🐙 Building With Hermes Agent transcript

Hello guys. Well, at least you know that I'm real seeing as my computer crashed. [laughter] How's it going everybody? How are you guys doing? >> It crashed. >> Thank you so much for joining. Thank you so much everybody for waking up or being up at 2:00 a.m. I am sorry it's very difficult to accommodate all time zones. It is currently 10 p.m. for me. Okay, so we have a few people from US. Okay, so hey Richard. It's been a while. 7 a.m.

East Coast time. Sorry, West Coast time. West Coast time. It's 1000 a.m. East Coast time. Michigan. Hello. Hermes hacked her camera. Don't say that. Don't jinx it. I have actually been playing with like hardware. So, it's going to be my next video. Okay. I'm like I just got to show you guys. This is very exciting. Um, I'm using a Pi and then connecting it, running local models off it, and also like using the ESP32 as well.

And it's actually able to like control my camera and stuff. So, it's very exciting. I'm excited. to show you guys. Ethiopia 5:00 a.m. Oh my goodness. Thank you so much for joining. Hello. I didn't even have coffee yet. I'm trying to entertain you then. Okay. All right. Let's get going. Hello. What are we doing today? We're going to be talking about That's a great question. Great question. Um, yes, I will be talking about Hermes setup cuz I've made a couple videos about it.

Let me know, comment Hermes if you've seen any of my videos that I made about Hermes. But I feel like I didn't really do it justice. And we do have a workshop that's currently enrolling right now. So this workshop I'm actually going to be like teaching you how to build all these things too. So I really want to like show you guys what it is that we have. I have spent like months building this up and I am very proud of it.

And I'm also like um cuz I'm connecting it to stuff like the aura ring. So thank you guys again for people who suggested that I get the aura ring. um connecting to that uh and also using like other like hardware and stuff like that. So experimenting using like really small uh devices and then being able to run Hermes and then also run different local models too. That's what I'm experimenting with and it's actually really cool.

Like okay, getting ahead of myself. Really cool. Really cool. Um Hermes bag. No, no, I don't have a Hermes bag, I'm afraid. Hermes. Yes. Nice. Okay, some of you guys have seen my Hermes video. Okay, great. Some of you guys have not. So, I want to show you guys what I have. Okay. Um, so I want to show you guys uh let me share my screen. [sighs and gasps] Okay, one second. Hold on. Let me share my screen. Okay, there we go.

That should work. Great. Hermes live stream as you guys can see. Okay, cool. So, I want to show you guys a couple demos first. So, the first one is Tacobot. Um, so live demos, guys. I hope it works. Oh my gosh, it's a live demo. So, this is like actually our Discord. Um, for us, our internal team, we use Discord for everything. We use Slack for like external team and stuff, but I want to show you guys like Tacobot. Um, actually, let me start off.

So, we use this a lot for work. So I can go like at tacobot remind me in five minutes to move on to livebot. Okay. So it should be able to set that reminder for us. And it can do certain things like for example I can go to our work updates for example and then I can say something like at Tacobot uh what is everybody working on right now? And it should be able to do that. And [sighs] let's see. Okay. So, you can see that it's connecting to notion.

So, we have our meetings in notion as well. So, it's connecting there, our meeting notes, and it can see what's over there. It can also read all of Discord, um, calendar, and a lot of other sources like GitHub as well. So, it's able to gather like what everybody on the team is doing. So, let that run for a bit. Okay. So, yeah. and it's able to schedule reminders for us. Let's go back and stare at this for a bit. Okay, cool.

So, based on the latest team meeting notes, here's what we're currently doing. So, setup guide for the Hermes workshop uh workshop calendar and lifebot access rollout. Uh what this is what Joy is doing. Jonathan and Cthur are doing this. Ken is doing this and the whole team is prepping for our Hermes workshop launch. Actually, have you guys seen the Are you guys on the newsletter? Comment workshop if you are registered for the Hermes workshop.

I'm very excited to see you guys there cuz we're going to actually I'm going to be like teaching you how to build all of these systems um during the workshop. So, comment workshop if you if you are going to be there. Um another thing that I can do here is say for example I we have like this YouTube trends tool, right? like for let's see let's see like here okay um I can ask it to like create a process doc on how this product works and it's able to like go through and actually explain how all our products work and things like that um and all of it is actually being stored in Obsidian so I know some of you guys have been watching the Obsidian video so I just want to show you like what that looks like uh in Obsidian as well let's go to Obsidian Yeah.

So everything gets stored here in Obsidian like here um YouTube topic engine and it would actually just like have all of these process doc. So then every time you need to have documentation for things you're able to very easily document it for any business owners out there or people who work in teams. This is like it it is honestly so useful to be able just have this bot that can tell you and document everything that you're doing.

Uh they can generate reports for you as well. So we have weekly reports that are generated um through Taco Bell as well. So you don't have to like go and make your own dashboard's own reports anymore. Yeah, pretty amazing, right? All of this is being run by a Hermes agent. Um let me go back. No, wrong one. This one. Cool. Yeah. So all this is being generated uh by a Hermes agent. So, it's actually one of the only that is one of the profiles that we have on my on my Hermes agent.

Comment Taco Bell if you get the vibe why that it's useful. And do you like do you think do you guys think like something like this would be useful for you? Like have you implemented something like this before? Hey, okay. Some of you guys are here from the workshop. Amazing. Thank you so much. I'm very excited to see you guys there. I do not have the funds for enrolling the workshop, but it sounds exciting. No problem.

I also have lots of free resources that you can check out from my videos as well. The workshop is for people who want to be able to build all of this within a span of 2 hours because it took me like months and months to figure out how to build it, right? So, for those of you who go to the workshop, it's going to be like a 2hour sprint and we're going to get all of this done. Taco. Taco. Nice. Amazing. Um, is this workshop suitable for beginners?

Yes, it is. It is completely beginner friendly. Um I did get your point. This is something definitely useful I need to build. Right. Like yeah like I will literally show you. So this is only one of the profiles that I have. Um and I want to show you like the others. So I have quite a few profiles but the other profile that I use a lot um is called Lifebot. So this one is Lifebot. Oh, by the way, I'm actually just going to put this here if you guys want to check out the workshop.

Um enrollment is open now. So enrollment for Hermes workshop here. So do check it out if you are interested in doing that. Let me just pin that comment as well. Pin message. Okay. Some questions. Who is the workshop for? I just joined. workshop is for people who want to build the things that I'm going to be showing you right now and do it in two hours as opposed to like be like me and spend like months doing it. That's who the workshop is for.

