this crazy stuff it it needs to integrate seamlessly but of course it needs to stay de centralized so um we've been looking at agents for helping with this since 2023 um and the basic hypothesis at the time was text to transaction was going to help us do this it was going to help us enable people's Grand parents or friends to start working with defi in a way that wasn't asking them to do more

262. 2:18:15

     or learn something just use a product uh and obviously it's not easy but we've been building uh this now for over a year we have um taken the approach that many many folks have which is a terminal with many agents and and a registry where people can go and start a conversation in this well the video might not work but in this video oh here we go we have a um a user who goes in has a conversation with the coin gecko

263. 2:18:47

     agent to get ideas on trades and potential Investments they pull up some charts they get excited about virtual and then they just say you know what I want some virtual in my account so they enable a swap and the agent performs and builds the transaction asking them to sign from their wallet and then it's done they didn't need to learn about where they should go by virtual they didn't need to

264. 2:19:11

     have three deps or sites open at once it was all in the chat and there's a couple of other agents that work here there's a d agent that lets you ask what are all the Dows I'm a part of proposals I want to vote on and so forth there's a an agent that lets you launch a safe on BNB like there's a lot of cool stuff but we also realized that this is probably not the answer this is for people already doing the stuff and if we truly wanted to make an execution layer as seamless

265. 2:19:44

     as possible we needed to provide tooling where it already was so am I going to be able to click next here no all right we'll let it run out but you can see these agents at work but the premise that we came to was if we want agents to actually be onboarding users to enable them to forget about defi we have to go to where they're currently onboarding we know today people are using uniswap and products like that and telling people about their earnings but they're not realistically

266. 2:20:19

     able to hear about to swap if you're a new user and start trading it requires a whole bunch of things so while we started with this registry approach we actually realized we needed to embed this into every dap that exists today and superpower in charge them to basically on board into any app without knowing crypto at all and our approach to that is embeddable chats so any dap today can embed agent chats directly into their product and choose

267. 2:20:48

     from hundreds of different tools to create a micro assistant that's perfectly designed for the use case of the D my personal most exciting use case is dex's where people will be able to one on board to the decks easily figure out how to connect wallets and accounts uh learn what's available on the decks bring in research tools and then have their accounts and trades passively managed for them and um we do have all of this public today if you want to embed the chat we had our

268. 2:21:20

     uh first embedded chat a few weeks ago shout out to a growth mate on the near Network yeah and a uh we have our first Dex embedded chat coming in the next few days hopefully fingers crossed um but I guess I'll leave you with this in order for us to bring the next hundred million people into crypto we can't ask them to learn things they don't want to learn or claim that there's better ux and

269. 2:21:47

     education needed is has gone on at this conference now for the seven years that I've been a part of it we need to automate away as much as traditional Finance does while preserving uh decentralization and autonomy for those who need it so if you want to build with us we're not doing this alone we have lots of model Partners we have lots of a um a data Partners uh we're partnering with more folks every day and if you

270. 2:22:13

     want to embed agents into your dap to out compete those who aren't going to embed agents into their products let us know and let's help you out thank you all right uh ladies and gents we're actually a little bit ahead of schedule for the first time in history ever uh so if you do want to go and drink the high fructose corn syrup or the coffee or you know whatever food we've got please do go ahead and indulge there's still a few

271. 2:22:51

     sweatshirts left um there's there's cups Etc pins but I would advise you not to go anywhere because I'm about to introduce one of my most favorite developers in the entire NE ecosystem his name is Josh he's going to do some uh wonderful demonstrations for you so Josh please come up on stage everybody give him a big round of applause for Josh for thank you David you're so kind one my favorite MC's in the

272. 2:23:29

     world um let's see I need to hook up my laptop I'm GNA do this HTMI here we'll do and here all right let's see my entire screen I'm going to start mirroring let's see is it doing it this is always this is always the problem I have a a history of being cursed with dongles and HDMI cables yes yeah is it your shirts on mirrored uh yes it is UN mirrored I just selected mirror yeah plug unplug plug it back in

273. 2:24:29

     yeah all right different dongle every time I swear shout out Noah and Laura my brother and my wife they're watching right now we sweat because they think it's funny oh yay all right now we're cooking all right all right all right cool cool cool um so yeah hey my name is Josh I am working with near Dev um we do development relations and a whole bunch of other things for the near ecosystem uh today

