NEAR AI Ecosystem - What Did You Ship This Week? #14

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

1. 0:16

   Near AI, what did you ship this week? Good morning, good afternoon, good evening. Uh, my name is Cameron Dennis. I help lead integrations and partnerships at Near AI and we are here for the 14th episode of what did you ship this week. This is think a pretty much decentralized or open uh standup engineering standup where people kind of show off what they're building. Uh these are a bunch of different projects within the near AI ecosystem. Uh we will maybe eventually start doing this for other

2. 0:48

   things that aren't necessarily AI related in the near ecosystem, but for now this is about 14 episodes in. If you're interested in seeing what's being built, highly recommend going to the YouTube and checking those out. Um, but just for a quick summary, uh, for those who are speaking today, please be sure to only share a window of your screen, not the whole screen, uh, because there

3. 1:11

   have been issues in the past, but here we are. And, um, we actually without further ado, want to introduce the first speaker today. We have Andrew from Doppelgangers. Andrew, welcome back. It's good to have you here. It's been a while.

4. 1:26

   Can you hear us? Hey Cameron. Okay. Hey Cameron, can you hear me? Yes. Loud and clear. Amazing. Hello everyone. Uh hello Cameron again. Uh it's great to be back. Um would love to share the latest updates in doppelgangers. So for everyone who doesn't remember doppelganger is uh creating your perfect personal assistant which uh I believe is you so hence doppelganger. Um the idea is uh we have a lot of inbound uh uh traffic coming in

5. 2:04

   uh emails uh your messages uh social media all of that needs your attention and there's only one of you. So, uh, we decided that, uh, it's going to be much easier. There's going to be a couple a couple of you. Um, one of the newest things and what I'm excited to share with you is, uh, the newest product that we have built.

6. 2:29

   Um, just give me a second to share my screen. Is it presenting? Yes, there we go. Yeah. So this is the newest platform that uh uh we are we have launched already. Um this uh is the biggest digital twin database in the world. Uh in essence uh you can create a digital twin of uh any public account. Um we are starting with uh Twitter or X and uh yeah um if anyone has uh any ideas uh which digital twin they would like me to

7. 3:18

   create um I'm happy to hear them out in the chat. Um why don't you just come up with one yourself? Yeah, you can clone me or clone yourself or clone I don't know. Man, well, I I'd be happy to clone you. Um, so can you send me your Twitter account? It's Cameron Dennis. I remember. Yeah, it's Cameron Dennis, I believe, again. I'll I'll message it to you, Drew.

8. 3:58

   Great. Yes, there we go. So, what do you exactly are you doing here? You have a Twitter API that you have access to that you're able to throw a username into this search bar and then you pull all my data. What next? Yes. Uh we basically use all the available public data there is. Uh so all uh your uh posts, replies, comments are going into a rack. Um and yeah, we basically have uh your digital twin. Uh generally this is a dev version. So it's free. uh

9. 4:41

   but uh the version that is going to go live uh probably tomorrow is uh going to be I mean like it's uh you can still create a digital twin uh but uh the idea because the API credits are not free uh we will kind of u allow anyone to use as many posts as they want to create a digital twin uh currently it uses just the latest Oh, yeah. Hello, Cameron. Uh, Andrew, I don't know if uh you're typing anything, but I can't see it on the screen.

10. 5:22

    Got to see what did Yeah, I just see what would their AI twins say and then it's kind of it seems frozen. Um, let's let me go check the Streyard. So, what do you expect this digital twin to do like now that you have this thing? What do I Why do I want this? Why why is it frozen? It's so uh it's so sad. Um yeah, so uh in essence uh it's an ability to uh talk uh with a an AI twin of any person out there and ability to create it. So, uh you can have a conversation

11. 6:01

    with Elon Musk, Jeff Bezos, uh uh I don't know, uh your school sweetheart or crush and uh yeah, I personally believe this this is quite exciting. Um obviously it's uh it's a tool. Uh so like all tools, it's limited only by your imagination. Um we also have uh the feature called merge. Uh so you can merge different people and different personalities together. Again uh it's their whole Twitter account and their whole public uh database available. Uh

12. 6:35

    we will add uh um LinkedIn and uh Instagram I mean like the whole meta suit uh uh next. So you'll be able to basically create a uh knowledge a shared knowledge base of for instance all biggest tech entrepreneurs or biggest crypto entrepreneurs like for example uh you can create mash vitalik Ilia and Anatoli from Salana all together and tell them hey how to build a product or how to how to create a foundation

13. 7:09

    um and these are just like a couple of examples I'm for while you're sitting here and I'm explaining this, you can think about more of these. Um, and uh the the fun begins uh when we will allow this to become a launchpad for agents.