Completely beginner friendly as well if you wish. If you wish. Um, if you're not free during the workshop time, do you record it? Yes, everything is recorded as well. And you also get resources and ideas like how to build like step-by-step guides on how to build things as well. So even if you not only will you will we be building this like I will be building it. You can also follow along build it if you care out there.

It is recorded and we also have like fullon checklist and guides that you can look at yourself and then build it too. How is it different from openclaw? So Hermes is actually pretty similar to open claw but in terms of premise but I find that it's easier to build around. That's the major thing. Like open claw is really cool. was like very magical when you first use it. But I found that after about like a month or so, it started to get clunkier and clunkier.

Things started breaking. U maybe because they had like new updates being done. But with Hermes, when something is actually built, it doesn't just break. So, for example, Taco I've been using for like months now. And so is Lifebot. Speaking of which, let me show you guys LifeBot. I'm also going to be going into more detail about like my building process a little bit later in the live stream as well. So, just want to show you guys the demos first because I'm pretty proud of them.

Um, okay. So this is like for example it actually uh notes down everything like I track everything using livebot like what the stuff that I'm building um like what I'm working on also I also track like how much food I'm eating and things like that. For example, I had an egg tart and half sandwiches today. So yeah, I'm able to do that. Um, and I'm able to ask it questions like, let's see, uh, [sighs and gasps] how is my progress for today compared?

Like, how is my work progress today, for example? It should actually be like not that good, I think, because I don't think I log too much time. But let's see. Um, yeah. So, I use it to just like answer. Okay, there you go. I use it to just ask it a lot of questions. Um, because it has like it's based upon my actual data, right? which means that it's able to give me very very specific and specialized advice. If you ever had like you know productivity people like tell you oh like you should use this counter you should do this you should do that where like I was trying to do fasting for example oftentimes like it's not tailored towards you specifically so then you end up not being able to do it but because lifebot is built using like your specific data built using Hermes that's how it's able to make sure it's giving you like very specific advice and it's I've been using this for about two months now and it's freaking amazing.

So, for example, it shows that my Pomodoro log today is for one session. Um, my health metrics are stale right now. It my health metrics I update this every night, so it hasn't come in yet because it's only 10 p.m. And also says that here's the things that I've done like review live stream slides and Tesla demos end to end before going live, you know, which I did. And yeah, right now we are running the 1000 p.m. live stream active now as it knows and it tells me like what my work in progress is um etal analysis log volume. this is well below your historical median.

So, it's able to compare that ex um executive execution shape. Today was not a built or deep work day. So, it knows like the type of days that I'm having, it's able to give me like the best advice for that specific day. Like it's it's very subtle sometimes. I feel like I'm not able to like explain it, do it justice. But because it's so tailored towards me, it's able to give me very very tailored advice. Like I can also ask it questions like how do I improve my productivity overall and log more deep work hours, right?

Like I can ask it something like that. Good morning. So excited for the workshop. Yes. I'm really excited to show you guys during the workshop. Like, I think you'll be blown away how much you can actually do, especially with the setup that I'm going to show you guys how to do it. Like, you're able to build things so quickly now if you know how to do it. It's pretty crazy. Um, okay. Let's see. It is reading. It's going through the skills right now.

Have I used runpod? Yes. Um, I've used runpod. use it mostly for to rent GPUs. That's what I've been doing with run pod. Once I have Hermes plus Hermes bot set up on my MacBook Pro with only 18 gigs unifi, how do I choose LMS? How do I choose different LMS once I get computer with more memory? Honestly, 18 gigs is completely enough, especially if you're using cloud models. Latency is crazy. Yep, I also do have a VPN on right now, so I think that's not helping.

Um, okay. So, see how it's able to give me very tailored advice like, uh, anchor your first timer by 1 p.m. Basically, my my number one lever for me specifically is that if I'm able to start working before 1, like by 1:00 p.m., I'm able to get a lot more work in. And this is so true. Like, every hour past 1 p.m. cost you 20 to 25 minutes of total daily focus. So, the rule is that lock in at least 20 20 minute timer before 1 p.m. every day, even if it's just planning or light prep.

This has helped so much. Uh log time duration for 20 to 25 minutes. I am someone who is incapable of focusing for long periods of time. So if I do like an hour, try to work for an hour, that's just not going to work. I'm just going to give up. Like literally, I like just just like I just give up. So Lifebot was able to figure out like I need to be doing 20 to 25 minute um pomodoro sessions to for it to work for me. Using 10-minute task bridges to protect scripting.

So depending on the type of task that I do as well, there's a different percentage of completion rates. So a hack that I've been doing is that for me like scripting is really hard, right? Like it it takes up like a lot of my mental energy. So I would do something like do some task first to kind of ease into scripting and then always like making sure that I don't stop working immediately after a scripting session. So it's like these very subtle like little tweaks that you can do, little hacks that you can do that's very specific to me that has really genuinely like it changes the way that I work and I feel so much more productive now.

I'm able to get so much done as well. So, etc., etc. It's able to do all this and all of this data. Um, like how I was able to get all these data. I'm going to show you guys as well. It is Where's my little guys? Oops. Give me one second. I had to restart my computer. So, let me show you my little apps. AK. So I am able to do this because of these little guys over here. So this one is I'm able to have the checklist of everything that I'm doing.

And then here it's like I have the pomodoros and every time I log time into this I can do like 1 minute test for example all of this and when I check things off it all gets logged into obsidian. So let me show you an obsidian. Um this is such a game changer. Honestly such a game changer. So my typewriter I'm able to log like every single day that I'm doing. And the more days that I have the more data uh life bot is able to get like Hermes is able to get.

Again, this is all running off Hermes right now. Um, and also with my workouts. So, there there's like workout stuff and my pomodoros, too. So, all my Pomodoro logs that I've been doing, the more data that it gets, the commentary that it gets, the better that it becomes. So, over time, since I've been using it for about like two months now, it has been it it's gotten so good at giving me very, very good feedback. So, genuinely highly recommend doing that.

Um, yeah. So, that is a little demo of what you can do with Hermes. And I'll tell you guys some other stuff I'm doing with Hermes, but I don't want to just like have this only be a demo. Um, but yeah. Do you guys have any questions so far about what it look of what it is? Let me see. Do you guys have any questions? Avocado. Cute and helpful to stay focused. Indeed, it is. I signed up, but the screen refreshed before I could look at the other offer that appeared.

