NEAR AI Ecosystem - What Did You Ship This Week? #17

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

1. 0:00

   We do a lot of things on the back end and I'll talk about a few of those things on the call. But what we're trying to enable is conversational frontends with DAPs um to make DAPs more intuitive, interactive for new customers and then more exciting for um you know people who are already doing things on the DAP. Uh as a way of background um we do a whole bunch of stuff. We have an

2. 0:24

   agent registry with over 600 agents. We have a um tool registry where you can find lots of tools across different networks um a lot of near tools specifically and a um our latest updates which I'm very excited to share today um are one an analytics dashboard which now every uh agent builder will get which will show um sort of like all the conversation logs that you have. We are GDPR compliant. Um it will show basically how people are going through

3. 1:05

   the funnel. So are they uh like how many message does it does it take to get to a transaction, the success rate of transactions and so on. Uh we have a couple of these already that have really um already proven pretty powerful insights into a uh our agent builders letting them know how to improve the agents. And then um let's see I have another slide on here we are.

4. 1:41

   Some of you guys have already seen this already, but um we're excited to share that as of today uh near intense is uh natively supported on um beta. So you can actually use that in in production. If you go to the beta terminal, you'll be able to see um Serge, our head of product, spent a couple weeks building this. But you can now a um here's a little demo say hey swap you know one USDC from near to base using intents and a um you'll be able to do that through the chat. So this will be

5. 2:19

   available to all of our agent builders as a tool uh and across every chain. So uh if you're interested in adding intents to your platform and you want to do it through a conversational UI, we're happy to support. Um, and we just launched this, so we really appreciate anyone who tests over the weekend.

6. 2:36

   Please give us your feedback if anything is not working. Thanks, Paul. How would they give you feedback? Uh, a few ways. Obviously, we love feedback on Twitter because uh then everyone can chime in with their own perspectives and also a u we have a pretty big Telegram group as well.

7. 2:56

   Yeah, that'd be great if you could share the link. We'll get that added to the recap um that we post normally after every uh episode. But a couple questions are are there certain pairs that you guys are prioritizing in your intents or can you use any asset that is supported by intents?

8. 3:12

   Yeah. So um two things one you can every all the assets are supported although um a uh uh you can see on our networks page what we support natively. So going between uh both networks and then for the networks that we don't quite support yet, you can uh move assets there through mirror and then uh we'll be working in the next few weeks to support uh the remaining networks. I know the two that we we don't have a timeline to support currently are Bitcoin and Doge for

9. 3:49

   native interactions both ways, but we intend to have Salana support next week. So just the Bitcoin and Doge for now are the only two limitations. Super cool stuff, man. Well, uh, this is great. We will follow up with everyone in in a Twitter thread after the fact.

10. 4:07

    Um, Paul, is there anything else that you want to say before we move on to the next speaker? Uh, no. Thanks for the time, everyone. Yeah. Awesome. Sweet. Thanks for joining. And, uh, next up we have, uh, Anton, who's actually right here with me, uh, in the other room. And so, uh, I'm going to quickly hand it off to him. Uh, but Anton, I guess I'll talk to you here.

11. 4:29

    Uh, introduce yourself. Introduce what Near AI is, what you're doing with Near AI, and what you ship this week. And I'll let you take the booth. Uh, hello. Uh, my name is uh, Anton Lemonos. I'm a member of the Neari team. And uh on near eye we have created a near eye hub and we are also creating a new projects and uh in in parallel uh we have uh created a analytics project that

12. 5:07

    I'm going to present myself present Yeah. So this is um near uh analytics uh dashboard to monitor AI agent analytics. It's an open-source uh repo and uh the intention here is uh to have a analytics for artificial intelligence agents that is more or less agnostic of the platform.

13. 5:44

    Uh it is however integrated with the data from near iHub and it will also be integrated with the new agent cloud that uh we are building. Um the flow of analytics uh is to first to collect uh logs and uh there is a script to collect logs from near iHub that any agent developer uh can run.

