athena-oracle/transcripts/ep_build_agent_os.raw.txt

# oracle/transcripts/chunks/c00.mp3
# model: tiny | lang=en

[0.0s] A gented offering systems full one hour course,
[3.5s] how to build an automate anything.
[5.8s] So today I'm going to show you exactly how to build your own agent
[8.9s] offering a system based on what's working for me.
[11.0s] If you're wondering what an agent offering a system is,
[14.0s] basically it's a way to have all of your agents
[17.5s] inside one beautiful dashboard,
[20.2s] where you also build out custom workflows
[22.9s] and you have everything one click away.
[25.1s] And the amazing thing about this is
[27.4s] that you can basically have everything in one single place
[31.6s] that's ready to automate and offer it as you want.
[35.0s] So let me give an example.
[36.6s] For me, I suspend a lot of time create videos,
[39.3s] and so we create this video agent, the basically does it for us
[42.5s] in one single click.
[44.0s] We have, for example, an SEO content pipeline where we can research stuff,
[47.9s] we can find keywords who can generate content,
[50.2s] and then we just deploy to our website in one single click.
[53.3s] And basically everything that you want to automate,
[55.4s] you can automate using these systems.
[57.4s] So for example, Hermes agent, not very fine in the thermal.
[60.7s] If you're just using the thermal, not that great.
[62.7s] Even if you're using for example Hermes desktop,
[65.1s] you still can't get it exactly how you want.
[67.9s] With this system, for example, we have Hermes Jarvis, which is a voice-operated AI agent
[73.3s] and looks and feels exactly how I want it to look.
[75.7s] It's got all of my projects over here.
[78.1s] And what else I have, for example, at Hermes, I'll recall the pause
[80.5s] and let's trend it news and if it melts.
[82.1s] And we're not seeing that to both stores to show off my own own product.
[84.9s] I just want to show you what's possible.
[86.6s] With this sort of stuff, for example, this one,
[89.0s] this is Hermes Astros, which again, is a custom workflow.
[92.0s] And you can rename this, you can change it how you want,
[94.9s] you can build whatever you want using these systems.
[97.4s] And this, for example, actually takes the latest news from Twitter,
[101.2s] gives us the trending topic ideas.
[104.9s] Looks at different titles we could use for that content.
[108.3s] And then also we can just plug out to the video agent.
[110.8s] So it can automate the video.
[112.0s] We can plug it into notebook, lemon.
[113.5s] Create, for example, free infographics and all that sort of thing.
[116.1s] And then we could create a serial content for you.
[117.8s] And you can build all of this for free using your existing subscriptions as well,
[122.3s] which is very important to note.
[123.6s] So, for example, if you have chat cheapity already,
[126.4s] and you already subscribed to it,
[127.9s] well, you can actually use chat cheapity to build all this out.
[131.4s] I actually personally used Opus 418 inside,
[134.4s] call it an Avise Fable 5 to help me a lot along the way.
[137.0s] And for example, if you have something that's a problem you want to solve,
[141.4s] let me give an example.
[142.6s] So a lot of you ask me like, how do you automate leads and email outreach using something like
[147.1s] Hermes?
[147.8s] So with Hermes, we have a tool here.
[149.6s] The congenerate leads, we just type in exactly what type of leads you want to get here.
[154.2s] It generates a list leads.
[155.8s] Then it has all the campaigns.
[157.2s] It can write the email campaign itself.
[159.3s] Then it can actually send the emails with these campaigns,
[161.9s] it pulls in all the lead details.
[163.7s] And then from there, we can manage the metrics.
[165.7s] We can see how many leads we've got, how many we've validated,
[168.2s] how many ascendable, how many we actually send, how many replied,
[171.1s] and we've got our inbox and sent inbox fully automated with Hermes agent,
[175.8s] it's like one beautiful dashboard.
[177.4s] And that just saves so much time.
[179.7s] So this is the power of this stuff.
[181.4s] This is how amazing it is.
[184.0s] And you will learn how to build all this today.
[187.0s] And I'm going to show you exactly how we can get started.
[190.4s] So if you're learning why I'm the right person to show you this,
[195.3s] I literally spent about three to four hours a day
[198.7s] coding this out and making it better and better.
[200.5s] And there's so many mistakes that I learned along the way,
[202.4s] so many things that I can prove.
[203.8s] So with that, let's get straight into this.
[206.8s] So I'm just going to be using G55.6 as an example.