I did email you. If you emailed us, we will get back to you. Um, please give us a couple hours. We should be able to get back to you. Uh, thank you for emailing us. So, yeah, the we also are selling the like lifebot completely. Like you can just get it all. So, that's what it was. Anyways, do you really use LM successfully? I mean, all of this is based off of Hermes, which is using LMS. So, what is Hermes? Good question.

Hermes is an agent harness that allows you to build a custom agent that you can do things like Taco Bell, which I just showed you, Lifebot, that which I just showed you. So, these are all built using Hermes. And let me actually show you, let's dissect it a little bit. Okay, so we saw Taco earlier. Um, oh yeah, I got really excited. So, what are we going to be covering today? So, I guess I showed you guys the demos. Um, so we're going to be talking about Taco Bot's profile and also Lifebots's profile.

And then I want to show you the stack underneath. So what are the tools that I use in order to build this as well as the step-by-step process that I use to build something like this? Cuz before to build something like this, this would have taken me like weeks probably to like get this done. But with like a certain I do like a certain sequence of steps that I do. I've gotten it to a point in which I can get it done um probably like half a day like and during the during the workshop I'm going to show you how to do this like I've gotten it the way I can build using Hermes now I can like get entire things like entire projects done within like 10 20 minutes.

It's crazy. Uh and it totally works on like other devices and things like that as well. So it's like spent I perfected the way at least for me I perfected the way to be able to to do that. Um and then I want to talk about some other builds some ideas as well as some ideas that you can build using Hermes as well. Okay so Tacobot so for each Hermes agent it requires certain parts. You can think of it kind of like Lego right like you have um your Hermes agent and you can customize it depending on how it is that you want to do.

So for Taco Bell for example, we can see that it lives on Discord, right? So Tacobot lives on Discord and the model that I'm using right now is using Gemini 3.8 Flash because it's really fast, but sometimes I do want to have it completely local. Um, which I would use the Quen 3.635B model. And the home for Taco is on the Mac Studio that I have here. You don't need to have a Mac Studio. You can run it in a lot of different places, including like a VPS, um, on your computer, a lot of different things.

But I happen to just have it on my Mac Studio and you can and then you need to give it certain data access. For me, I gave it data access to notion, GitHub, calendar, meetings, uh recordings, and discord threads. That's why it's able to answer questions, right? It's able to answer questions uh that I ask it and it's also able to store that information as well. So, in a memory, it has what is called honcho shared memory.

So, it's able to have like specific memory. Not going to go into much detail about that. Um but it also has long-term memory, which is on Obsidian, which I showed you guys earlier. Any questions about Taco Bot, how it's built? Will you share hardware requirements recommendation prior to workshop? Yes, we will. But honestly, I would be shocked if you don't have the hardware requirements. The only thing I can think of that would make it make you unable to do it is if you don't have a laptop at all.

If you have like only if you only have a phone, I think that's going to be hard. But pretty much even if you have like a Raspberry Pi, that would be fine. um as hardware. Like really with hardware, I think people don't realize how accessible it actually is now. Like if you have like 4 GB of RAM, that's enough for hardware. Yeah, you don't get to have like local models, but you can always use cloud models. Um hi Tina, I'm using Grockbot for very similar kind of bots.

How do you compare both things for I think they're completely different, right? They're very different sorts of things. With Hermes, it's the whole point of Hermes, like why is it that we use Hermes as opposed to something else is because it allows you to switch out models. You're not like bound to to like Grock, for example. You're not bound to OpenAI, for example. You're able to switch out models and you can also make it completely local, too.

So, sometimes there's things that you maybe don't want these tech companies to have the information that you have, right? So, um you can use completely local models and have Hermes running locally, too. So, none of this information ever gets out. so powerful, especially if you're in a team and you don't want, you know, your team secrets to be leaking out. Um, workshop. Yes, I'm excited to see you guys. So, what amount of memory would you recommend for local models?

I think to have a decent local model, 16 gigs is I would say you can get a very very decent model with 16 gigs. 8 gigs I think is okay. Like you can have a decent conversation with it. Um, you know, like I you want you guys want to guess what do you think is the smallest mo model that you can run that um is able to run a model like smallest amount of RAM available to run a model. I'm just curious if you guys can guess.

A little hint, my ESP32 is able to run a model. Like this tiny little thing over here is able to run a model. Take a guess. >> [snorts] >> 0.5. Yeah. Yeah. 0.5. Exactly. Um, so the ones I think I believe it is three megabytes. It was able to run a model that is three megabytes. That is and you're able to have like you're this one you can't really have a conversation with, but it's able to generate stories. These are called like tiny stories models.

And yeah, like that's that's it. This is how like how small of a device you can actually use in order to run models. So don't be worried that your device is not good enough to run Hermes or run models. Yeah. Good guess. Yay. Good job. [sighs and gasps] Um, okay. So, I I already showed you guys Lifebot. So, that's the way that Lifebot works. Um, if you guys have seen my Obsidian video, Lifebot is heavily dependent on Obsidian because that is where it gathers all of that data and stores it as a AI database in Obsidian.

And Lifebot, which is a Hermes agent, is able to access that directly. Um so because of that database like very very high quality data that is what makes lifebot really really good and here are the specs for lifebot. So communication wise it's on telegram as I have just shown you the model that I use I currently am using deepseek v4 pro um but I also sometimes use GPD 5.6 Terra, uh, depending on, I don't know, the the vibe that I feel like that day where sometimes there's like free credits for certain things, I would use that.

Um, but they're both really good. And if I want to use a local model, my favorite model is still the Quen 3.635B. I honestly think it's one of like the best models to be able to run locally cuz yeah, like it's just good enough where you don't need like a massive machine to be able to run it. Uh, but it's able to pretty much do everything that you need. The home is also on the Mac Studio. Again, it doesn't need to be.

It can be on any device. Like if you have a laptop, chances are you're going to be able to run it completely fine. Data access, as I said, Pomodoro, typewriter, Apple Health or Ring, and food tracking and exercise logs. Like I have this all stored in the AI database. And in terms of memory, um is the same setup as Taco. Uh so Honcho shared memory. So it's able to get information as well as long-term obsidian memory. Any questions about Lifebot? question.

Obsidian is more of a archive/ library than persistent memory. How do you deal with outdated or superseded material in Obsidian? Great question. Uh so the short answer is that I make sure that what is being written is of a high quality. For example, if I'm doing logs and stuff, right? It's not going to be information that's going to become irrelevant because it's very very well formatted. The long answer is that I use something called an LLM wiki.