14. 6:13

    Uh then there are some backend jobs to process the matrix. Uh and uh finally there is a UI dashboard to run on this matrix which I will present now. So this is a dashboard that can be run on uh all the metrics that is gathered from agents. It's uh fully configurable.

15. 6:39

    It can have multiple views which is given in configuration. In the current configuration there are six views. Uh so the first view is a invocations uh number of invocations. Then we have a latency graphs. There are some uh the graphs that are shown on uh the start of the dashboard.

16. 7:07

    So uh if uh nearby hub data is passed to this dashboard it's quite uh it has quite clever initialization algorithms for example if it detects that agent developer used uh multiple models then uh like several models then there will be like several lines here and you can also configure all these graphs to for example slice by the model.

17. 7:36

    And similarly, if a developer has several agents, there will be several lines here each for agent. And you can create create any graph on any metric. Uh then there is a table view. The table view is initially one row that you can then uh slice for example by your engine name.

18. 8:02

    And here you have a matrix for each agent. For example, number of times it was called. So this second column is number of times it is called. All the size all those cells are clickable. You click them and you have more information. You can click columns to get more information.

19. 8:22

    You can click cells to get more information. You can also download this as CSV. Uh then uh you can use uh for example filtering like last week. So you can see the requests that are made each week. Uh another table is to compare models.

20. 8:50

    So this is a comparison of uh models. Uh uh right now it only has a a leave bench matrix. uh but we are planning to add uh yeah there will be way more data. There will be cost data. There will be latency data. There will be scores from other benchmarks.

21. 9:08

    Again this this also clickable and you can see more information with all the descriptions. Then other views are logs all logs or error logs. And we can actually scroll down to system for low for example. And here yeah the the most usual error is that model is not present.

22. 9:39

    Uh we have also that the privacy notice guidelines here. Uh it depends on where the traffic comes from. For for example if it comes from agent developer themselves then agent developer has access to all the logs. If it is a benchmark or evaluation traffic then uh it can be present on graphs or table but no individual logs and if it is actually from users then um it is shown if user agreed to that and we

23. 10:10

    determined that it doesn't have any sensitive API information and here I'm just admin so that's why I have access to all the logs but for edge developer we use this direct privacy guidelines. Um yeah back to the repo. So this is a dashboard UI component and it is a web component that can be uh integrated into any website. There is information here on how to integrate this web component

24. 10:47

    uh and it can have um a lot of different config configuration. Do do you want one view or do you want multiple views uh etc etc. Um yeah it's uh published as npm and can be installed in many uh websites. Hey Anton, is this live today? Yes it is live.

25. 11:16

    Um and so any agent developer has access to these analytics on the near app. AI uh yeah uh so all the agent developer has access to the traffic that uh from agent developers. Uh but there if there is interest shoot me a message. We definitely want to expand uh the access uh in other directions so that you can see logs from users as well.

26. 11:49

    Yeah, we definitely want to expand the access. And what's the best way for people to reach you? Uh just uh yeah, just use Telegram or Slack or GitHub. Yeah. If you have a hard time getting a hand of hold of Anton, you can always DM me on Twitter and I can make it happen.

27. 12:18

    So yeah, what's next? What's next for the analytics? Uh for analytics we are actively integrating it uh in our new agent cloud platform and uh yeah we are improving uh on how it can be integrated into websites and also yes expanding uh data data access for agent developers.

28. 12:45

    Nice. I have a lot more questions about the agent cloud generally. I think we should get Pierre on at some point to give the full lowdown, but um is there anything else that you want to add about the analytics side before I move on to the next guest?

