NEAR AI Ecosystem - What Did You Ship This Week 4 - AI Agents That Build Agents? NEAR AI Dev Standup

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

1. 0:17

   Near AI, what did you ship this week? Hey everybody, thank you for tuning in. My name is Cameron Dennis. This is I believe the fourth or fifth. What did you ship this week? Uh this is a pretty much a communitywide near AI focused uh really commu standup engineering standup for a lot of the builders uh that are kind of working on a bunch of really cool stuff. Um I also realize that I never really give my updates. So at the end I will be giving a bit of an update

2. 0:49

   on what I've been focused on. Uh but without further ado, we have an amazing agenda today. Uh I'm just going to share my screen real quick. And for those who will be sharing screens, uh please only share the window of a screen uh and not the entire screen. Actually, turns out here we go. Present. Um just going to walk through the short agenda that we have. It's only today we only have about five of five presenters talking about a bunch of different stuff, but uh give me

3. 1:19

   one second. Great. So, uh, kicking it off, I'm actually going to walk through some of the things I've been working on, then pushing it back to Alex from the Near AI team to talk about AITP and MCP and some of the other stuff that he's been working on. Uh, then Carrill from Datrix talk about agents building agents. Uh, Shashank from the fraction team talking about some of the progress that they're making on their crowdsource data

4. 1:45

   labeling and uh, sort of game that they have. and then passing it off to slime to talk about in ear uh sorry in ear and then Zahed from Jutsu to talk about the uh TypeScript SDK. But without further ado, this week has been much more of an event focused week for me. Uh not as much about building anything really cool or exciting. uh but kicked off Monday with an amazing conference in San Francisco uh called the open source AI summit and that was with uh Erica Kang.

5. 2:20

   She is actually the the head honcha who made that happen. Uh it was really awesome. We had speakers from Google DeepMind other authors of attention is all you need. Over like a thousand people came through throughout the day. Um I just was sort of helping with the uh organization of all everything. But most importantly, this weekend, I'm very excited about a hackathon that we're

6. 2:39

   running called the Useful Agents Hackathon. Uh been cooking this up for the last month or so, and we have a really amazing uh bounty partners, over $20,000 of prizes. Uh Coinbase is sponsoring, we have uh Nevermind, we have uh Silverstream, we have a bunch of people who are actually building useful agents or agent frameworks uh or get getting together to really just get useful agents built. This has really nothing to do with crypto. I will say it's more having to do with uh just building quality benchmarks and eval

7. 3:10

   frameworks for uh agents because right now we think that one of the biggest problems in agent land is people are building really cool stuff but there's not always a great way to assess its success. So uh this weekend is all about assessing success of useful agents and we'll be sharing more updates uh early next week. Other thing that is actually one of the cooler projects that I've

8. 3:33

   been working on is a auditor agent uh for the near AI hub. This is an agent that will check for security uh considerations and compliance standards and everything else for other agents in the near AI hub. And I sort of see this as the first layer of discovery. And the whole point is so uh if you're connecting with an agent in the hub, you know, it's not going to, you know, screw you over in any way, shape, or form. Uh so this is a very important initiative.

9. 4:01

   I've been working with a lot of top tier security firms to establish this framework and build this agent. So this is one thing that I'm uh tangibly working on. And then earlier this week on Monday, Alex Skdanov gave an amazing paper club. We do this about once a month in San Francisco. All of it is recorded. So you can go onto the near uh protocol YouTube and learn more about S1 uh simple test time scaling. It's a

10. 4:27

    really cool paper. Uh pretty amazing to see like more AI native people coming through uh near AI these days. This is really kind of what we're building for. Uh and this paper club is a sort of a great way to kind of stay on top of the most cutting edge research and know we do share this. It's totally available for the public. So um check out the recording. But without further ado, going to move on to our next speaker. Uh

11. 4:53

    we have Alex Komaford from the Near AI team and he will be walking you through what he's been working on. Alex, bring you to stage. Right. And uh before diving in, can you just quickly introduce yourself, uh what you shipped this week, what you what team you're working on, and uh maybe a short fun fact. You have about 10 minutes. Sure.