So, it's able to prune that aformation over time. It actually does it automatically. Workshop. Yes. You see everything I just showed you, right? Lifebot um and and tacobot like we're going to be you guys will be building this yourselves on your computer or whatever device it is that you're choosing to VPS whatever machine. Um yes, we're going to be you will have your very own versions of you can build these things. Um let's see which Gemma model which Gemma model not sure I understand the question I do have an anthropic cloud subscription can I use it with Hermes you can use your claude code with Hermes yes that's fine okay [sighs and gasps] next slide okay so makes sense right very cool comment lifebot if you understand why lifeb is is is amazing and useful and important.

Um, if you get if you get if you get a livebot, if you get it, if you get it, if you get it, you get it. If you get it. Yes. Yes. Okay. I'm really glad that you guys get it. Like, lifeb single-handedly has really really changed how productive I am. And not just that, like also like tracking stuff. Um, I'm the kind of person that's like I always wanted to track my exercise. I wanted to track like my food intake and all these things, but I'm just I'm like way too lazy.

I can't handle it actually having to track it myself. So, I just get Lifebot to do all of it for me. Someone said Wifebot. Hey, you can try you can try Wifebot. I don't [laughter] know. I cannot confirm or deny if that is good or not. Um, okay. Let's move on. So let's talk about the actual tools that I use. Okay, so the stack underneath all of the Hermes agents that I build. So obviously the first one is having Hermes agent which is the home base.

So um it has a desktop app, a terminal or a dashboard. So you can have all these different things. Um let me actually show you try to do this without like crashing the system. Okay, I might end up having to crash it. Okay, I'm gonna try um I do I'm going to actually log into my Mac Studio right now and I want to show you like the Hermes itself what it actually looks like. Hopefully it lets me do that. Okay. Okay, I think we're being a bit laggy because my laptop might not be able to handle this right now.

Okay, I'm going to try. If it lags, then alas. Wait, that's not going to work. Uh, okay. Okay, so this is me logging in to my Mac Studio, right, where a lot of these Hermes agents are running. And you can see that this is Hermes over here. Um, yeah, it's really slow because I'm trying to live stream at the same time. But this is the Hermes desktop app and it's really easy. This is what I mean. You don't even need to know how to code.

Like you can say like install life file for me. I'm not going to run this right now just cuz you know we're live right now. But this is what it looks like and you can have like different agents that you can talk to directly um as well. So there's like different sessions that you can talk into directly. So that's one one way to access it. You can also access it through terminal. So for example, you can go in terminal over here and say something like oops you can also do say h live chat and that loads the Hermes agent as well.

So I can say you can see that has Gemini 3.8 flash. You can be like, "Hi." Yeah. And it's able to talk to you. So, there's like um this is what it looks like. Just kind of show you guys what that looks like. That is what it looks like. Um so, you can do it through your terminal using code, but you can also do it using the desktop app. So, either way, if you know how to code, you don't know how to code, you can do it both ways.

So, it's actually really, really accessible. Okay, let's see if anybody has any questions. Does anybody have any questions about what Hermes looks like? How many of you guys have actually tried running Hermes agent? Just write Hermes if you've actually run your own Hermes agent just for me to kind of get a vibe. Let me just also move go back to the slides. Okay, cool. How many of you guys have ran? You done your Okay, nice.

So, a lot of you guys have done it, right? Okay, great. So, a lot of you have tried. So, for those of you that have tried, who haven't tried, um, do you feel like what you've built has been useful for you so far? For those of you who haven't tried, I hope that kind of shows you like it's really not super scary, intimidating to actually get started using Hermes. There's like they have a full desktop app now, so you don't even need to know how to code, which is pretty cool, right? me.

I don't know how to create more than one agent/bot in Hermes. You can make multiple profiles. That's what they're called. Multiple profiles. So useful. I have Hermes, but I don't know how to set up the different agent. You got set up. Hopefully can learn that from you. Yeah. Um you can So I have a lot of free content on this where you can join the workshop where we actually be taking through this with you like step by step.

But really like I promise you it is not as difficult as you think it is. um once you actually get the hang of it. I don't have API key so I can't use it a lot. That is fair. Olama versus Llama CPP for Hermes. Great question. Yeah. So Olama is actually built on top of Llama CPP. So it's a easier way for you to access um local models if you use Olama. However, if you want to have more control and also be faster, you would just use Llama CPP directly.

I use llama CPP directly. Uh you do need to know like you you got to like do a little bit of code if you're going to use Llama CPP. So, you know, it's not that hard, but you do need to know a little bit of code. So, if you can, I would use Llama CPP. Uh but if you want to just get it like really fast and running, you can use a Lama. That's the distinction. Okay. All right. So, anyways, so what do you need to actually build what I just showed you, right?

These things I just showed you. So, obviously, first you need to have a Hermes agent. It's free. Just download it. um then you need AI chat. So with AI chat, what I mean is I like to use cloud or open AAI um in order to think through what I'm going to be building first. I'll show you guys in a little bit uh what I mean by that. But I actually always want to like think through what I'm building first with AI before I actually go and execute it.

Um you also would need an AI coding agent. I mean you don't need a coding agent but it does make things a lot faster if you have an AI coding agent if you're using codeex cloud code where you can even use open code an open source version because this can help you execute a lot faster if you're using an AI coding agent. Um another thing that you need is an Obsidian AI database. So queryable record which I just showed you earlier of past build specs like a lot of information that you can store into your Obsidian AI database.

And then this one is kind of optional but I think it's really helpful. It's an Obsidian LM wiki. How many of you guys have heard about LM Wiki before? Comment LM Wiki if you have heard about it before. Um, put simply, an LM wiki is a managed library of different sources that you can query and it's fully managed by AI. So, it's like you're able to take notes and then that a the AI is able to like go through it, scan it, get rid of stale stuff, update it over time, and you can also query and ask your questions, too.

So, I kind of want to show you guys. I have a LM wiki that's dedicated to Hermes actually because every time I find something new that I think is interesting, I would put it into my LM wiki so that when I want to build using it later, I can ask my wiki bot um who access this BLM wiki to be able to give me information about it. For example, I'm just giving you example. I'm like maybe I come across this really cool like holographic memory, right?