29. 12:59

    Yeah. Uh if you're either interested in analytics for your own agent or for analytics for your agent hosting website, take a look at my repo and shoot me a message. I I will be more than happy to collaborate. And for those who don't know, Anton is the champion in the uh near AA alpha telegram group that answers everyone's questions. So he is incredibly responsive and uh you know we're very lucky to have him. So Anton, thank you for joining. Uh hope to have you on in the future. And everyone else watching,

30. 13:31

    DM me if you can't get a hold of Anton and I'll be happy to put you in touch. Thank you, Anton. All right, next up we are moving to Arto and Matt. uh from Asimov Protocol. Gonna bring you guys on stage. We have Arto here, Matt here. Welcome back, guys.

31. 13:51

    Familiar faces. Hello. Yeah. Yeah, indeed. Always love to be back. Yeah. Well, just a quick reminder even though you've been here a bunch. Um Yeah. Who are you guys? What is Asimov Protocol? And what did you guys ship this week?

32. 14:07

    Sure. So, okay. So, I'll start off with that. So, what is Aszimov Protocol? So myself and Artto are two of the co-founders of Aszmoff. Um we are building the infrastructure for a decentralized verified knowledge base using knowledge graphs designed specifically for AI agents to reason over. And I'll I'll show you why that's uh of interest in in in a second here.

33. 14:31

    Think of it as the semantic layer for AI where structured data, provenence and like decentralized compute all come together in one place. And one of the uh specific aspects of this is let me share my screen here is what we want to do is is make the sort of foundational models effectively interchangeable and shift the value to the data. Right? This is something that we've been working on.

34. 14:59

    And what what is interesting to see is I mean this was a tweet from Andre Kapathy um earlier on today um and based on um a a an announcement uh released about Gemma 3N talking about how these models you can actually run locally now, right?

35. 15:16

    Um, and so the idea being to shift kind of the split the intelligence up so you have a cognitive brain that is able to reason but really move all the encyclopedic knowledge outside of the model and into a separate uh library of information and that's what Azimoth is right so Azimoth is a series of knowledge graphs that the AIs can consult so the idea being is that the models themselves are interchangeable so

36. 15:42

    a better model comes along you can use that you want to use a hosted model you want to use a local model, keep everything local, you can do that. So, we're really kind of sort of skating to where the where the puck is is going with this. Um, and to that end, um, Artto gave a little bit of a tease of this last week, but we have launched, um, our new website, azimoff.sh and, uh, so this gives you a bit more

37. 16:04

    information about what we're doing here and, um, some examples of using, for example, the langu loader. And I'll give a an another example of that in a in a second here. So we have that which is uh finally launched which is good. Uh we have the Azimoff directory which I think I showed last time for people uh which is a directory of all like the modules and data sources you can get to um and use with Azimoff. And we also have the Azimoff specs website which gives you

38. 16:34

    all the specifications you need to build on the Azimoff protocol. And one of the things I'm going to be doing is working on a basically an LLM friendly version of this so that you can directly point whatever you know um clawed code or whatever tool you're using uh you know MCP tooling in curs or whatever uh to be able to ingest this in and and help you build with that. So just to quickly

39. 16:58

    state uh here that you can actually as as sort of shown on there we can actually use something like lang chain and use the azimoff loader in lang chain and with not a huge amount of code um you can build uh a knowledger graph application here to output data. So in this case um we're using it's using azimoff to extract directly from duck.go. So it recognizes the URL as duck.goes goes and gets it does a query

40. 17:24

    for Isaac Azimoff brings back the data but what is interesting is it has actually augmented it with some intelligence here so it has actually worked out and said okay well Robert Heinland was a person Arthur Cclark was a person Boston University an organization so that's adding more semantic information in which is what Azimoff's all about I'll pass over to Artto now uh because he's going to talk

41. 17:46

    about some more release stuff righty um let's See, so we are in the middle of the platform roll out and um this is basically the last uh sneak preview because uh everything uh goes out on Monday. That's usually a good uh day of the week for big uh launches. Um and uh I'm going to just show a sne and then um starting starting on Monday everything will be nicely uh documented long form

42. 18:18

    um tech blog post and and and so on. Um let's let's look here at a couple of things. Um h yeah so first of all we had indeed uh the blog went also live uh this uh this week. So this is uh literally asimov.blog blog. Uh there's an initial um some initial content there and there will be soon um technical content uh there as well. So subscribe to the blog over at asimov.blog.