274. 2:25:04

     I'm going to be talking about optimizing your near AI agents uh using something called reg or retrieval augmented Generation Um and you may say why why do I need to uh use this rig and uh it kind of helps a lot of the LM shortcomings uh part of that is preventing hallucinations these AIS aren't really giving you the right information all the time um and it's going to help prevent that from happening um it's also going to show solve um short contact Windows

275. 2:25:33

     there's only so many tokens of chunks of data that you can give an llm or a model during a prompt and um you may have a larger amount of data that you want it to parse through to feed it um to this model um but what I'm really uh Keenan is it's a key component for user privacy um and that's kind of near theme is you know user owned internet user owned Ai and all those um core values that we try to really move towards um it's a great article by uh pine cone they're um a

276. 2:26:07

     vector store solution um but it talks about yeah why rag must be a part of your data privacy plan and um some cool you know tidbits out of there is it really enables control of your data um you could control the data that's provided to the llm at Generation Um and the data does not need to be stored in these models you can have it separate you don't have to feed it and train it

277. 2:26:29

     and and trust that all these things are going to you know hold your data securely um you know and it's shared as a per request basis you don't have to again have it on there all the time and then it allows for quick removal from the system so like earlier we're talking about tees and trusted execution environments you know the ability to give it some information run something on it prove that it wasn't exposed it all in any way and then shut it down and

278. 2:26:53

     wipe it out it's yeah just totally awesome uh and puts me at ease using some of these uh AI systems um so how it works is kind of like an overview um you basically have you know you as a user or an AI agent even could submit a query um you have this retriever um system that's in place there's a whole bunch of different types of ways that you can do this but it's kind of a simplification of it you have a retriever and a vector

279. 2:27:18

     store that stores all this data it finds the most relevant content and then it brings it back up to the retriever gives the query uh plus the original query plus the context this data that it found to the llm that then will give you back the information to the end user so let's do a demo um recorded a demo we will show how to quickly make an AI agent on your AI uh simply run near AI agent create it'll walk you through this um prompted tutorial you just enter

280. 2:27:50

     an agent name you enter um the uh description of your agent and then the initial prompt so however you want to have it initially set up and later we can um modify that information later it says hey is this what you want to do proceed and then it gives you a bunch of um uh options it shows you where your code was stored and then some useful commands about how to run it locally up upload it to the agent Hub or run this agent from your local registry so what

281. 2:28:14

     we're going to do is we're going to copy this little nugget and we are going to copy paste past the uh copy pasta in our terminal and then we're going to ask it a question um so what types of access keys does near protocol have which is a really cool feature about near um and it gives us this answer which is kind of interesting um full exess Keys yep function exis keys but this thing uh it's not a thing view function access Keys it hallucinated it doesn't that's

282. 2:28:40

     not a thing we don't have to sign for transactions view calls are free um so it said oh we have these Keys We like all right that's kind of interesting um let's ask it another question how many many keys can an account actually have um and it thinks about it and it also hallucinates again it gives you 128 which is also not correct at all um it thinks that can have 128 Keys it's a lot of keys but it's actually not the answer

283. 2:29:05

     um so we'll go back to it and say you know are you sure like again what are the the types of keys uh that that you can have on near and it's like oh yeah I forgot there's uh these other things as well there's like a function called permission access key and a method called permission access key um which again do not exist at all uh it's just making it up and I think it's like deriving it from this permission access key thing that it thinks uh it's about

284. 2:29:29

     um and then I start messing with it I'm like are you sure those are all the keys are you sure there's not more types and it's like oh yeah I forgot there's actually all of these types you have these storage access keys for reading storage and Standalone and Method call and yeah so it just keeps going and going and it's hallucinating it doesn't really understand and to its credit it's trying to figure it out it says it apologizes for its responses I tell it's

285. 2:29:53

     it's okay AI you're fine um but yeah so it's trying to figure out what exactly it you know the information is based on what it knows from this open source model um so what we can do is we'll exit out of there and we're going to run into um we're open up an IDE and we're going to check out the code for this agent um and all agents are real simple um I just opened up my registry I have my different accounts in there I'll go