Wow, so interesting. What is holographic memory? Who knows? And it's like what's this is in Chinese? What's that mean? you know. Um, and I can do something like I hope this works. Live demos. [laughter] Oh, please work. Yeah. So, all I have to do is put the link over here through my wiki bot. Again, wikbot is also a Hermes agent. So, see how flexible Hermes agents are. Um, it's also a Hermes agent. Let's wait a little bit.

Hopefully, it's able to digest the information that I give it. LM wiki. Okay, great. Some of you guys know. Nice. A lot of you guys actually have done LM wikis. That's pretty cool. Please. Why is it being so slow? Okay. I hope I'm not like crashing it. Let's answer some questions. Oh, please. I know. Live demos, you know, live demos. I mean, it's actually worked thus far. So, if this doesn't this doesn't work, I'm not going to be super shy.

I wouldn't even be mad because honestly it's been like pretty decent up until now. As you see, I was like literally testing it at like 6:42 p.m. and it was fine. So, but it's a live demo. Okay, let's see if it's going to work or not. I'm going answer some questions. Oh man, obviously it's going to not work right now. Okay, I'm going to try again. I'm going to try to query it. Um, tell me about how to make a conbon board.

Oh, okay. No. Okay. Never. I'm sorry. It was It was It was just being slow cuz ah darn it. It's okay. Let's answer some more questions. Um, don't worry. It happened to me in my presentation yesterday, right? Like obviously if it's going to fail, it has to do it. It has to do it during your live demo. Always. Always. Um, show us the live logs is working on LM as it's working on LM Wiki. Uh, I'm not going to do that right now cuz I got a SSH into my Mac Studio and I don't want to crash everything cuz I'm live streaming right now.

So, going to have to trust that it's doing Hey, see, look. Look, it's doing it right now. Actually, you can just read here. Yes. [laughter] Um, you're brave just for trying live demos, right? I've been showing you guys live demos the entire time. So, thank you. I I feel I feel like I'm I'm being acknowledged for how brave I am right now. Thank you. Glad you have issues, too. I have issues every day, like constantly. Uh, what do you use sation to your Mac Studio?

I just uh use Oh, I use tail scale. Yes. What is the cost of ingesting data in LM Wiki in terms of token usage? Great question. It depends on the model that you're using. If you're using a local model, it's free. Any good resources to find models that work with one or two gigs of GPUs with Olama? Uh, you can just ask an AI to do that. Just ask like what what works, you know, that works. Hey, it actually worked. Wow, isn't that amazing that it actually worked?

Okay, so let's see, let's see, let's see about holographic memory. So, holographic is experimental local first persistent memory engine for AI agents. Its main features are blah blah blah. So, it tells you like and then it also tells me how to create dedicated combon board. So, blah blah blah all these different things. Okay, so what does it actually look like? If I go to my Obsidian, I can actually show you what it looks like.

What my little this version of my LM wiki looks like. Please work. Please. Um, okay. Obsidian over here. Um, [sighs] so go to my wiki. Okay. Whoa. Okay. Queries, concepts. So, I have a lot of concepts that are here. Raw. Okay. So, here are the raw things that I ingested. I think it's because I interrupted it earlier due to my impatience that you don't actually see the um I'm just going to try this again just because I do want it to ingest.

So it should ingest. So let me just try this again. Um it's going to ingest it at some point. But you can see like here's the things that all the raw things that it's stored. So it would actually document all of this and then it's able to make a bunch of connections. um based upon these. So this is what the LM wiki actually looks like. So not going to go into way too much detail right now just because not that much time.

Oh wow, we're already at 10:47. Wow. Okay. Yeah. So then um it's able to ingest it and then it's also able to query it. So that's why it's able to tell me like, oh here's how you can build something like a combine board, right? Um and it's able to cross reference that with my actual Hermes build. So it can tell me exact steps for doing things. So that's the other tool that I like to use when I'm building. Um, yes. Let me go back to the slides.

I get so excited when I talk about the builds. Okay, cool. So, let's let's let it do this. It's currently extracting and ingesting all of this into DLM wiki, but let's go back to the slides first. So, these are the things that I use in order to do my builds. All right. Um, does anybody have any questions about any of these tools that I just described before I talk about how I use these tools to build it? Open obsidian stuff about my life.

No. Obsidian versus Ellen wiki comparison. Why need both? Obsidian is a tool. LM wiki is the implementation of an LLM managed uh Wikipedia page. Bas not Wikipedia page like what's it called like wiki like wiki resource. You're just you're using Obsidian in order to implement an LM wiki and to visualize the LM wiki. So completely different things. Does LM wiki work the same way Hermes does? So, I actually used Hermes to build my LM wiki.

My LM wiki is managed by my Hermes. Everything I'm showing you guys because this is a Hermes live stream. Like everything I'm showing you guys, I built using Hermes. Yes. Pretty cool, right? [laughter] I'm telling you, like Hermes is freaking pretty magical in my opinion. Um, okay. I'm going to Okay, so it says that it has been captured. So, it's going to show up in Obsidian now. So, let's actually see if it shows up in Obsidian.

I want to show you guys real quick before I move on. There you go. See, it showed up in my Obsidian now. Great. Um, okay. Let's go back. I want to show you guys the actual process that I built this cuz I think this part is really important. So, I don't want to rush that. Uh, okay. Yes. So how do I actually build this? So from idea to running the agent, right? So what I do is like I have a I basically collect a bunch of things that I want to build where I think it's interesting for my um for my Hermes build, right?

Because you can customize and build so many different things with it. So I collect it and I use my LM wiki in order to send it links of like cool things that I find so that it's able to store that in that wiki. Um it's filed by Wikbot which is managed by Hermes. when the time comes I'm like I actually really want to build this thing I would query the LM wiki like I did earlier say like I want to build stuff using the combin board it would actually tell me like step by step how does I do it so in order to implement this um I would then go to a chat like claude or open AI and then I would come up with a runbook so I can't actually show you my claude right now because I don't have a VPN turned on because I worry that it's going to kill my stream right now so I'm going to show you what the runbook looks though I work on this runbook um with Claude.

Where is it? [snorts] Please show up. Where'd you go? Run book. Okay. Okay. Yes, it is not beautiful. So, sorry about that. You don't have to read it. Like, it's just the concept of it. So I basically come up with this run book using claude based upon the information I have in my in my LM wiki and then it would come up with like for this is like literally one of the builds that I was doing right so it's called the Hermes expansion road map and it would come up with different phases this is when I was building taco bot which is the loot team bot um and then phase B which is lifebot uh phase C retry patterns blah blah blah so it literally has this rum book and then what I do which is this is the part okay are you guys ready this is this is like the magical part okay After I built this runbook, I would go and just use a coding agent and just give it to the coding agent and ask it to build it for me.