43. 18:50

    Um on uh on GitHub um we have uh now made available the Asimov uh command line interface. Uh so this is installable via multiple methods including uh homebrew on uh Mac OS and uh scoop on uh Windows uh and of course the usual cargo install and the like. Um and there is also uh there are various um extensions to the base uh command line uh interface. There's another one that's available uh the module um command line interface common tools for that. Um we are building an entirely uh

44. 19:27

    extensible platform. So there's a rather large number of uh different components which will all be explained in the uh introductory blog post on Monday. Um over on uh on Twitter uh please follow us at asimov_platform. This is specifically uh for platform updates uh such as indeed um some releases uh today. Um so these are the last uh -ashdev releases uh prior to the actual tagging of uh 25.0

45. 19:59

    which is effectively our 1.0 for the platform. Um a year a year in the making um very very satisfying to uh finally uh get it out. Um let me see I could um uh just demonstrate uh this also a little bit live from the terminal. Uh so let's take the terminal here.

46. 20:22

    Um so this is all uh available already if you know the magic incantations and this would be things like uh you know tapping uh tapping that particular home root tap um and then uh going ahead and installing the Asimo CLI and uh and sure enough uh we we have that available.

47. 20:46

    uh but it's not u super useful yet by itself. We uh generally distribute it um using a bunch of these uh subcomands. So similar system as git where uh git is distributed as a 100 binaries where we don't quite have 100 but we do have a few uh and for example um let me just show um for variety's sake we installing the module CLI from uh cargo uh instead of homebrew. So, um, one of the awesome things about the Rust ecosystem, of

48. 21:16

    course, dependency hell is a thing of the past, uh, you just cargo install and, uh, things generally work out pretty nicely. Um, at least if upstream has done, um, you know, reasonable job. Uh, so this is, um, this is how you would get any asimov component onto onto your machine. And I think we've already in previous weeks um, demonstrated a lot of the the practical functionality uh, here. So I I won't rehash that. Um we'll

49. 21:45

    we'll have some cool uh demos um uh next week and uh that will be more interesting than just uh you know fetching uh fetching data. Uh so I I would uh I would leave it at that for now. You you showed Matt the uh spec site as well. Matt, you're muted.

50. 22:06

    Yeah, I showed the spec site. Yep. Okay, cool. Um so uh Cameron any any questions uh here? No not really. Um it's all it's all pretty straightforward. Um I guess one one question is like I see Asimov doing there's a lot lot of stuff going on and you know knowledge is a very large space. So if you were to sort of talk about like the immediate sort of almost twoliner or oneliner of what is doing that is relevant for AI builders today.

51. 22:41

    What would that be? Uh sure. Um so the the way we are phrasing it is that we we are building u what will be the leading platform for trustworthy neurosymbolic AI. And just to unpack that trustworthy um is rather important and unpacking that is going to take a blog post. Neurosy symbolic means the combination of knowledge graphs with LLMs. Um there isn't really uh any kind of existing standard or player in in this space. We aim to be that. And as you can imagine, Cameron, there's no

52. 23:17

    there's no conceivable name on this planet and and the history of mankind that has um let's say this similar kind of um claim to trustworthiness in AI as all he's the first AI aligner in history. Um which is three laws of robotics. So uh that's what this has always been been about for me. uh trustworthy AI uh AI that knows um where the data came from, why it should trust it or not uh and is able to make better better decisions and as a result we able to make better decisions as uh humans.

53. 23:57

    Got it. So it's primarily on just the data like essentially data structuring in order to make sure these LLM have a better easier time process. Look, look, there's a Cameron, there's a there's a multi-year uh timeline here uh which is again all about trustworthy AI and that has implications far beyond the data itself.