[209.6s] I quite like soul.
[210.6s] I think it's great to build with.
[212.2s] And you can see that we've got GVT and chatGVT work over here.
[216.5s] It's sort of snap.
[217.7s] Now for example, we set up Hermes agent.
[220.6s] That's fully ready to go with use GVT.5.6 as a brain.
[225.3s] Inside Hermes agent and now we want to say create a dashboard with it.
[229.8s] So what we can do here is we can actually go inside the chat here and we can say, okay,
[235.6s] now create a mission control dashboard where I can operate chatGVT,
[244.6s] codex plus Hermes agent.
[249.5s] And we can say make it clean, make it beautiful, make it interested to use
[255.4s] make sure it actually works.
[257.0s] And then if you have an idea of what you want to create,
[259.8s] so let's say for example, we have a certain design style with that sort of thing,
[262.7s] we can actually use that for inspiration when you're building this out.
[266.6s] So if we wanted to build this setup inside chatGVT,
[270.6s] we can just take screenshot, plug in and just say,
[274.1s] I've attached an example of the style I wanted to create.
[278.5s] And so we can hit enter now and that'll begin building this out.
[283.4s] So if you want to start from zero and you want to build it exactly how you would like it,
[289.5s] in a way that's easy to use, you can go from there.
[292.4s] That'll also show you another way that you can do this.
[294.6s] So if you want to use someone else's template,
[297.8s] so for example, we have a
# oracle/transcripts/chunks/c01.mp3
# model: tiny | lang=en

[0.0s] version of mine inside there, but bottom that we actually update every time I code out for
[4.8s] three or four hours, I will actually save that. So you can see the sit file here. If you
[9.8s] want to do it, how do you install this? See we do the same thing. You can use whatever agent
[15.3s] you prefer so that could be checked with the it could be for example called whatever you prefer.
[20.3s] And we can start a new project. Start from scratch, save that new project. And then we can
[26.7s] add this so we can add the sit file. So we can grab the sit file over here, which is
[34.7s] there. And we can say install and configure this agent or S on a local port we don't already
[43.4s] have running. So I like to run these things locally as well. And we'll come onto that and
[48.2s] why in a second. But now we have two options. So we can use the setup from the Air Profit
[53.7s] for the bottom and that's just kind of like ready to go and all the other option that
[57.6s] we have is if you just want to build this out free from scratch you can go and build it inside
[62.9s] chat to you with you as you can see. So now it's going to say okay what should the dashboard be
[68.1s] able to do. So says before implementation, I need three decisions that material effect the
[74.8s] architecture. So we have Asian control. We have workspace. We have application format. Now for
[81.1s] me, I'm going to say just allow it to run locally for me in Chrome. And have one fix to
[90.4s] folder and allow it to be fully operational. So this is something important to note. CC when it's
[96.2s] asking you it's asking you like do you want it for operations so that it can actually like edit
[99.6s] files. It can do things autonomously for you or do you want it kind of like a chat. For me, I
[104.8s] don't want a chat because if I just want a chat I'll just go to chat to chat to be here.
[109.8s] And that is exactly why I actually created the agent OS because when you're just going back and
[116.1s] forth inside the chat here, it's super limited. Like it doesn't feel authentic, it doesn't feel
[120.1s] powerful. It doesn't really have everything you want and it's very, very limited in terms of what
[124.8s] it could do. So that's why I prefer to be like way more powerful. The whole point of this is to build
[128.9s] something absolutely amazing that can be a little bit of an elephant and just runs like magic. So that's
[134.2s] why we are setting up like this system. So you can see over here this is beginning to install now as well.
[142.9s] Alright so you can see that we've got two options here. We have the agent OS been installed locally from
[147.6s] template and we have the Hermes agent OS over here too. See two different options. Now both of these
[152.8s] will run like you can see I actually want to switch this to fast mode. So I'm going to say switch to fast mode now.
[159.1s] Just so that we can get faster outputs from both of these. That's already enabled which is great.
[164.4s] And now when we're using here you can see that it's beginning to use this local setup. So you'll see
[172.0s] a number like this. Now you might be wondering why you're running it locally, why not just run it on VPS.