That's it. It just does it. It like that's that's what you do like just open cloud code or like open code or codeex whatever you want to use and then just give it like I want to say like I just go and type in like I want to run this runbook, right? So I would just give it like phase a like this is what I want to do and just give it to your coding agent and it's able to implement this. Isn't that crazy? So in some ways like what is the most like useful thing in terms of what you're building is the actual runbook itself.

Once you actually have the runbook in terms of what it is that you want to build and you figure out what it is that you want to build the rest of it you can just hand it off to a coding agent. So what I said here the slow step is the point. So put the effort into figuring out the runbook up front even if it takes days or weeks so you're not spending even more time fixing the build afterwards. Because if you don't spend time on your run run run run run run run run run run run run run run run run run run run run runbook then you're not going to be building the right things and then you're going to be very sad as you're trying to fix things.

So this part is the most important part. I was telling you earlier like why is it that I can like build this entire thing on live stream and I can like during our workshop I can do this in two hours, right? But it took me like months to actually get to this point is because I was trying to figure out what the runbook is supposed to be. But once I figured out how to do it, the execution step is so fast if you're using a coding agent.

Okay, any questions? I think this is like legit like probably the most important part like the process of execution of building your Hermes. Um, is this so any questions? Why does Hermes sometimes give me answers that are trashy? Why does Hermes sometimes hallucinate? Um, it can be a few reasons. Maybe it's because you're not using a good model. It could also be because you're not prompting Hermes. So, you're not prompting it, giving it the right prompts.

That could also be the case. also one of the reasons why you want your runbook to be really good because if you are that's probably like two of the major reasons why that could be the case. Do you read it? Read the runbook. Um I do read it. I skim through it like just to make sure it's okay. But as I'm building it with Claude War Open AI um as I'm building it, I would be like asking it to just going back and forth like I try to understand everything that it's actually building.

So I do try my best to actually understand what it's building and not just like oh whatever like do whatever you want you know I do try my best to actually understand it. Yeah, I don't read it like I I don't like look at every single command on it uh because that would take me forever. But I do read generally speaking like what is in there. Like for example, we look at phase eight, right? So I I get the steps. So build plus test.

So I'm like, oh okay, like stage one is building both bot presents in both servers from day one. Um staging is done by access control. And then like okay, this looks about right. And then I go like stage two is team roll out. Oh, this makes sense, right? Because if I'm building a bot when I was putting a taco bot into Discord, I wanted to actually test that it worked first by itself before I put it into Discord where the entire team has access to it, right?

So that's why um I do have this like general staging that I would do. I also asked it to give me like time estimates which is kind of like it's not always very accurate because of the coding agents. You're able to do it much faster than the time estimates. So it would tell you that oh it's going to take like a week or like you know 30 minutes or something but if you use a coding agent um it takes like a few minutes to implement something like this.

Um is it actually true that Hermes gets better or improves the more we use it? Yes it is true. It creates skills by itself. Yeah. Your runbook is PRD plus implementation. Am I correct? Yes. I think that's a good way of Yeah. Yeah. You know what that's a good way of putting it. Yes it is. It's like PRD plus implementation and done so in a way that I can give to coding agents so that they're able to build the thing that I want specifically how I want it.

What model are you using for your Hermes? Depends on which agent. I use different models for different agents. How do you optimize which model Hermes uses for each task? Oh, okay. That's kind of like general understanding of different models. Um, for example, if I want something really fast and I don't want it to be costing me like a lot of money and I don't want to be running a local model, for example, I would just use probably like Gemini or uh GPT Terra or GPT Luna.

If I want something to be super fast, I would do that. Uh, I've actually I really like using Claude Code for example and Claude, but I very rarely use Claude models for my Hermes agent because even though it runs on my like I have a subscription and I would run using my like Claude subscription, I just feel like it's really expensive and it doesn't work um it's not really worth the amount of money. It just it burns my credits really hard.

So I actually don't generally use any cloud models for Hermes. Um, do you have courses to learn more fundamentals on how you build? I do. I have a lot of videos on that. Should check them out. Um, how do you prompt to get a good run book? That is a very good question. I kind of like I can't really show you right now because of the VPN situation, but I think that that is a skill. Like I just like I've gotten really good at figuring out like what's the right questions to ask.

So, I think you're just going to have to like see it at some point. Um, you just need to like think about all of the edge cases as you're building the runbook and you gain that intuition over time. Sorry, I hope I can I wish I can give you a better answer than that. Um, how to add knowledge in Hermes since it doesn't support rag by using something like a AI database like I showed you in Obsidian. Use Tina's videos for learning.

Her videos rock. Oh, thank you. I appreciate it. What model are you using for building the agents? Oh, that's a good question. So, I'm using coding agents to build it. So, I primarily use codeex and cloud code. Those are my two. I sometimes use open code as well, but I primarily use codeex and cloud code. call code is what I would um use if it's something that is pretty like linear and just like it just needs to implement code directly right but if I want to build something that has more like images um that has needs like you know visual stuff I would use codeex for that yeah okay 108 right I'm going to continue on just cuz we're running out of time.

Okay, so yes, I have some other builds which I don't really have time to show you guys right now as demos, but um these are some other builds. Hopefully, you know, some ideas that you have. So, for example, I have a coder one that is able to draft a spec. I can draft a spec on Discord and then it would go and just ship the build directly by itself. So, this is like autonomous coding. Um, so I have that. I also have a watchdog which is kind of boring but it's a alert channel that makes sure all my agents are online and doing okay at all times.

Um I have the wiki bot which I showed you guys already. This guy over here oops oops ignore that. I have rebal bot which I'm building as well. This is an always on research agent that is turning things that it finds like ideas that it finds online into different ideas for content um also for execution like for like building software and stuff like that. So this is Revelbot um that I'm building as well. And then the newest one that I built is called Sensei.

Um and it turns books and resources into skills. So, I don't know about you guys, but if you ever like read a book, especially like a non-fiction kind of like self-help kind of book, it would give you like exercise. It will give you like this framework. You're like, "What?" And as you're reading it, you're like, "Wow, yes, amazing." You know, makes sense. I'mma execute on this. And then by the time you finish reading the book, you just kind of like forget about it or you're too lazy and you don't actually do any of the exercises.