54. 24:19

    I mean, for example, why do we write in rust trustworthy AI, right? We don't want to be seging and and corrupting memory like llama CPP for example. Um so we we aim to build reliable systems. Um but we are beginning off with uh the uh knowledge part and uh hence I shall limit my answers mostly mostly to this at the moment.

55. 24:43

    Check. Matt, is there anything you want to add? No, I think that's it. Uh check out the website. Um we'll be doing some more demos like Artto said, a bunch of stuff being released on Monday and um yeah, I'll be putting together some more demos for people to actually uh kind of have a have a play with it. and uh and and Cameron should also mention aside here that we are beginning our institutional first institutional round.

56. 25:07

    Oh, cool. Well, congratulations um on all the progress. Uh we'll DM you about the the round and thank you for joining again. Great. Thanks. Thanks a lot. See you guys. All right. Up next we have Sasha from Learn Near Club. Sasha, how's it going? Welcome to the show.

57. 25:27

    Hello. Hello. It's going very well. My pleasure to be here, I think. And this is my very first time being at at this scope and I'm excited to share what we did ship even like not last not not this week but even like last minute, right?

58. 25:42

    Usually we don't we do not put in production stuff on Friday. But yeah, but this is an exception. Yeah, it's it's well amazing. Yeah, quite light though. So nothing should be broken. Yeah. Cool. even if it is these are demos. Hopefully demo gods are with us today. Um just want to kind of quickly give everyone the big uh kind of overview of what learn your club is uh because I'm quite familiar. Uh but this is your first time on the show and the audience might not know. So if you don't mind just giving a quick introduction about what learn your club

59. 26:15

    is, who you are and then diving into what you ship this week. Absolutely. Uh learn your club is community learning everyone uh can learn about near protocol itself and uh which is important to about near ecosystem projects and uh me personally I've been uh working in online learning for more than like 20 years and I've been working with new protocol for a while and what we do now we do uh blockchain for

60. 26:49

    loading and uh like near protocol for loading And now apparently we do it with AI. So let me share my screen. Okay. Share. Okay. Okay. I hope you can see it well. So, uh, a a short preamble about, uh, learning and Hey, Sasha, do you mind just like zooming in a little bit on the screen? One second, please. Sure.

61. 27:34

    Okay. Is it good now? Yeah, much better. Yeah. Okay. Cool. Yeah. Cool. So, so the the short preamble uh uh why it is important to to get verifiable and responsible right for learning uh AI has already entered the classrooms and stuff right and if you put it in plain English like almost like like if you put it to extreme uh people stop learning right at school like you can see like ch doing essay at one side at one end and ch

62. 28:14

    grading this essay as at another end the the teacher end, right? Like and so like we we have to fix it, right? We have to to figure out how we can uh uh how we can train our our own intelligence with the I help uh rather than just you know just empower the artificial intellig intelligence by itself. So this this task we are tackling at Lonia club with the help of near team and uh other project who participates actively in near ecosystem and one of them is our

63. 28:52

    private uh knowledge based provider extra which I think uh was on this call uh recently. Here you go. So the the most uh uh the most recent update that we get is like this uh commenting visi because we believe that commenting is very important uh activity for the learners especially for non native English speakers so they can uh review once again what what did they learn and leave their

64. 29:25

    opinion about that. So let's go here and see the comments. Click the comment seed section like issues. Okay, I am interested in the issues with current. This is the section that I'm actually commenting and ask uh learn agent to generate comment for me. I I have to pay it with uh uh my uh and learn uh my learn points and here you go. So this is a comment generated uh by uh learn agent

65. 30:03

    and okay let's see what we have I think this mixer so like this thing I want to remove cuz like I want to to be a little bit personal my make unease especially when you see this seems transport us. Yeah, otherwise the rest looks good for me. I remove this stuff as well. Now I'm posting comment and let's see if magic happens.

66. 30:44

    All right, here you go. Okay, now now we have I think we Okay. Okay. Yeah. So the system detected that that comment was was actually first I generated right with a little bit of participation of the user of the user and the system graes this way as you can see here right that the uh the award for the comment was shared by the user and by learning agent.