14. 7:27

    So, you'll be able to infuse that those personalities that I've just mentioned into an agent and let it out loose in the wild like potentially launch it launch a token for it. Um again uh this uh is a tool um and uh depending on how you use it uh it can have different applications. Uh I am personally really excited. Yeah. Yeah. Go ahead. Yeah.

15. 7:54

    Maybe you're about to get to but like as a business like what are the use cases that you're trying to target first? Because if I had a digital twin I'd wanted to like answer my telegram messages that I define as like tier three and four. Um, I would still want to answer my tier one messages, but uh, yeah, like how does this actually like besides like a, you know, personalized deep research tools, the other sort of

16. 8:16

    use case I see that you're describing, like what are these sort of like verticals that you're trying to go to market with? Um, verticals that we are going to go to market with are initially going to be on the font side. So, it's like AI influencers, AI girlfriends and whatnot.

17. 8:33

    Uh uh but to answer your question in particular, I'm super excited uh uh for a sales agent because you can have uh Wolf of Wall Street as your sales guy uh or maybe Peter Teal or maybe like a mash of different personalities that uh basically it's everything that you believe you're missing within like your uh personal profile. potentially I'm more I'm a more introverted person. So I would uh I I know a couple of friends

18. 9:06

    who I believe would be amazing sales. So I just use their personality and turn them into an agent and see if it is or not. Uh again uh it's this is an agentic tool. Uh so if uh you going to see a lot of success in converting leads within the certain industry, you can easily sell your agent to other people. Uh once once it's ready, once uh it's tweaked and once it's working, you can go ahead and monetize it. Uh that's the beauty of

19. 9:34

    crypto. Uh you can launch a token for it and anyone who will be using it is you'll be getting a cut. Uh it can even be a D uh where different uh people come together uh to build an agent to launch a token for it and figure out some use case that we haven't thought about. Um a lot of people are telling me that the first thing they're going to use this for is uh Tinder, Bumble and other data.

20. 10:05

    Cool. I mean, I guess you're still going to need that plugin into Twitter or Bumble or something. And so, like, how do you actually plan to make that connection because right now I'm not certain Twitter is already there. Um, Bumble, uh, yes, uh, true. Uh, you'll need to get that API, but, uh, there is a web version of it. Uh so you can always go around with uh some sort of a uh um Google Chrome plugin uh uh or even

21. 10:35

    an agent uh that takes control of your like one of your tops or one of your screens. So remote desktop connection like like most people are doing it right now like most people are doing it right now when they don't have any API there are workarounds. Got it. And so what did you actually ship this week?

22. 10:53

    Um, this uh has been live for a couple of hours. Um, and uh, yeah, that's so this being this being throwing in your te your Twitter account and then having it scrape the Twitter profile and then having that being used to train your personal agent. Okay.

23. 11:14

    Yes. Uh, unfortunately the screen the screen has frozen uh because uh, mine has chat open. Okay. Maybe I if I kick it from the studio. Uh kick in. Okay. Well, next time uh send me send me a recording. Uh maybe we can display it later. Yeah, it seems to be totally frozen. Um but thank you for joining. Is there anything else that you want to say before we move on to the next guest?

24. 11:40

    Um yeah, it's doppelgangers.ai. Go check it out. Uh we it it's free for now. Uh so you can create uh any digital twin for free, any doppelganger, your doppelganger. Uh all of that is already available and go check it out. Happy to hear your thoughts. Awesome. It's also live on mobile, correct?

25. 12:04

    Correct. Awesome. Cool. Yeah, go check it out. Thank you, Andrew. All right. Next, we have Manuel from Silverstream. Manuel, how you doing? Wait. Hey, you go. Hello. Hello. Um, I think you guys for the from doppelganger should use our APIs to integrate with you know those websites cuz they're like yes this is us.

26. 12:28

    Yeah. Uh I'm not going to say the story that that you know we he first told me how the company started is very close to that use case. Yeah. But uh that is your story to tell not mine. And uh but just for everyone who doesn't know you, doesn't know Silverstream, can you just give them a quick background on what Silverstream is, who you are, and what you ship this week? Yeah. So, we build web agents. So,

27. 12:54

    as um as the guys from the Per was talking about, not everything has APIs and you know, not just because they don't want you in some sometime the systems are just old. And so, that's where browser agents come in. They click type like people do and we help developers with this new developer program we launched in the the last conversation to interface with the inter with the web. So some of what we did