So imagine if you can turn a book into a personal coach that's able to like go through everything and actually get you to execute on the books that you read. So I built Sensei um in order to do that. So testing it out. Um this one I did on my BPS as well. So yeah, this one is a really cool one. I think if you if you want to experiment with something and you like reading books, especially like non-fiction, self-help books, things like that, this is a really cool one to build. and some other ideas that I'm currently looking into as well.

Maybe some ideas for you. So, it would include things like productivity like dictation uh to structure document. If you're someone who likes to talk a lot into Telegram, for example, you can like dictate it and can just directly make it into a structured doc. Uh people who have like X to notebook LM podcast. So, you can pull like Xlisten books like you know just pull them there and then compile them into a notebook. um meeting auto transcriber self-explanatory.

So these are some ideas um health and wellness like journal into obsidian so you can dump your day's log at night and it gets filed properly into your obsidian. So this is another Hermes agent you can build. These are all like different types of Hermes agents ideas that you can build with on the life side plain English home control. So if you have like um what is it called? Like Alexa or you know like any type of um IoT devices that that can control your home.

You can also have a Hermes agent that controls that directly which is pretty cool. So you get your like Jarvis situation deal watcher. You can have an agent that is able to look at prices for you. Um say if you're want to buy something, you want to see when it's going to go on sale, you can have a Hermes agent that's constantly looking at it and will alert you when there's a sale that's happening for the thing you want to buy.

Um, on the OP site, group chat digest, we have that already. Um, so it's able to summarize a bunch of chats. I don't know if you have like a lot of WhatsApp groups or Discord groups. So, it's able to summarize everything for you. Compiling skills into code. This is a really good one as well. um over time if you build a lot of skills using LMS if you want to it might be because like LMS um if you're using text it's inherently going to be not as stable as deterministic right so you can also compile your skills that you're building your pipelines into code to make sure that it's much more stable over time so this one is a little bit like abstract but it's a really useful one if you're someone who wants to make sure that your workflows are very very streamlined and you don't potentially have issues that pop up.

That's that's a good one to build. [sighs] Yeah. So, question, which one are you excited to build with Hermes? I hope like this kind of gives you some ideas of things that you can build like all the things that you can build using Hermes. And put into the chat. I want to know like what kind of things do you want to build with Hermes Sensei? Very good ideas. I have a lot of books. Right. Yeah. Um, how many telegram bot I can connect to Hermes?

You can have as many Hermes profiles as you want and you can connect a Telegram bot to each of them if you want. Yeah. Sufu can sense SU to debate a topic. True. Uh do you have local agent? Yes. So Hermes agents are inherently all local agents and if you use a local model it's a fully local setup. So is it it is like a Jarvis in real life if we tune it correctly. Yes, if you build your runbook properly, if you do that properly, then yes, you can absolutely do that.

How do you get the books to the bot? Good question. So you can have PDFs or audio versions of it. Um that's usually the best honestly if you just like look at summaries that people have like a lot of people have like text summaries as well. So you can use that to ingest into your Hermes to build out the skills. Do you stream on Twitch? No, I do not stream on Twitch. No, I do not. Maybe I should. How are your electronics learning going on?

It's going good. I was showing you guys earlier like I've been building a lot of stuff with electronics. Um kind of like building with models, but I started off like just connecting with like literally just like sensors um and like different LED lights and stuff. And I'm trying to figure out like how to use AI models to control them. So using an AI model in order uh using AI model with Hermes. Usually with Hermes or without Hermes in order to control these like hardware stuff.

Yeah. So it's going good. I would say I'm still too scared to actually solder. I feel like that's like the ne next step for me is to learn how to solder. And I'm just kind of scared. [laughter] I did have to take a hardware class when I was in school, but I was I think I almost failed that class. I'm just Yeah. I don't know. I think at that time I was just not into hardware. Hi. Hello. Is there a better choice between Telegram and Discord?

Yeah, there's a lot of choices. I just happen to use Telegram and Discord a lot. You can use WhatsApp. You can use like they I think there's like 10 15 different ways that you can use I can't like think of them off top of my head. Like I message like there's a lot of ways that you can communicate with it. It doesn't need to be Telegram or Discord. Is it hardware you're talking about IoT? Yes. Learn to solder at 14 years old.

Dang, that's pretty cool. My local AI on Hermes agent is so dumb. What do I do? Depends. Make it less dumb. What's wrong with it? I mean, like I said, I really think like learning the fundamentals, like going through figuring out it took me like a really long time to figure out as well. Like seriously, it really did. So, if you're just like trying to figure it out all by yourself constantly and assuming that your full-time job is not learning things, which is my full-time job, it makes sense that it's not like great in the beginning, right?

Like it's fine. Um, what's the best free API provider you would recommend? I think like if you have um it kind of depends because some people sometimes they have it on how to say local models, right? Cuz that's completely free if you just download them locally and use them then they're completely free. So that's I guess the the best API the free the best free API if you just download the models locally. Open router. Yes.

Did I say open code? Yes. Open router. That's what I meant. Open router. Yes. Um, yes. Okay. Any other questions? Let's see. H. Okay. Yes. Little plug again. If you are interested in building all of these things, I will teach you like all the basics. Um, as well as how like the runbook, right? Like how to actually go through like I'll be actually building it on the spot. Like we'll be doing the workshop. You'll be following along.

So, pretty much everything I've learned over the past few months and all the experience that I have, I would be showing you all of that into like a 2-hour workshop. So, if you are interested, you can check it out. And yes, I do have a guarantee as well. 7-day 100% money back guarantee, no questions asked. So, I am confident that this will be really, really helpful to people who actually are interested in building the things that I've just shown you guys.

So, that's why 7-day 100% money back guarantee. Sounds good. Any questions? Oh, and for anybody who wants the slides, uh, I do send out these slides as well. So, feel free to here. I'll put that here. You can sign up for this and then if you sign up for it, I will send you the slides that I just showed for this live stream. There. If we can't be there for the live workshop, can it be available for later? Yes, it will be fully recorded.

You also get the runbooks. Um, everything will be there so you can use it anytime. When is boot camp coming? [gasps and sighs] Ooh, good question. I can't make any promises right now. We're restructuring a lot of stuff. That's why I haven't done like another boot camp yet cuz I do feel like the boot camp I want to like make sure everything is like very up to date. So yeah, not yet. When is the Hermes workshop? It is on October 3rd.