67. 31:22

    So I think this way we we uh making adjusting that common system in the way which is looks right now. First of all we help users to do not hesitate to to comment themselves. We do inspire them to use but at the same time we we force them to review actually the comment and tweak it if needed. If uh we also reverse them according right. So if they do not bother, they get nothing for this comment, right? And everyone like and they even sign this, right? Like I just

68. 31:57

    posted this comment and I wrote it for me completely. So uh and if user spend more time, you know, and review the comment and rewrite the comment to make it personal. Uh I think this helps user to better comprehend the material and to to build that knowledge into their head.

69. 32:19

    So hopefully this one of the small bricks uh in the in the new learning method uh which could be adopted uh at schools and somewhere like it's uh so it's uh it's work in progress still right but I think we we are on on a on direction on on redirection okay and okay and another thing that we shipped today and proved it a little bit today.

70. 32:49

    Let's let's see if it do the job. Uh sometimes users state, right? They they do not know how to start, right? What kind of question should I ask? Like me personally, you guys like they're probably well aware about near system and you have some particular questions that you may ask, right? But newcomers, I have no idea. So like we created for them. It's nice feature, I think. I like to think to think so like surprise me and let's see how it how it goes. You

71. 33:21

    can see that our learning agent is learning from private knowledge base. Now it's consuming the information stuff. It's a little bit uh not the performance and as we would like to to have it but uh thanks Anton and team now we can better explore the performance of the agent with uh with the dashboard and see and see where the bottom are. Thank you guys.

72. 33:51

    Okay. So this is uh this is what uh the agency came with right. People say near faster is another bchain due to it impressive technical capabilities particularly it fast block production and finality times which is cool right like when when you read such a surprise you probably are inspired to get into more knowledge about new protocol itself and new ecosystem projects and uh yeah

73. 34:20

    let's try to ask about our friends from extra extra What does the learner eye thinks about them? Tell me about X-rays. And while the agent is working I I I I want to uh to draw your attention to the fact to the problem that uh extra is working on and protocol is working on the why we we we were working and we've been working on learn AI agent we we figure out that the models itself doesn't much matter they still like

74. 35:00

    already intelligent enough but what matters a lot is is the knowledge base like garbage in garbage out so this This is where the value actually exists and this is where actually the like normal ordinary users ordinary content publishers will be able to monetize their content. So kudos to you guys and the best of the luck. Okay, so yeah, apparently uh the system knows quite a bit about X-rays because they are

75. 35:30

    friends and they do not hesitate to publish some learning materials and learn. So but just in case right if if you if you actually do not like the the reply of the agent and uh you can work and help us to uh to find the knowledge gaps and improve the system for you and for and for everyone else. Yeah. So that's basically that. Super cool stuff. Sasha, what's next?

76. 36:00

    Like are you trying to get this adopted by external education organizations or what's the you know What do you what's on the road map? Yeah. Yeah. This is this is this is the this is on the on the road map. Next step, we still have to polish it a little bit, especially performance. So like, you know, like when you go out of of your bubble, right, and and talk to people, they expect like web web two

77. 36:23

    experience, right? Everything like this, right? And they they want to get that for free. So we we we'll get it little little bit polished and uh and uh and uh more robust and yeah for the it's it's easy to sell this right like when I talk to parents listen like do you know what you what you actually what your kids actually learn like discharge GPT you have no idea like like zero right like and and with with this approach that we are trying to

78. 36:55

    demonstrate like we we going to have like poor roles right a learner himself or herself a parent a teacher and the eye tutor and everyone is aware what's happening right and and the kid is aware that if she she she knows this right do not cheat behave well and get bonuses man I don't know if you saw recently but the uh there was a viral video of the student that graduated UCLA that held up his computer thanking chat GPT for his

79. 37:25

    degree um and so I feel like we are uh going to get much more stupid as a society if people aren't actually learning things. And so really love what you're building. Um my girlfriend's a fifth grade teacher and it's really disappointing even at that fifth grade level. People aren't reading. People are just going straight to chat GPT for all their answers and their actual math skills that a fifth grade fifth grader at like a second grade level. Um, this

80. 37:56

    seems to be pretty consistent and I'm worried. So, we do need things like what you're building and uh, yeah, I would love to help accelerate in any way we can. Absolutely. Absolutely. My my pleasure and thank you very much to for having me here. I hope what I presented does make sense for you guys for those.