286. 2:30:18

     under my near AI Dev near account and I'll open up this agent that I just created it's two files you have your agent python file metadata. Json file and this is basically the how you configure your agent um you can see it's the Llama instruct is our model and then Max tokens it's kind of important talk about that later um but yeah this is basic the code for your agent super super simple and it just did our initial prompt so what I'm going to do is I'm

287. 2:30:43

     going to go to docs. near. I'm going to check out a new tutorial that we just pushed out was this documentation rag um and in there um it'll get take a step by step for all this stuff I'm going through it quite quick because I have limited time um but yeah what we need to do is we need to give context to this model we need to tell it hey this is information so this is um two documents

288. 2:31:07

     from official doc.org about the protocol um and we're just going to give you some commands so you can um download that easily to your local um your local agent and give it some context um so we're going to grab these markdown f files uh I'm just going to change directory real quick um go into near AI uh. near my username in my agent and the version number just going to download these two

289. 2:31:30

     files real quick and you can see I have now these markdown files that are both access keys and CLI uh commands the CLI is also part of the tutorial that we're not going through this um uh update or this demonstration but yeah the other thing we're going to do is we're going to update our agent with our code is to load those files you can see in the very top um here's two files take a look at those and what we're going to do is use those files to um uh with our

290. 2:32:02

     query from the user um and then use that to to um answer the question um and often times I forget what agent I named it so if you just run a agent in uh near agent interactive gives you a list of all the ones that you've created I'm going to create number three and then I'm ask the question again how many access keys are there in there in near um take a second to uh respond and it

291. 2:32:26

     did it correctly this time now it's got actual data to do it there's two types uh there's full access and function call it's cool that like there are a lot of uses that you can do with function call you can like really scope them down to a whole bunch of different things and I think that's what the model was trying to figure out it's like actually with this function call you can create

292. 2:32:42

     different types of keys but ultimately there's only two um so yeah we'll um yeah we um went went over yeah basic full and function um we'll say hey how many keys can uh an account have um and I believe it'll it'll it'll spit out um you can have multiple but there are two main types so it's still not hallucinating on us it's still telling us yes these are the right ones it's just these two um and then like okay

293. 2:33:07

     well what's the limit for the number of keys that I can actually add to an account and see if it'll throw me something off like the other one did and um yeah it tells us yeah there's actually no limit theoretically There's No Limit um you might run into I don't know like TR do like a million keys on an account I don't know haven't really tested it but theoretically you could do almost near infinite amount of accounts so yeah um basically that's kind of a

294. 2:33:30

     simple way of how rag works but if I go back to My Demo slide we skipped one part and that's this Vector store basically what we did is we just took these markdown files and gave it to the llm and gave it with the prompt um that doesn't scale there's a big problem with that and you can only have X amount of data that you can feed the the model um when you're asking it a question so what we have to do is use something called a vector store the vector store

295. 2:33:55

     what it does is it takes your raw documents it like chunks them up and embeds them in these Vector points uh and they all become these numbers and then we store them in a vector database so the chunking is basically just going through your document it's grabbing this these sections and doing a little bit of overlap so that way um there's a little bit of context with it and you can adjust all these settings um within your uh within your program um and then it plots it out and it's it's like all

296. 2:34:20

     right we have all these array points here are all the DAT here all the markdown files or whatever files that you're pointing out and it's like all right here's the question how can I create an account and how you know what's closest to that answer and then it grabs about a few of them and then it returns it back to that individual um so yeah basically let's go through real quick on how that part Works um

297. 2:34:45

     basically all we're going to do is go back to our tutorial we're going to go to Vector stores um scroll down a little bit again goes through all this a little bit more in depth if you want to learn more um and we're going to grab the entire docs data set from n.org so grab the entire directory now um which we couldn't have done before because it would have ran out of um uh um it would have received a Max token limit and would have errored

298. 2:35:10

     out don't need agent thatp or I'm sorry we don't need our two docks um anymore because now what we're going to do is upload it into a vector file also we have the vector file code for you so we'll just create a new file vector. py add these um add this little Boiler code um code and yeah basically what we're doing here is we're just running this um Vector store it'll start grabbing each one of these files that are in this data set it's abouta like a hundred or so maybe more um and it's chunking and