28. 13:20

    this week and then I will go deeper on what what we are um we on boarded we have this new developer that is working with us from the new ecosystem. We will announce this collaboration in the end of the year but in the end of the month but with them we help them in understand the web page. So they have these processes that now are done by end and through our agents and the near up they can automate this process and now as a business owner right this is something

29. 13:49

    that scales on compute not on account anymore which means you can scale your business and it can be a venture return rather than just a linear consulting business. So yeah, we use these agents automate the web the team itself I this was my research I've been working on agents research for 10 years now it wasn't called agentic research before it was reinforcement learning but that's

30. 14:14

    what we do my co-ounder is um he was um lead security at Google cloud so we know how to set up and build these agents and now we are trying to help other developers to work work on that what happened this week. What happened in the last two weeks? One, we have we had a minor breakthrough uh where we now can get to 99% accuracy on specific tasks. So when we work together, you tell us this is what the agent should be doing and while normal agents, open source agents or even like

31. 14:47

    the open AI ones get to 60 70% accuracy, we can bring that to 100% for you. Um we we it was a bet we made on the science side. It played out. Now we're trying to work with bigger companies to scale it up and show it to the world.

32. 15:05

    That was very exciting. And on the other side, now we have the capabilities to recreate the internet uh inhouse. So you want to work you know with HubSpot but that's somebody else's website and we can clone it easily. So you can test it in, you know, kind of a staging environment yourself.

33. 15:25

    So between uh this first collaboration big capabilities uh breakthrough and now this new ability to clone the internet. This is mostly what came out last week. Um man 99% is fantastic and is actually a prerequisite for people to actually use these things. Yeah.

34. 15:47

    The caveat is that we need to be very clear about what you want, right? um we are not 99% AGI yet we are 99% non-general intenses and but that's something that we have a repeatable process where for us is easy to achieve right so I'm happy to have more conversations with developers last time it was more about the partnership program now we don't have bandwidth to get yet another person yet but to have

35. 16:16

    the conversation so if some I think I put our uh Twitter account here if you want to write write us what would you like agents to do? Uh we are soon going to be able to get more partners in the program and so you will be on top of the line. Yeah, I actually have a perfect partner for you guys. Um, it's a consulting group that automates uh practices for like retail businesses like totally not focused on crypto whatsoever and they have no idea what's really going on in the world of AI either, but like are going to benefit

36. 16:47

    from automation and so and it will free up their people to work on more meaningful things like uh for a long time I've been worried about AI taking over you know jobs. It turns out AI is taking over the crappy jobs and the people are not getting fired but they go to do something something more meaningful in the same company. So I'm really excited about what what's happening now. That's awesome. At some

37. 17:12

    point I'd love to get a demo on like how it all how the platform really looks. H yes yes we I think this is scheduled for mid of next month. uh like now we had I think two months of Q internal QA things are are pretty good now and um for I I want to work with developers at first because in the end of the day the people that actually build useful things are developers uh B2C kind of cases are just to showcase as you say the demo to understand the

38. 17:47

    capabilities cool what kind of developers like what is the background of these developers that you want to work with are they like already a part of existing companies are these more independent people hackathon participants like what's your target within the near hub I want to see people that are have this like I want to enable people that thinks all in the world in the right way like

39. 18:11

    what the crypto environment is good at is thinking about what the future should be like and putting together all these correct technologies like long-term thinking technologies we have our go to the motion for the enterprise case that there is no need for that but I want to see those people that really believe in a better future and want to help them to create this better future and so this is why this channel I think it's important

40. 18:36

    to us so if you are a developer that wants to build something a little bit more ambitious ambitious that you know what is today your capabilities I'm really happy to help got it uh what about like background like programming languages that they know they need Python experts typescript obviously you're not writing smart contracts soility doesn't matter but yeah it doesn't really matter I guess you know as you say like TypeScript or any

41. 19:01

    JavaScript uh language or Python those are the best cases mostly because they're the easiest one and fastest one to iterate uh we can work with C++ agents I don't think that's a good idea yeah well I'll get I'll get some people um this is All very exciting. But how are you actually like facilitating this? Are you like hopping on calls with each of these people and kind of walking them

42. 19:30

    through? Okay, got it. Right now, that's why we limit the amount of people we can get as this uh developer partner program. Like we will work with you to figure out whether like our agents are the fit. At the first call, it will mostly be about me trying to convince you not to use it. And if I fail, then we continue to the following cause to see how can you use it. Um, you know,

43. 19:54

    and I kind of want to I want to double click on this uh benchmark that you you say that you've like had this breakthrough. How do you know that you have this breakthrough? So we the way we do science internally is we work on a smaller simpler benchmark so we can create more experiments and we know that small benchmark is predictive of the biggest one right for the general population we need to show the bigger one working for