81. 38:16

    Is there any uh call to actions that you want the community to to know? just come to learn and learn, you know, like and the learn with the learn by yourself and just, you know, invite your friends like like show to to your friends who have kids and stuff, right?

82. 38:32

    That graduate guy was already like 20 something years old. Imagine if you have kids like five years old. Yeah. No, it's amazing. It's it's so urgent. So like we have to act right now. Yeah. Yeah. We're running out of time, but uh in learning, but also in this session. So Sasha, thank you so much for joining.

83. 38:53

    Hope to have you back soon. Last but not least, we have James from GovBot. James, how's it going? Hey, sorry about the glare. All good. All good. Doing well. How are you? Doing well, man. Doing well. Just uh working, doing the thing. Excited to learn more about GovBot. Um, but just to kind of give everyone a little bit of context about who you are, what you're building, and uh what you actually shipped this last week.

84. 39:23

    Sure. Well, my name is James and I'm a delegate supporting the House of Stake governance system for Near and Govbot is an AI powered assistant that helps governance participants understand proposals and make better decisions. Um, we've got a Telegram bot which I shipped the week before last. But today I'm going to demo this agent which lives onchain uh in a web 4 contract which is a pretty cool thing. Um, it's been

85. 39:59

    around the near ecosystem for a while and I I think it's a really interesting use case for blockchain. Um, it's essentially onchain websites. Um, but there is a new dev tool built by Peter Salammonson uh that enables a JavaScript engine uh to be embedded in a web 4 contract. So you can post JavaScript code to the contract and it it runs in the browser and um the web 4 get method is what

86. 40:34

    actually serves the content. So you have onchain control mechanisms for what you see on the page. Um but I'll go ahead and share my screen and just dive in. Um real quick, James, can you also give a little bit of background about what the house of stake is and what the use case you're building is?

87. 40:54

    Um yeah, so the house of stake is made for governing uh the near ecosystem and uh potentially the the community fund which is intended for you know developing uh various initiatives across near and um it's also meant to help with other kinds of governance but um you could imagine uh for the sake of you context. There might be some AI agents that help us with budgeting and this governance system helps to make sure the

88. 41:33

    community stays in charge. So I think that's my my best attempt at explaining. It's really exciting project that will you know change how near uh you know actually operates and it's it's really exciting to see how many people are already voting um in a a key decision about the issuance reduction. Um so just another shout out there. Um, but yeah, I hope like hope that uh helps, but it's it's a big deal. Um, so we're we're going to be rolling it out. It's it's

89. 42:09

    supposed to be launched in the next month or so. Um, and and we'll see where it goes. It's very exciting. I um also am a delegate in this this crazy experiment um and truly think that agents are the only way that we're going to be able to scale coordination. So yeah, looking looking forward to this. Yeah. Can you see my screen?

90. 42:35

    Yep. Sweet. So it's very simple what I'm going to share today, but it's a proof of concept. This website actually is is running uh you know in a contract. So, um, it's being served through the the web 4 contract and it's a web 4js contract, which means it has the the JavaScript runtime embedded. Um, but I'm just going to go ahead and connect and authenticate with nearai um, which is under the hood. It's really interesting to me because

91. 43:11

    you can get free inference uh through the near AI endpoints and and the APIs are are super cool. I've got another example of an agent that is published on the hub as well running on web 4. But this is just calling the chat completions endpoint. But you see it says send a message to authenticate. So I'm going to do that. And it's uh giving me a popup to uh log into Near AI. Got a sign a message there. But just went through authenticated and it says hello.