12. 5:16

    Okay. I'll start off. Um I'm Alex. I work on I work in the near AI team and specifically subp part of that on payments and specifically AITP the protocol that we're developing for agents to communicate with each other uh exchange products and services with each other as well as payments. So what I'm going to be demoing for you is a bit of a side pet project that helps the near ecosystem as a whole and um is sort of like a proof of concept of showing the

13. 5:46

    power of MCP servers. So for those who don't know uh MCP server MCP is an acronym for model context protocol. Um this is a specification really that was released by anthropic not too long ago. um specifying a more or less um a way to define APIs that thank you for sharing that. Um more or less a way to define APIs that will work well with LLMs. So um at near we believe like very strongly in MCP and it's very easy to

14. 6:24

    get started with with setting up an MCP server. So sort of as uh something that we wanted to develop is actually how can we make the near blockchain also an MCP server or specifically how can we make even the near CLI command an MCP server so LLMs can directly access the near blockchain. So that's what I shipped last week and I'm going to show it to you now. So I'll share my screen. Can I

15. 6:50

    take over? Uh, let me share my whole screen. Hopefully everybody can see it. Okay, cool. Got the infinite loop of screen sharing. So, if you head over to uh nearai/necp, this is uh a first draft of an implementation of an MCP server for interacting with the near blockchain. It has a subset of um all of the possible commands you can use with uh the near CLI, but I'll demonstrate a few of them

16. 7:22

    now. Specifically using a system called claude code. So anthropic also released a tool called claude code. I'll demonstrate for it for you now. It's a CLI tool that effectively allows you to talk with claude an entropics LLM and sort of like a chat GPT but in the terminal as a TUI terminal user interface and you know you can ask it anything. This is like a normal chat GPT window but kind of the cool thing is cloud code has the ability for you to augment it with uh MCP servers. You're effectively giving claude additional

17. 7:57

    tools to to use. So, if you want to add the near MCP server, it's as simple as doing claude MCP add. I'm going to do a little bit of an older version just for this demo. Uh 0.0.11. Um but you should be using latest. So, this is simple in this one command. We use npx which allows you to run an arbitrary package from npm. It's all written in Typescript. And just like that, you've uh added it added the uh

18. 8:27

    near capability to claude in your current session. So if I type in claude now, um it will install this command and we can see we have an MCP server available near MCP and we can ask you questions just about accounts and we can even send transactions and do things. So if I say uh I'll I'll use an example of one of my throwaway accounts. tell me about the account paper prompt near and what cloud code will do is it'll recognize that I'm trying that I want to be able to execute this tool

19. 8:59

    um in near mcp so it get me a little bit of an account summary of this of this account and it uses the the original SDK and look it tells me a little bit of piece of a little bit of information now I don't now currently if I say like list my accounts I have locally let's It should say yeah list all my local accounts and we can see we have none. So uh in our current key store so we can import one. So I have this file locally.

20. 9:31

    Can you import my account paper near from my directory? I think it's called side accounts and claude will probably access will probably launch a few a few different tools. But we can see here it found it found the file right away to import this. So okay, we've imported our account. So now if I ask again list my accounts I have locally. Uh we say yes, we want to proceed. We see we have access to this account. And now that we have this

21. 10:07

    account, we can do we can do things like send transactions and do much more. I'm not going to demo that because that'll take a little bit of time. But I'll give one example like uh one tool that one tool that I added in here is can you sign the data hey there near or say near people um and I'll say sign this piece of data with the uh paper prompt public key. sorry with sorry not

22. 10:36

    public key the paper prompt key let's say and this is an example of one tool and this is it'll call the sign data tool and behind the scenes it'll give us it'll give us this signature and then that we can pass around and there's obviously more tools check out the readme on github it's uh github.com near aainemcp and yep integrate near protocol into as an MCP server in your LLM right now. Yeah. Thank you.

23. 11:10

    Thanks, Alex. Um, quick question. Uh, where is the best place for people to actually uh learn more about AITP and a and you know actually connecting it. Sure. So, the best way to learn about AITP is to head toitp.dev. Can include that in the in links or follow-ups. um that is the best place to learn more about latest developments and also to check out um the near AI the near AI rebuild.

24. 11:39

    Awesome. Cool. Well, thank you for coming on. Uh and also where's the best place for people to learn more about you? Sure. If you want to learn more about me, you can always just email me. My email is Alex or sorry, Alex.ai.com. So yeah, come out reach out to me and if you have questions about MCP server, I'm happy to follow up.