44. 20:19

    ourselves we know they have the same complexity but they don't have the full about about themselves. So we know we solve the our problems in the small case and now it's just about time matter of time and of course we we work with people that handle the you know more complex benchmarks and work and and cases and we just show them how we solve the new problem and we're like do you think this is a valuable solution this is is this meaningful you know when I was in academia will be like is this a

45. 20:50

    publishable paper and will it be great and the comment we got from the owner like the creators of the bigger benchmarks were were like what the this is meaningful uh this is a meaningful change. So wait when you do have this meaningful breakthrough like what next because it feels like the space is moving so quickly that even if you went to go write a paper and do the whole academia thing like by the time you go talk about that benchmark in the conference it's going to be there's be a new benchmark

46. 21:21

    or there's be another breakthrough. So like h what next like it seems that you just need to like build this into production and get users. So right now we are in this space where people that build benchmarks you know there are the uh people building capabilities and the one building benchmarks the ones building benchmarks are really thinking like hey if we build it too simple they will not do it. So what they're starting

47. 21:46

    to do is they create a benchmark where if you can saturate that they know it will be very useful and they will be selling it. So it's almost as you know I have a customer they want to do this thing I'm going to publish a benchmark and if you eat it I just sell you to the customer I get a fee on top of that. Um so in this case it's like that the moment we can show the biggest benchmark these guys are like we created this because this was what we needed and now

48. 22:12

    you can do it so I need you right. uh we are finally getting you know the academic and the real world cases to collide and because the measuring whether success was achieved versus achieving success is easier uh so they in some cases it's easier so in these cases people are more eager to work together you know do the decision I I used to work in academia and like there's a bit of there's this lag that exists and so what I'm actually really interested like it's pretty much a a

49. 22:43

    marketing push at this points for you to then say, "Hey, look, we have this like we break through. Hey, corporate partners like we do this better than others. Come work with us." But I guess how Yeah, go ahead. Imagine you're like now there's so much noise in the eye market that if you want to buy any kind of any product, you need to create an eval benchmark, right? So what these

50. 23:07

    companies do is they create it anyway because that's how they make the buying decision like which LLMs to use. you just benchmark all of them and so they already have those ones. So by necessity companies started to have their own internal evolves and they just evalu solution against their own.

51. 23:25

    So I think the academia didn't move just the business world came our way. Uh which is great. Uh but on the other side like now people are way more um more intelligent about what they want. they they can call like the new generation of AI model from you know code they get released and people like oh this is only 1% better I don't care right or maybe this is a meaningfully better model and so I just want to buy it man this opens up a huge rabbit hole

52. 23:57

    that we probably need to talk over a beer or something about like the role of academia in in industry because academia is I'm not saying dying actually like enrollment rates are higher than ever but yeah like the role of it in industry is changing drastically because these companies have these resources internally.

53. 24:14

    Honestly, I think like academia as always had the one job we kind of forgot over for a little bit and that one job is to think about 10 20 30 years timeline but because now we have this kind of singularity looking like system. what you your pattern of 50 years is now five years and so people need to start thinking about what is the 50 years within an exponential uh growing thing.

54. 24:38

    So we just need to retrain the academic academic mandate to be like hey you need to think longer. I know you're thinking the way you were thinking but technology is exploding so you need to go longer and the safety stuff is that like it's thinking longer. Hey man, let's do let's do another podcast or something about this because uh this is this has changed from what did you ship this week to let's unpack academia. Um all good man

55. 25:03

    this is great. Um so congratulations on the breakthrough. This is super exciting. Um I want to see proof. Uh give me that benchmark. Give me that eval. That'd be great. Um and whenever you can start talking about this and like proving it more publicly um would love to support. I mean like that's what we are focusing on. like we saw the we saw the spark and now we need to make the fire.

56. 25:27

    Cool. So so exciting. Well, thank you again for hopping on. Um next up we have Zahed from Jutsu. I see my Zahed are you here? Yes. All right, welcome back. How you doing? I'm doing good. Thank you. And I love seeing the proof also and I like love to show the proof that we have done something and shipped something this week. Yeah, we'll uh we'll get Manuel to open up at some point. Um but yeah, we love to see your proof.