Have you tried OpenCloud 2.0? Lots of changes since you last used it. You know what? I haven't tried it in a while. So, you know what? Maybe I will try it. Thank you for the suggestion. Yes, please. Okay. Yeah, if you want to do the slides, then please um sign up for the newsletter and we'll I'll send you the size in like a few hours. Yes. Oh, yes. Boot camp. Yes, boot camp. The AI agents boot camp. Yep. We're revamping a lot of stuff.

So, do sign up for the newsletter as well. We will be giving you more information about that. And if you do the workshop, you also get $100 off for the um boot camp as well. So, if you wish, is this beginner friendly? Yes, completely beginner friendly. No coding required. You can code if you want. Uh but no coding required at all. I'll be showing you how to do everything, which I've spent months and months doing trying to figure this out.

So, trust me, it's going to save you so much time. Will you be launching IoT workshops soon? Would you guys actually want me to do that? I've been thinking like, should I make more content at least like starting off with content first? Like, should I make more content about like IoT um and like how I'm building stuff and running models like using devices and things? Maybe I will if you guys are interested. Can you show the Quen model you said earlier was very good?

Yeah, I actually send you I'll do you better. I'm going to send you a link for it. Um, okay. Let me find it for you. This model is the one I use locally the most. There's also a um smaller version of this. The quen models are really good in the sense that there's so many different sizes available. So if you can't do the 36B pram, you can do the 27B pram. That's also pretty good and they have lower ones as well. So just choose whichever.

Like I think if you're starting off with a model, um going with the Quen family is usually a good choice just cuz they have so much variability. So much not variability, variety that you can choose from depending on how big your device is. That would be my suggestion. Thank you so much. Thank you so much. I really appreciate you supporting the channel. Thank you for the super chat. I can't actually see your Oh, thank you so much for the super chat.

I appreciate it. Happy Cape Dogs. Thank you so much, Le. Appreciate it. How would you recommend attending your class if you already have a lot of things partially set up? Oh, yes. Great question. Everything that we do is going to be on different profiles. So, if you do have a Hermes setup already there, don't worry at all. Like, everything we're doing, it's not going to affect your current Hermes. So, do not worry. Uh, there's a new 27B that is faster.

Yes, I think in the Quen family, right? Yes, I think so. I haven't tried that one out yet. How is it? If anybody has tried it, do you think I can run locally with Lenovo Yoga ThinkPad? Honestly, you can I can run local models on this little guy over here. Okay, so this has like I think 8 megabytes of RAM. So, yes, you can pretty much run if you have like a laptop, you'll be fine of any f of any sort. If you have a laptop, you'll be fine.

Uh, thank you, Rexie. Have you tried a new Quen 3.8? I have tried it, but I haven't really tested it extensively. It's not I think Quen 3.8 is not open. It's the one that's not open source, right? I haven't tried it like extensively. So, I did try I just tried it, but I'm pretty sure this is Wait, is Wait, wait. Is it the open source one or is it 3.7 that's not open source? No, no, no. 3.8 is the one that's open source.

Yes. Yes. I think I think No, this one is good. This one was good. Oh, yeah. This one was really good. 3.7 is the one that wasn't open source. So, I tried that one, which I wasn't sure. Can't make it on 3rd of October. Well, if you want, it is all recorded as well. And you can always email us if you have extra questions, too. So, that works. But up to you guys if you wish. If you wish. Um, do you try DeepSync models?

Yeah, one of my favorite models is the DeepSync V4 Pro. That's probably my favorite model to use because it's cheap and it's also very multiunctional. Like um, it can do a lot of different things. Bonsai is the new 27B smaller in size. Interesting. Okay, thank you. I need to check that out. There's always so many new models coming out. 35B A3B is so much faster than 27B on my 8 GB of VRAM. How are you running a 35B model on a 8 gig VRAMm laptop?

Are you just quantizing it a lot? That shouldn't work. I think um D6 Pro is deprecated. Deep Flash is the new one. That's Wait, I thought No, no, I think those are two different families, right? Is it? Are you sure? Wait, but I've been using Deep Seek Pro. I can't keep up. There's too many updates. Okay. Okay, good to know. I'm going to try that out. I think I still am using Deep Sync Pro. Okay, cool. I'm going to try it out.

Um, [sighs] why Hermes and not Open Claw? That's a good question. So, I did start off using Open Claw and somebody did say that, you know what, maybe Open Claw is good to revisit again. But the reason why I switched to Hermes was because I wanted to build more like capable more like extensive things that that's more stable over time. And I found that Hermes was a lot more stable and also because it's self-learning. I felt that it was just easier to use as well.

That was the major reason. Although the open cloud does have more integrations for sure. So that is true. Okay, I'm going to stay for like a couple more minutes and then I shall go to sleep because it's 11:15 p.m. Um, open call open claw to this. Okay, I will try it out. But I don't know. I'm kind of liking my Hermes build right now. I think al the one good thing though is that you can actually transfer Hermes to open claw and open claw to Hermes.

I don't know if you guys knew that. You can actually do that. 11:15 p.m. Where are you based? I am based in Hong Kong right now. Open clown 2. Are you still connected to notebook LM? What does connected mean? I I use Notebook LM. I don't have like personal connections to Google or Notebook LM though. Do you pay for Hermes agent monthly plan? Yeah, I do. Um, so I pay for Hermes the news research the news it's called the news subscription.

News portal. Yeah, news portal. Yes, news portal subscription. I do pay for it monthly. I do like the $20 one and I kind of like top it up if I'm doing something that is a little bit more extensive. Okay. Oops, I disappeared. Okay, cool. All right. Thank you guys so much for joining. I hope this was helpful to you guys. I really hope this is helpful. This is a lot of fun. I haven't done a live stream in a while. Um, okay, cool.

Thanks, Richard. Deep See Pro is still there. Good to know. Great. Yes. I hope this was helpful for you guys. I hope you can see like how much stuff that you can build with Hermes. Uh I've had so much fun building with Hermes in the past like few months. So I really hope that you guys are able to have fun doing it as well. And for those of you who are attending the workshop, thank you so much for choosing to attend this workshop.

Um I promise that we will try our absolute best to give you the best value possible and everything that we have learned, experimented, done for all these months of pretty much grinding it out every single day. We're gonna compact that all together so you're able to build these things that I'm showing you right now as build anything really. After you learn how Hermes work and the mechanism for building, how to come up with the runbook, how to use coding agents, literally you'll be so shocked by how many amazing things that you can build using Hermes.

So really excited to be able to unlock that for you guys. Thank you so much and I will see you guys in the next video or live stream. Have a good rest of your day. See you.