299. 2:35:41

     automatically embedding them for you you can do the C you can also customize this um however you'd like we can show we'll show you that in the tutorial if you go to doc.ai um but yeah basically what it's using is like an open AI compatible client it's not actually open AI it's just the client that we use um it's actually uploading it to the near AI platform um and yeah it's just looking through these entire um markdown files and it's grabbing all the information

300. 2:36:06

     that we have there um and then in each file taking in uploading it and then we're storing it in in these chunks and then now we're going to have this ID and that's really key is that now you have the Spectre store ID that you can now use with your query when you're trying to get your information from so we'll grab that little ID we're going to go to our rag agent um and yeah we just shipped this uh earlier today we're actually embedding our agents now in

301. 2:36:30

     documentation and this is an example of a live implementation of it want to get some docks out on how to you can embed your own agent in there uh as well um but yeah we'll just go ahead and uh update our agent. py file to now use this Vector store and we're going to copy pasta this snippet code and here we're going to go through and um just update the vector store ID so we grab

302. 2:36:54

     that ID that it gave us earlier and then we'll just copy and paste it over our our own Vector store ID um but this part is pretty easy it's it's just like you know here's your query and then you have your vector results you just go and you know you query the the the vector store and then you can see here we have six so like give me six chunks that are as close as possible to the answer now we'll take that send it over to to the

303. 2:37:19

     model and it will uh answer the question hopefully um in in the best way possible um yep and it just returns an answer so it's all pretty simple you know python script it's not really too crazy I think anyone honestly could probably figure it out even if you're not a developer I'd be happy to talk to anyone that's not a Dev that wants to try this out after um after but yeah um we're going to go

304. 2:37:44

     ahead and now run our agent um and now we can ask it more stuff we can say hey what languages does near support for smart contracts and now it knows more information um it knows that hey there's actually rust and JavaScript that are two main smart contract SDK we're actually working on a python one right now um it compiles it to web assembly knows some more information on that and it's actually giv us some boiler play code um to get started building smart

305. 2:38:06

     contracts on our platform and then it gives um some links to the SDK documentation um below so yeah that's kind of like a speedrun obviously through how to use rig how to implemented how to you know build these agents and um yeah as you can see we used that Vector store we got the content and we following uh this diagram and I guess you know I have just a couple minutes left and what I would say is you know near in a sense we're kind

306. 2:38:36

     of moving from one let you know people like oh you know what are you guys working on what are you focusing on at the end of the day we're just trying to build a better internet and build one that is user owned and basically the internet Belongs to You belongs to the people that actually use the internet internet um not to a couple of Corporations um you should be in control of your data you should be in control of your assets you should be in control of your identity no one should own that but

307. 2:39:00

     you uh and you should be in control of your choices um it's one thing that you know I think the internet was awesome and kind of veered off course and I feel we really need to make a course Direction back into this realm um so yeah and that's kind of why I'm here and continue to contribute to this platform so anyway Let's Build a Better web together come join come talk to us try it out for yourself scan this takes you

308. 2:39:25

     to doc.ai come talk to me after this talk and um yeah that's pretty much it uh thank you very much Josh thank you so much as always okay ladies and gents believe it or not we finished ahead of time right right for the like the first time in the history of ever uh but the food will be open until 4:00 so if you want to go and uh fill your belly feel free there's coffee there's there's lemonade Etc um a couple of important tidbits if you are

309. 2:40:05

     expected to come back tomorrow for the near AI agent games which is going to be super fun and super exciting that doesn't mean you can leave any of your gear behind there's a separate event on tonight so take your laptops uh take your take your phones uh make sure you don't miss any of your Tinder messages Etc uh Etc um and there's one other thing I needed to mention what was it a

310. 2:40:29

     survey that's right so you're going to get an email it's going to say hey um how beautiful how much how much weight do you think that the uh the MC lost in compared to uh redacted to here you need to answer that question with about 40 pounds thank you very much thank you very much ladies and gentlemen yeah well it's all the power of OIC um seriously be honest with us um give us your real genuine feedback it's going to help us improve for the next time and uh if you're going to be here for near AI agents tomorrow I will see you then

311. 2:41:00

     until next time God bless thank you very much byebye baby down and high I'm in your