[177.5s] Here's something to consider. Okay and this is something that I saw when I was looking at all in the
[182.5s] stakes from open core and a lot of the issues that people face is that there were two things with agent. So you
[188.2s] have to be very careful of. Number one is that you have to be careful because when you're building
[194.3s] with this sort of stuff, if you give access to multiple different devices then someone accidentally could
[202.2s] access your agent OS and if it can edit files if it can control your computer if you can do everything like that
[207.9s] it's not that secure. If you if you've given the world access to it. So that's why I don't recommend that. Instead
[213.9s] what I recommend is that you use the system like this and you run it locally. Now you can run it on
[219.4s] VPS. We have plenty of AR profitable and members who do that and then they just limit the tools that
[224.2s] they give access to but it's up to you how you want to do it. Now you can see that we already have the agent
[228.9s] OS installed over here. So this is fully installed as you can see here. We've got a friend setup. It's set up on a
[236.0s] separate agent OS local port. So for example if we go over here this is my actual agent OS and then we have
[243.6s] the template over here. So it's pretty easy and the thing to note here is when you're using this you just need to
[248.8s] configure everything. So it's now running using the template that gave it but I'm going to say configure it
[255.7s] here. I'm just going to make this simple if you're doing the sound. We're going to rename this project and we're
[260.8s] going to call that template setup. And then we're going to use this project here and we're going to call
[268.8s] starting from scratch. We'll just call this new agent OS example to a two different options right there.
[279.6s] So now it is said okay before I configure it just we're just going to go through step by step.
[288.9s] Set it up at each part that doesn't require an API and we're going to say yes and figure it up.
[294.7s] Right just so that it can start working directly now we have
# oracle/transcripts/chunks/c02.mp3
# model: tiny | lang=en

[0.5s] the other setup here. So this is the upper version working locally. As you can see and then I just want to make sure one thing as well
[8.7s] that we don't have overlap in ports. Yeah, so this is
[12.1s] starting on this port which is 3,200 and this one is running on that port which is 1,200. Now you actually see when you're doing this like it does a very good job of
[24.4s] building
[25.4s] how locally and also emulating the style that we want. So we'll see how this goes.
[34.0s] So it says course, scaffolding is in. I'm wearing this out. Now while we went for those two tasks both the new one and the template version to be done,
[44.5s] something else to consider is the memory setup. Now there's so many different memory options.
[50.2s] For example, if you've never come across it, Hermes agent memory, so many different options for that. You can use loads of different assistant memory. There's lots of different options for this.
[61.2s] The reason that you would have a memory is so that you're not starting from scratch and also that you'd link everything together.
[67.2s] So if we have a look for example at this setup over here. When you're building this out, all of these agents can use the context from our memory system. Why would we want that?
[80.2s] We would want that because for example, one of the reasons you want to have an agent or S is that you don't go in between different apps and have to configure everything separately.
[88.2s] So normally what happens is this is a big problem for most people is like they will go over to chat,
[95.2s] to chat, go over to chat, and then go over to the cloud and these two things they don't think together. They don't work together. So what you want it said is a memory system. How do you build a memory system? You can go into Obsidium Obsidium is my preferred memory system.
[110.2s] I'm sitting in basically stores everything in my town files in a beautifully organized system like this.
[116.2s] Now if you're wondering how to get Obsidium, you can just go to Obsidium and it's a free app that you can download. It's actually open source.
[124.2s] So you can just get Obsidium and then you configure it like that. It won't look as beautiful like this, but you can ask the agents to organize and smart and help you and organize it properly so it looks more efficient.
[138.2s] So step one is building out the dashboard which we've got chatchipity working on over here. Then step number two is that we are building the memory system inside all of this.
[150.2s] So that when we're using cloud and when we're using chatchipity and if we're using hermys they all sink together. They all work together.
[159.2s] So the reason for that is that then the agent is no exactly what you've worked on which means you'll get better responses. You'll get more personal responses and it'll be easier to use each one.
[170.2s] So if we have a look at our memory system over here and we scroll in this is our memory galaxy. So every single dot that you can see is a individual memory and this is all linked together inside one beautiful galaxy.
[184.2s] Where we've got all of our files over here that we can preview and we can have a look at really easy to use.
[191.2s] So when we're doing this it means that we can sink our memory together and we can have every agent basically understanding what the operations have worked on.
[202.2s] And the other thing to note about all of this system is like everything all dot inside here is for example like a person or a project or a business or a tool that I've used and it all gets sinked inside one place.
[215.2s] Now when you're doing this swap you don't have to create all of this yourself you should get your agents to do it. So you can say to for example, Claude, hey Claude, based on everything I've worked on recently just create an obsidian file and organize that and then you can say to chatchipity.