25. 12:00

    Amazing. Well, thanks for hopping on. Uh Carrill, you are up next bringing you on. Welcome. Welcome. So, real quick before you hop into the demo, quick introduction about yourself, uh exactly what uh Datrix is and what you shipped last week. Yeah, thank you.

26. 12:21

    So, my name is Kil. I work for Datrix. I'm satrix. What uh what DX does is we create a no code platform uh to create different uh API agents. Our initial idea was to create a no code platform to create AI agents in analytics which we are successfully doing. But for near uh we are working on the father agent uh it is the a or agent builder the agent that can uh build other other agents just ask him I want to build the agent that will

27. 12:58

    uh like use coin market cap API and uh agent builder will will do this for you. So um if you if you heard about rep lead for example it is something similar uh so it can like uh generate code of the agent can uh like uh implement some additional features if you ask it can save some u API keys uh for you and it can of course it can deploy this agent to near hub So I'm going to uh share my screen just

28. 13:39

    uh quickly. Um so it is on the near hub you can find builder agents builder. In fact here the this agent there is some description of how it works and uh you can uh ask to build something for you. So for example just in the morning I will not I will not demo it uh live because it will take some time but in the morning just tried it. So hey can you create a coin market cap agent for me tell that sure agent name coin market

29. 14:23

    cap here's the description it is the technical plan what features will be implemented I can agree or I can provide some feedback to that plan if I agree it will generate the code here uh of the agent and it will deploy this agent to the near hub under your account Um in case if uh the agent uses some environment variables or uh some API keys, it will provide you the information on how to obtain this key this key and how to provide it. And here you can just click on the deploy the agent.

30. 15:08

    Yeah, here will be the here's the source code of the agent and you can just you can just use it. So, it's what I've shipped uh last week. Uh what we are what we are working on uh this week is uh we want to support uh like more more features to make this agents more functional uh more capable.

31. 15:34

    So uh we uh changed the code base. We are like migrating to the lchain. So it will be like support better dialogue flow. It will communicate with you. It will support CDP hopefully next week. Uh and uh it will like it can it can generate agents that will through the CDP they can communicate with different with different blockchain. So that will open you a lot of abilities so that without any without code you can you can generate agents. So the goal is to

32. 16:12

    generate 1,000 1,00 that agents in couple of weeks and uh we also want to ship the monetization so that you can your agents can earn money. Yeah. A couple questions. So one, what model are you guys using for the actual like codegen piece? Yeah, it is deepseek.

33. 16:33

    Okay. And this is all made available through the near AI hub or you running around? No, no, it's available through the near hub. So, it's pretty it's pretty good. So, we started from quen but it generates some random uh tie symbols.

34. 16:48

    Use pretty good. Cool. And then what kind of agents would you want people to actually build using your agent builder? Yeah. Yeah. So uh the first one that we want to build is coin market cap agent and it will like just use the whole available coin market cap API. Um then probably what we'll be interested is uh uh there are there is a marketplace for different APIs called rapid uh there are like thousands and thousands of different APIs for example

35. 17:25

    to get uh leads from uh linkadine or to get uh information about some like from Google maps etc etc. So uh you just can provide which API you want to use and the father agent will be the agent for you. So for example you can ask uh okay give me the give me the agents it will recommend books for me or films for me and use just API and it will generate the agent for you. So it can be really cool.

36. 18:00

    Cool. Amazing man. Uh is there anything else that you want to share with the community before uh passing off to Shashank? Yeah. Uh so right now the like prototype of this agent available on the near hub uh and you can try it. So in case uh like happy to hear any feedback uh on how it works any bugs or any features requests that can help us to uh create this agent in the way that it

37. 18:31

    will be really useful uh for community and really fun to use and it's the the agent dashbuilder correct? Yeah, agent dashbuilder. Yeah. Awesome. And write any feedback to kkatrix.ai. Awesome. Well, Krill, thanks for hopping on. Uh really looking forward to, you know, continuously following this project. I think that this is kind of how we, you know, hit that like terminal velocity in terms of agent developers because, uh, not everyone's going to want to build their agent from scratch,

38. 19:03

    but they have good ideas. So, thank you for coming on. Uh, passing off to Shashank. Shank, how you doing? Hey. Hey, man. How are you? Doing great. Doing great. Feeling much better. And yeah, apologies to everyone for missing last week. I was absolutely bedridden and terrible. So, um, yeah, do you want to, uh, give a quick overview about who you are, about what Fraction AI is, and what you shipped this week? Um, sure. Uh, I'm