57. 25:58

    I know him. I trust him. So uh and I know his work. So just like I'm talking about jutsu and post. Well, let's see it, man. Let's let's see what you got. Okay. So, is it the usual like a Yeah. What's who are you? What is post? What did you ship this week? Okay. So, hey everyone. Uh I'm Zahit, founder of Jutsu. So, Jutsu is a AI agent platform specially focused on the enterprise. But today I'm going to show our consumerf

58. 26:33

    facing product called post. So, Post is a uh AI social media manager agent that can uh kind of write like you and cons consistently post you in social media uh especially if you're a busy founder uh who probably better spend time on fundraising or other building the product so post can take over the the social media part of it. So this week we uh kind of a launch five four especially four features. Uh one of them is we are calling it branding but it think about

59. 27:12

    it's more like a what doubleganger showed you uh where you can clone uh someone's uh profile. In our case you you can basically clone someone's writing style from their LinkedIn profile. Uh and then co-pilot we launch a a very advanced copilot and pricing and referral. So let's uh yeah let's start uh see the demo. How do I present? Share screen. Share screen.

60. 27:44

    Okay. Okay. So first of all post is now launched. So you can go and um kind of a start using it. Uh you can sign up for the free trial for the first month and then uh based on your need you can subscribe to different versions. So it's it's live go to post.ai and when you sign into the system um this is how it looks like. Uh today uh especially you can create post.

61. 28:18

    So you already know this thing. Uh today I'm mostly interested to show you the kind of the cloning feature or we call it uh create a writer. So imagine like a uh most of you can use a tool like chat GPD or cloud and to write a LinkedIn post, right? So it's it's simple. You can just go and say like hey write a post about X. But the biggest problem in if you're using charged or or claude is it does not understand your brand. It

62. 28:48

    does not have any context about your writing style. It does not have any context about who are you targeting, who is your uh which location you are targeting if you're a business, what kind of tone you want to use when you will write a post. So those kind of things uh if you use charged or plot it does not have that context. This is where you can basically create a writer.

63. 29:13

    So a writer is imagine like uh there was example of uh let's say Cameroon. So if he want to uh kind of a create a uh writing style like him uh he can do it. So I was trying to find Cameron Post. Um so we just need two or three. I don't see but let's say manual. So I don't post very much on like the problem.

64. 29:41

    Imagine like uh we can give uh this one as example. So we can say let's go to his profile. Uh I can add that this one. And then let's find one one more. Uh this one maybe uh this one is very small. This one probably okay. So now if you have if you just take these two and then let's add another example and then uh I can you can name it. Uh typically naming is better because I'll show you why. So you can you can name the persona.

65. 30:24

    Okay. So that now you click generate and it does uh generate the u like embedding and all the thing that we need. So now imagine like this is there. So if you now want to create a normal post now all of a sudden you have access to all the uh all the post uh all the character or all the writer that you we call it writer that you created. So uh now if I want to write a post um let's say there are some um kind of a predefined that we can go like a finding

66. 31:01

    the trends um and then let's do a s uh generate. So what it is doing now it is taking this topic it is taking this writing style and then it is also doing the web search and then it will come up with a uh with a post that resemble uh manual.

67. 31:22

    So let let it uh give it few seconds. So like you can see like he used a lot of this kind of thing and then roughly like a this is how probably he organized his his writing style. If we go back to his writing style uh which one is this one don't know which example we give.

68. 31:49

    So we go brand kid. So, so you can see like it find emoji and all these things. So, this is one way like a I can you can see like I have many different uh writers or writing style. So, I can switch between them whenever I'm generating code uh generating the post. So, imagine this can be your uh all your um LinkedIn example and you can generate your writing style. So, that's that's one piece. The second piece is the co-pilot.

69. 32:22

    So sometime people still want to use uh uh use large language model to kind of a not only just generating a post but also like a okay which time I should post or or uh like you still want to utilize like a newer model for doing certain things. So uh now with our copilot so it has access to very powerful tool. Uh what that mean is basically it can create post uh just from prompt and then

70. 32:56

    it has like a you can see it will adapt the u UX based on uh based on the things that you are asking. So like a just for example I was asking about weather uh then it will show it different way but when you are trying to generate a post it will do everything and then it will kind of a visualize how it will look like in in LinkedIn and then you can of course go and uh you can add images if you want. So think about I want to add

71. 33:26

    this paper for this thing uh with this uh post. So you can do all those things and then you can schedule whenever you want to schedule. So then the other thing we did is u kind of a created a referral program. So imagine you are using post and you found it interesting and you want to share it with other. So you can earn by sharing with other. So if you go to settings referral so you

72. 33:56

    have a unique referral code uh that you can share with your friend and then you can earn uh next 30 days. So like a uh if you are if you want to share you can pretty much use post forever for free by sharing with other people. So this is another very powerful referral program. So an incentive for you first if you find it useful then uh if you want to use it for free go use the referral code and invite friend. Uh then all the other thing like a uh subscription now we have a different