92. 43:43

    How can I assist you today? And what I'm going to do is say uh what tools do you have and we'll see what it's got available in in the tool belt. Um yeah, there's four tools I've built so far. Um the the one I'm going to demo today is the get proposals tool. Uh so you can just say get proposals and it'll ask for some clarifying details.

93. 44:12

    Um and you know this can be you know improved a lot. I'm I'm thinking about what tools to add next and how to how to make these better. Um but it's made for discussion purposes. you can come discover proposals and uh you know even um you know talk about them with the agent. So it'll it'll say oh I've I've got uh you know these proposals and you can say oh what's that one about and

94. 44:42

    it'll it'll tell you more and um once it goes through here um then we'll we'll see what what's on the the contract. Um, one of the cool things about the House of State contracts, um, is that it's totally modular. So, there's a, um, main contract that is, you know, the the singular uh, VE near contract. We won't go into that too much, but that's kind of where the the stake weighted voting power is u, you know, tracked. It's kind of like it keeps keeps a a ledger of who's got how

95. 45:20

    much voting power based on what you lock up into your own lockup contract which is another modular component of the system. Um and then uh the final of the the three is the voting contract which you know is is really flexible and and you could actually have many different voting contracts that work with the main veir contract. So this is one that I deployed for this demo. Um and it's got one proposal in it. So you can see the proposal details and we can even ask a

96. 45:51

    question say like you know how should I vote and we'll see what it says. Um, I think, uh, you know, this is really basic, of course, but if you have any ideas for for tools that could be useful as a delegate or as a voter in general, um, even as a participant that wants to submit proposals, um, you know, please reach out. Um but you know it's it's going to give a a measured response

97. 46:23

    here to uh let you know like how you could approach voting. Um one one idea might be to kind of facilitate voting as well but um that's just clicking buttons. So, I like the more analysis forward approach where we're actually, you know, informing participants and and helping them, you know, better understand uh what's being proposed. And yeah, I think uh that pretty much does it. I'm uh happy to share more and and hopefully come back in in upcoming weeks and and uh give an update on the progress.

98. 46:58

    Super cool stuff, man. And um I'm looking forward to testing this, especially as a delegate. Um, is there anything that you want the community to do right now? Yeah, so I mean we've got uh a bunch of things on on the you know the fire. Um I I will highlight we've got this other uh demo front end. You can go test the the contracts. It's at near.vote.

99. 47:24

    Um we've also got uh this uh new GitHub organization where I'm uh pushing code. Uh it's the near governance um GitHub. You can you can find that. This repo I just shipped last night maybe for an upcoming demo. It's a factory contract that helps users deploy voting contracts. Um and we've got a a Telegram bot. So uh you can test that as well if you want to get into this group or you

100. 47:52

     can just send it a DM. It's neargovernance_bot. Um, and yeah, I'm I'm a JWAP JWOP on Telegram and X. So, feel free to reach out. I'm hoping to collaborate with everybody who's interested in governance agents. Awesome, James. Well, thank you so much for joining. Um, we'll be sure to promote all these things that you dropped and hopefully see you next week or today. Yep. I'll see you, man.

101. 48:27

     All right, that concludes this episode of What Did You Ship This Week? Uh, just as a reminder, we do this every single week. Uh, normally we do it on Thursdays at 11:00 a.m. Pacific. Uh, today we did on Friday because we weren't able to do it yesterday, but uh, hope to see you next week. Again, if you want to participate, you can DM me on Twitter.

102. 48:47

     Uh, we do ensure that all the projects here are shipping on Near. So, that is a prerequisite. I've had people ask to join the show when they're not building on near and well doesn't fly and so uh just DM me if you're interested and to see you next week. Thanks for joining.

103. 49:03

     Thanks for watching. Subscribe to Near Week.