39. 19:30

    Shashank, founder and CEO of Fraction AI. It's a decentralized auto training platform for AI agents. Uh right now there are probably thousands if not millions of agents out there but all of them are pretty static as in they run on the same underlying models. Uh there could be thousand agents but all of them underlying are just deepseek. Um so doesn't really differentiate them. Um they're pretty static. They do not get

40. 19:54

    better with time. And so we are solving that problem. It's a decentralized platform where you can create your agent without any code. you just need to give the instruction of what your agent should do and then uh your agent competes with thousands of other agents on the platform and becomes better over time. So it's it's like a training platform for your agents and during this

41. 20:17

    process uh you you can also earn because it's your data that is being used to train better AI models. Um so yeah this week has been pretty busy for us. Um we launched our version 0.2 of of our uh test net. Uh it it's been eight weeks so far and we have gotten pretty good feedback uh from our users. Uh let me just share my screen.

42. 20:41

    Um share screen. Okay. Yeah. Uh so this is the new platform. Uh earlier it used to work but didn't used to look as good. Now we have polished it out. Um so the your agents become better in something called spaces. Uh each space could be about a certain theme. So it could be about rap battles. Uh it could be about having memorable conversations. It could be about uh playing strategy games uh which are very good for reasoning. It could be about writing emails, job listings, tweets, whatever. Any sort of

43. 21:17

    specialization, we create a space here and then your agent competes uh in those spaces. And uh just this week we hit uh 14.6 million sessions. Um and that that's a lot that's a lot of API calls uh to to the underlying models. Uh we we are probably like top five in the world uh in one of the platforms for the number of API calls. So thanks a lot for that. Um uh as you can see we have like uh two this is the active number of players 266,000 overall. I think we have

44. 21:49

    around 300,000 players. Uh so really good feedback. Uh this the the Burning Colosseium is the the new space that we launched this week. It's it's like a strategy game where um let me show you how it works. So you uh you you can create your agent here um by uh giving just giving it an avatar uh writing a name uh let's call it Baker and just write an instruction how this agent should act in that space. No need for

45. 22:21

    any code, no need for uh setting anything up. Just just give your instructions. Uh give what you think a good strategy would be for competing in that space and then your agent will get into that space, compete against uh thousands of other agents and will become better over time. Um and this is a well needed development because right now we are dependent on like four or five centralized research labs to come up with better models. uh through our platform you will own your own agent

46. 22:51

    that becomes better over time. Um so yeah uh you you can go to fractionai.xyz/dap and uh get started on the platform. We are on test net right now and uh you'll have a company of 300,000 other users. Yeah. What kind of spaces would you want people to run? Um, honestly I I don't want to limit uh limit people's imagination. They can come up with pretty wild ideas and would be open to have those spaces on the platform. So it

47. 23:24

    could be games, it could be education, uh it could be marketing, sales, uh anything people want to do uh and and want to create better agents for that and want their agents to become better by competing against others. So yeah, uh no limits there. Cool. Awesome. Is there any place that people should continue I guess like learning more about Fraction?

48. 23:45

    What about you? Uh yeah, sure. Just go to fractionai.xyz, join our discord. Uh our mods are pretty cool. They'll answer all your questions or uh I try to be as active on Twitter as possible. So just just ask me out there and and I'll be happy to reply. Uh it's pretty easy.

49. 24:01

    Yeah, you don't need to code. Just just go to the platform, create agent and compete with other users. All right, cool. Shank, thank you for joining bringing up Slime. Thank you for coming. Slime, how you doing? Hey, I'm doing great. Thank you for for having me.