73. 34:32

    subscription model. Right now we are using um stripe but pretty soon uh we are thinking about adding uh near and other uh like a um cryptocurrency to pay pay for the subscription. Uh yeah, so the these are the things that we kind of a mainly shipped this week and if you are interested to kind of run your uh social media in autopilot so just go use post and the last thing is link uh Twitter probably it will come next. So right now if you want to use in

74. 35:13

    for LinkedIn and if you want to automate uh give it a try. Thank you. Zahe, this is awesome. Um, I have a uh social media manager joining the team soon and so we'll be getting them to use post which is exciting. Um, question though around the referral program. What are you using a third party service for this or did you build everything in house? You know us, we build everything in house. So I

75. 35:39

    showed a showed this referral program to a uh mentor. He said like that itself is a product uh but that's not a product. So yeah, referral program on top of uh Stripe. So we have we build our own. Cool. That's awesome. Well, I'm talking to a bunch of teams right now that are interested in a referral program. So if they want to use it like we can we can help them uh with that. So imagine like adding crypto web three also in the back end. It's easy. So because we build it, we know how it works. That's amazing.

76. 36:11

    Um, any closing thoughts besides use the referral program, like check out the referral program. Uh, I mean, obviously check out Post. I'm going to start experimenting with it tomorrow again. Uh, give you some feedback. Besides that, what else should people do?

77. 36:30

    So I oh one of one of the other thing I'm kind of planning or thinking about doing is something I'm try going to call it uh a podcast founders who code because lot of founder like me who are builder uh struggle a lot with go to market and kind of a figuring out the other side of the business not the technical side. So I have been learning a lot like a last few months you know that like a so and I think I also want to grow and learn from other founder. So having some sort of like this one like

78. 37:03

    you have like where you share what we ship I think it will be great to kind of a have something for builder technical builders who don't necessarily think about the business side and learn after a year doing building all this cool stuff. uh so I think that's something I want to uh do and if you are interested if you're a technical founder if you want to share your uh your thoughts your

79. 37:29

    experience being a technical founder how you transition or how you you solve that problem that will be wonderful and I think the other thing will be um I'm we are looking for a our kind of a first I would say like a BD or sell sales B or like a someone who can uh demo a lot of the capability that we have like a Tutsu you know like we we have a really really good product which is like a Lego boxes and we are

80. 38:03

    helping enterprise to solve real problem uh and we are work already talking with many enterprise so I need now a person I have idea of which is the customer profile and everything so I need a person who can do the top of the funnel filtering out so that like free up my time. So, I'm looking for a person. So, if anyone in this call or if you know anyone, that would be wonderful. I do know people and would love a job description. That's my prerequisite. I need a JD. Okay, cool.

81. 38:37

    Yeah, I'll do. Awesome. Well, cool, man. This has been great. Uh, super excited to try this out more and we'll give you more feedback tomorrow. Thank you. Bye. Awesome. See you. Last but not least, we have Harsh from the Bitcoin agent in Alpha Alpha Devs Labs. Harsh, how's it going? Hi. Hi, Cameron. Can you hear me?

82. 38:56

    Yep. Loud and clear. Can you just give everyone a quick overview of uh what is Alpha Dev Labs? What is Bitcoin agent? Kind of give a little bit of the story of how you got here and also just introduce yourself. Yeah. So, hi everyone. Uh I'm Hush. I am founder of Devs Labs. We are a team building AI agents and specifically focusing on near AI right now. And uh last week we shipped Bitcoin agent. Uh so let me just get started with uh what is Bitcoin agent and what can you do today with

83. 39:29

    this with this agent. So yep let me present my screen and get started. So uh Bitcoin agent is uh an AI agent that you can use to interact with Bitcoin L1 using just your near account. Right. So just a little bit of history, we started uh during one trillance agents hack and we built a test net prototype where you could basically send a transaction on bitcoin test net with it and uh from there on we picked up and

84. 39:58

    we got introduced to a lot of more people in the ecosystem and we built our MVP uh in this time and recently we have launched version one which is the mainet version. So without further ado, let's see what this agent is capable of doing, right? So u so this bitcoin agent you can basically go to a dashboard which we have developed. There you will be able to see the agent wallet and we'll be able to fund this wallet with the with real bitcoin and this is the account

85. 40:28

    which is uh on bitcoin man. This is not something which is like u uh like on derived or some kind of a bridging or anything like that. So you'll be taken to a to to a chatbot context which is right now running in pity and you'll be able to see your bitcoin balance and this wallet will be directly accessible through your near account. So we will have a demo where we will be sending

86. 40:52

    1,000 sets to one of the u bitcoin address and uh you will have to just approve the mcp uh mpc payload that you will receive uh the uh receive from the agent that uh that will basically handle anything from creating the payload to relaying it on bitcoin uh mainet and you will have the transaction settled on bitcoin without worrying about handling keys on different chains or or basically if you have one year account you will be able to access any number of chains we did it

87. 41:29

    with bitcoin right now we plan on to do it with other chains so that's a quick demo I won't get into architecture much but the way this thing works is we leverage uh chain signatures which are available uh on near uh near mainet and we get those uh sign those signatures and really them on on Bitcoin, right? So, we stitch all this thing together.

88. 41:53

    Here are some links you can try it. So, just um uh to conclude what we are trying to do and uh what our goal is that uh we are building in for agents on here. That's the main focus that we have. We are developing new MCP tools uh to make chain signatures easier uh to use with uh any agent, right? So, we started with agent that could send transactions on Bitcoin L1. So ideal progression here is something like this.

89. 42:20

    You should be able to transact with your Bitcoin. Then you should be easily able to lend and borrow with your Bitcoin and eventually trade and manage your portfolio from an AI agent interface without ever leaving near or even worrying about which chain or wallet to use. Right? So we want to use this experience and knowledge that we have uh with MCP and near chain signatures uh to develop this thing uh as an open source tooling and help agent ecosystem grow.

90. 42:46

    So while building Bitcoin agent we were able to test and work with most of the infra that near has to offer and definitely I can say that near is the uh near is the chain for AI. So moving forward uh we look forward to actively contribute uh towards building this MCP infra targeting near sign uh near uh chain signatures as of now. And uh yeah uh and last note bitcoin agent will soon be live for uh anyone to use with neari.

91. 43:13

    It is live today. You can use it inside bit and AI agent interface. You can fund your wallet and test the transaction. You can send uh bitcoin to anyone with just uh from that chatbot u chatbot context. Right. So yeah. So that's all from my uh end uh teams looking forward to contribute and keep building on your AI. Harsh this is great. I'm super excited about this future of you know using agents to simplify crypto transactions. I think crypto is way too complicated for the vast majority of

92. 43:44

    people and then obviously Bitcoin is like the gold standard. So, uh this is really cool. I'm super excited to have you back on the show when there's more progress made on the MCP side of things. Uh for those who have not seen the previous episodes, Alex Kmerford from the Near AI team built an MCP server that can communicate with any near contract API so you're able to write in natural language and execute transactions. Uh what we'd like to do with the Alpha Dev Labs team is to sort

93. 44:11

    of extend this to not just your transactions but transactions across any chain um and then have this MCP server be pluggable for anyone else to use. Yeah, we are exactly working on this and actually uh internally we tried it with cloud desktop and we are uh adding all the uh tools that we have developed to the MCP server that the MCP server that you shared with me in the last meeting.

94. 44:34

    So we are hopefully like in couple of days maybe in a day or two we should be able to have at least bitcoin signatures getting reloaded from from uh the native mcb server. So it showing yeah this is like you're actually manifesting the vision for blockchain for AI because you know you first need the protocol to scale you need the chain abstraction to make sure blockchains are visible and then the agents piece is a kind of at

95. 44:57

    say like the end userf facing component because those agents and users don't really care about what blockchain they're on or anything else and so and they also want to interact in natural language so I love the direction of this uh super exciting what can we help with like what can the community there's for you know 1,800 people watching right now. Uh what should we do? Uh so recently when we launched the agent, we

96. 45:22

    got a really good feedback from the uh internal team who tested the agent. Uh so I will want everybody to basically share u their feedback because I want people to use it from the uh consumer point of view because we had a lot of developer feedback on our codebase uh and we were able to like uh get all of these things fixed but I want kind of a user experience and consumer side of uh

97. 45:45

    feedback on it because we are really early in this. We really want to understand how a user uh would want to use chain um in in with with this kind of uh uh interface. So I I think this as a complete change because in near future you will not be clicking buttons around or basically visiting links. Uh it will more more be uh about interacting. You will have a foundation model. You'll be connecting different services through

98. 46:11

    these MCP servers and every single uh product and service that you use will eventually is going to have MCP server. So I would like to see how uh people uh use it and how these things evolve so that we are on the on the right path while while building all of this. So if somebody's providing you feedback, what kind of form should they fill out or what kind of form do you want this

99. 46:33

    feedback in? Um because standardizing feedback and so you can actually interpret it properly to change product direction and you know edit features. Um I'm assuming you don't want a bunch of people bombarding you in DMs. form that they can fill out. Yeah. Okay. So, so I think I think we can do some kind of a private beta where we can launch um a form on our uh social media uh handles