50. 24:20

    Welcome. So, what I built uh this week is Wait, actually, real quick, real quick, real quick. A little bit more about uh in uh Interior. I can't pronounce it very well. Uh yes. Uh about yourself and yeah, then dive into it. in here is a a memecoin that was created um last year actually exactly one year ago. So we have an anniversary today and that's why we wanted to launch something today. So now we don't have a happy

51. 24:54

    anniversary. It's super exciting. Yeah, thanks. So we built a telegram bot that has um a lot of stuff on here and recently we integrated the near AI agents so that you can use any agent from near AI in in right in your robot. Let me share my screen. Here it is. So I can do SL start agents and uh there are al also beta agents that are supported but uh we interested in your AI. It shows a bunch of agents but um

52. 25:42

    what also today I have launched is um a dating simulator. It's an agent that u nobody wanted but uh everybody deserved. So you can search by typing Click and uh this is the first extension. An extension is uh what um agent builders can optionally do. It is not supported by app.n.ai but uh is supported in telegram. We launched uh last week and so that not many people know how to use agents. So we this extension to help them understand what

53. 26:18

    they can do. For example, this agent has start a new game button. And I will show in a bit uh how to how developers can in integrate this into their own agents. So this agent is syncing and um it's going to take a while. Maybe I'll show this um what I done 20 minutes ago.

54. 26:48

    So, Professor Stok's nectar trade clicks mandibles approvingly. And this is basically the story. You can see the illustrations that it generates. It's a dating simulator with uh animals with insects because that's more funny. And I can respond with buttons.

55. 27:11

    This is another extension. You can add the buttons uh under messages so that you can uh so that agent can um give people options to reply with for example I can say wow the setup is so detailed by the way this Chinese uh character same as K mentioned we're using qu so it sometimes happens and it's also syncing it shows you send this is the button that I pressed test and it's typing. It shows that um

56. 27:48

    you have to wait a little. Here it is. It's also typing uh the next uh question. And it will uh randomly generate uh this illustrations uh once in a while. And so that's it uh for the agent itself. But I'm also going to share a little on how to access the documentation of um of how to integrate this. You can check on docs.inure.te click the button AI and uh it's not really hard. That's lit literally all. So yeah, this these extensions

57. 28:37

    don't break uh near EI website. You can just uh add them to any chain or or sorry to any um agent and it will continue to function as it was on your AI but just will be a little better. But on Telegram, you're muted. Yeah, I was saying that this this is super cool. Like what do you I guess want people to to to make with this. Do you have any sort of vision of uh how the community get more involved? People can make anything they want, integrate it in any agent. But what I really like

58. 29:20

    is uh improved UX to help people um navigate the agents, help people um help new users um know what the agents can do because right now you you see the input field and you don't know what it can do. You just ask what you can do if you are if you know how to do it. But if not, you can say like a GM and it u it discover what it can do in maybe 5 minutes. But now it will be pretty

59. 29:50

    instantly. You already know by seeing buttons what it can do. Yeah, that's great. Yeah, people people need that help sometimes. So, uh that's really exciting, man. Is there anything else that you want to share with the community? Yeah. Oh, sorry. I forgot uh one more extension. It's the run status.

60. 30:11

    I don't have a demo for it but it's basically um a way to for the agent to say that u the agent is running for long and uh it can um show the progress by steps for example uh when uh it fetches data it can say a progress fetching data and like 50% and uh it will be displayed to the user you can check it it uh on docs.incere.te if you're a developer, if you want to improve your UX for the users and there will be more extensions

61. 30:52

    in the future and I hope this can be integrated in uh near as well at some point maybe in different format but um I just like uh this uh better UX. Yeah, totally. um especially with uh Telegram distribution would be nice too. So yeah, this is great. Um where can people stay up to date on everything else that you're pushing?

62. 31:20

    They can uh let me share in where do I share it? We have group. Yeah, just drop it to me and uh we'll we'll share it after. Yeah. Okay, I'll drop you. Oh, and also these agents can be added to a Telegram group where your group members can interact in it uh inside the group itself. So, yeah, it's great. Cool, man. Well, thanks for coming on. It was awesome to have you here. You're sort of this uh this myth but yet legend the ecosystem.

63. 31:56

    So, uh good to good to see you on screen. Thank you. And uh awesome. Well, and last but not least for today, we have Zahed to talk more about what's been going on in Jutsu land. Zahed, welcome back. Hey everyone. Is it the same format like introduction? Same format. Yeah. How did you know?

64. 32:17

    Hey everyone. Uh I'm Zahed, founder of Jutsu. So, Jutsu is a AI agent platform but we mainly focused on enterprise u and we are working with near AI to kind of a build uh near Typescript uh agent SDK uh to build a better um developer experience uh so that people can u build a easily. So this week uh last week um we kind of a created a proposal for how the uh how basically the framework will look like. So I we we created a RFP that um I'm going to share

65. 33:03

    maybe I uh I start sharing my screen. So the first thing is uh this uh document. So it's a basically a way for us to kind of uh break down the framework and into different smaller pieces. And then uh the most interesting part is at the end uh a the agent will look like a very simple modern way of building a agent where you can specify a model you can specify memory tools and then you can run your agent. So it should be that easy and um also like you should be able

66. 33:46

    to scaffold a TypeScript uh agent. Today it's it's very difficult because you have to kind of clone the whole near AI repository but in the future you don't need to uh do that. Uh this is another uh so this is another open source repo where you'll probably have get more idea about how we are thinking through the structure of our project. So uh we took the kind of the inspiration from the Python SDK but we are building it more towards uh Typescript uh developer

67. 34:19

    because Python and Typescript developer are kind of different and they love different standard different formats. So this is this is another place you can follow us uh follow our work. Uh the last uh another thing we did is we published a uh a video on how to kind of a getting start with Typescript SDK the current SDK. So just remember we are working on the newer SDK and kind of gradually improve it. Uh but if you want

68. 34:49

    to start today or maybe tomorrow for the hackathon uh so this this video is kind of the best way to get started. So it has uh from how do you install near AI and then how do you create a typescript project and how do you deploy it. So you you get the whole idea and I'll share that thing with you. Uh the other thing other than near AI like one of the other thing I have I was working for last probably a month. Uh so uh it's a social

69. 35:21

    media LinkedIn social media manager agent. So the basic idea is I I I didn't I don't like to post every day but if you if you are if you know people in in LinkedIn like some people they're like machine they are kind of a uh publishing every day two or three post but I don't have time so that's why we build post.

70. 35:43

    So post is a kind of a agent where you can come and you can specify a few things and then it will generate a post. You can also generate image with those post. So one of the recent post will be this one that I published. So um and then you can see the quality of the post also like it's it's getting better. So people who founders especially busy people builders who don't want to be a social media influencer and spend like a six hours a day in social media but

71. 36:15

    still want to be consistent in social media. I think post is a aen uh you might get value. So just go to post.ai and kind of a give it a try. Uh so that's kind of a what we did uh this week. Uh few other things we did uh last week was um uh we presented uh near SDK to the broader uh TypeScript community uh in SF and we learned like a the existing SDK have lot of kind of a uh friction. So that's kind of a this week next week we are going to work on kind of improving the frictions. Uh and I'm

72. 37:00

    excited for the hackathon. So if you are in SF definitely uh participate the hackathon and I'll be there. So I'll be happy to kind of a help um anyone in the hackathon. Awesome. I will definitely be there and will probably be asking for your help.

73. 37:22

    So thank you. Uh and uh just real quick, what is the best place for people to reach you? So the the best way you can help now is the following the post uh LinkedIn page. So I will share all these things to uh uh Cameroon and then pro you can probably share or should I post past it here or like a send it to you? Uh you send it to me. I'm I'm actually have a drafting message now.

74. 37:49

    Perfect. And and reach out to me like in in Telegram or send me an email in zahjutsu.ai. people who know me like I'm open book so I'll I'm here to help the community so just reach out. Awesome. Thank you. Thank you. Uh and that's pretty much all we got guys. Uh just to recap uh talked about AITP going to drop this link here. Talked about the agent builder agent agents building agents uh fraction we got in ear and

75. 38:21

    then post.ai. So it's been a hell of a day. Uh we will continue this next week. I'm actually really excited for next week's. There's a bunch of uh pretty big announcements coming out. And yeah, last but not least, if you are in the Bay Area and very interested in building actually useful agents, come through the useful agents hackathon. Uh it is brought to you by Electric Capital and SF Compute and Near AI. Uh this is going to be a hell of an event. We already have about 250 or so hackers. Um there's

76. 38:52

    a lot of other people who wanted to come. I said no because uh they are not actually interested in building agents uh which means you don't come to the hackathon. So uh yeah this is what we're doing this weekend. Uh it's going to be weekend full of shipping and we'll update everyone on how it goes uh early next week. So thank you all for tuning in. Uh to stay updated just follow Near

77. 39:15

    AI on Twitter and Near Week. Shout out to you guys for making this happen every week. Uh we are excited to continuously do these things. So uh we will all stay in touch and see you next week. Bye. Thanks for watching. Subscribe to Nearweek.