100. 46:57

     and uh we can just gather it from from maybe 20 to 30 people who can use this uh MCP. Perfect. Yeah. Uh create that beta form. I'll get the nearweek team. We'll all work together to drive traction for it. and yeah, let's build like, you know, truly chain abstracted AI because uh sooner or later the chain's not going to matter. And I love that you see the vision of like websites not going away completely, but you know, everything

101. 47:27

     will have an MCP server and people are just going to interact with their agents. And this is still super early, but you know, I know the new AI team is very bullish on this sort of thesis of uh you know, everything will be agentic, but not 100% yet. We need better uh just reliability. So, Silverstream's uh sort of I guess breakthrough in reliability for the browser use agents as long as

102. 47:52

     you can very specifically you can tell them exactly what to do. I think it's a huge step forward because like especially with money and this is something that I'd love to probably talk about more uh asynchronously is uh around the security of LM messing up.

103. 48:07

     They hallucinate all the time. Uh I have a lot of examples of this. Uh but how do we make sure that these LLMs or agents don't hallucinate with people's money? Yeah self yeah I think self custody and like at the same time having a agentic u web I think that is something uh which we need to balance because the other companies uh don't have u those uh kind of uh responsibility because uh when we

104. 48:36

     uh talk about decentralization uh user ownership is u as uh important as any other thing right self-custody is as important so like navigating this security is going to be u really really crucial. We need to uh do it the right way. Right.

105. 48:56

     Awesome man. This has been great. Uh I have one last thing to share but before uh you hop off is there anything else that you want to share with everybody? No no uh it it is really uh great like it was my first experience coming uh on and uh it was blast. I got to learn a lot from other founders who are building and it was really great. Yeah. Yeah. Well, you're welcome back whenever.

106. 49:18

     Sure. Thank you. Awesome harsh. I'll see you. Alrighty. Last but not least, uh I have something to share which is fun. Uh let me see my computer. Give me one second. Present. Great. So, uh, the Proximity Labs team was not able to make it this time around, but I wanted to make sure that they got a super amazing shout out, uh, for shipping new documentation for shade agents. And for those who don't

107. 49:48

     know, shade agents solve one of the biggest problems in web 3 right now, and that is building truly decentralized agents that can control accounts across any chain where you can verify that even if the machine that's running that agent goes down, you don't lose your keys.

108. 50:03

     This is critically important if we expect agents to actually facilitate commerce or do anything important in the future. And uh they ship these new docs and these new docs give you solid uh graphics of how everything works. Uh there's, you know, very clear function calls and just overviews on exactly like how to build one. There's really great uh templates and examples that uh

109. 50:27

     they're only getting better. I'm actually working with a bunch of teams right now to increase I guess the different types of utility shade agents can provide. Uh but these docs are brand new and would love people's feedback on exactly what you think and how they can be improved. Uh I'm super optimistic about this direction. I think every agent will eventually have to be verified and that verification will have

110. 50:52

     have to happen on chain or inside of a TE and then have some sort of testation that goes on chain and this is something that is here live right now that you can use and uh one other shade agent that is absolutely worth looking into in case you haven't already is this mindshare agent. Uh let me pull it up properly. So this mind share agent is able to, you know, take information from Kaiito's API and trade on behalf of a user. And so this is live right now. It's fully

111. 51:23

     autonomous. I'm super excited about the use of shade agents in Dows, uh, decentralized autonomous organizations for those who don't know. And these things are primarily used to govern certain protocols, but the problem is that humans don't really like to vote and like to participate in governance.

112. 51:40

     And so I do believe that these shade agents will provide that uh flexibility and trust uh and actual use case to uh vote on behalf of a community in a doubt uh by taking things like sentiment off of Twitter or sentiment inside of a bunch of Telegram groups and uh and voting. And so I'm working on a couple projects that uh can help facilitate this. Hopefully we'll have them on in

113. 52:04

     the future. But I did want to flag that there are brand new docs with really great examples. Shout out to the Proximity team for for building this and shipping this. Uh I think this is a really big step forward for the ecosystem. So uh we'll add those docs in the chat uh or like the followup after.

114. 52:24

     And that's about it. That's what we got this week. Uh we'll have more next week. Uh and until then follow Near Week. Uh they're going to be coming out with more short clips highlighting exactly what people shipped last week with a lot of the call to actions that they flagged.

115. 52:39

     And that's all I got. Hope to see you next week. Thank you everyone for joining. Thanks for watching. Subscribe to Near Week.