[231.2s] Based on my local obsidian file can you just improve this but that's how it works.
[236.2s] And one thing I'm just going to do here is I'm going to say don't remove the previous system that was set up on the local port just create a new one.
[247.2s] So we still need the agent or else on the old port because what it's done is basically take a mile port out and then replace it.
[254.2s] So make sure the old one on that's famous this local port is back, it's back up ASAP and then you can use it totally different local host.
[266.2s] So that's something to note as well as like when you're running this locally if you're running in other projects like other get her projects in the background.
[273.2s] You don't want to have them conflicting because you can only run one local project on one local year.
[282.2s] So we've talked about memory and we've talked about how to do dashboard now also it's important to note with the memory as well something that'll make it a lot easier.
[291.2s] You can use something like omie and omie can actually record your screen take notes can listen to your microphone.
# oracle/transcripts/chunks/c03.mp3
# model: tiny | lang=en

[0.0s] phone, you can understand what you're working on and then it can create a list of memories
[3.7s] here.
[4.7s] So you can see for example, 21 minutes ago, one hour ago, one hour ago, it understands
[8.8s] everything that will work on recently.
[11.4s] That's really useful because then I don't have to stay at the nose myself and actually
[14.6s] get more detail than any of the other things that I'm working with.
[18.1s] So when we're using OME, which again is another free open source project, if you think
[23.6s] they do have a subscription if you're using it a lot, but if you're just using a basic version
[28.8s] of it, you can get access right here.
[30.8s] It's available on OME, it's MIT license, it blisters for you, understands what you're
[36.9s] working on, takes notes, and then you can actually export that to Obsidian.
[41.7s] So what OME, that goes to Obsidian and we go from there.
[46.0s] So if we go to Home inside OME and then we go to the settings and we check this out, we can then export
[57.8s] this directly to Obsidian as well, which is really, really useful.
[63.1s] So this is being an out now.
[65.3s] You'll see here, this is quite a big project for chat activities to work on.
[72.4s] So when it's being an out from scratch, you can see here that's beginning to do it out, said,
[76.2s] I said, I'm working on FIM, prioritizing the UI and the integrations here.
[85.2s] But it might take a few minutes, it's been coding out for like 10 minutes already, just working
[89.2s] on this brand new H&OS.
[91.0s] Whereas for example, as you saw before, we've already got the H&OS live with the templates, so it's
[97.4s] way faster to use the template than it is to use a brand new set of from scratch.
[103.8s] So if you want to save time, I would just go with the template inside the app, I'll put it in,
[107.3s] but if you want to build your own from scratch, this is how you can do it over here.
[111.0s] Now when we're waiting for those to run, one other thing to note here is that once we're
[116.8s] building this and we're creating stuff out, you only need to add what's relevant.
[122.8s] One feature that I built in and I recommend this for you too, is that you have an option to hide
[127.9s] things, because what we're trying to build here is something where we focus on the system,
[133.1s] not the model.
[134.1s] So for example, OpenCore, go back like two months ago, this was probably the biggest A-A-A-Gene
[140.5s] in the world.
[141.5s] It's only recently that people stop using it, and that includes me because it's Hermes to
[144.7s] go over.
[145.7s] So my point here is that you want to focus on the model, on the system, not the model.
[153.3s] So you can hide OpenCore and you can add it in.
[157.2s] And I would allow that for yourself as well, so that you can move things around, you can change
[161.2s] it, because things in AI are crazy.
[164.0s] Like G556, as soon as it came out, absolutely mind blowing.
[168.2s] But G555, I would never build anything like this with it.
[172.0s] So as you go along here, there's going to be things that you swap out that you don't need.
[177.0s] I mean, if you look at my setup as well, because I'm testing some of the things all the time,
[181.7s] I have loads of tools here, but for you personally, you might want to use Hermes or you might
[186.1s] only use Clawd and you just need to build out several workflows of each one.
[189.8s] So that's something important to note, it's like you don't have to use everything that you
[193.9s] see here.
[194.9s] It doesn't have to be as complex as this.
[196.8s] One thing that I learned along the way was the simpler you can make this the better, because
[203.2s] you won't use every feature.
[204.2s] For example, Kimmy code, I personally haven't used for weeks, so I should probably hide them.
[210.3s] I can always reuse it if Kimmy K-Table and A comes out and then it's absolutely mind blowing.
[216.2s] So you see how you can swap things in, you can swap things out, etc.
[220.1s] You've got the mission control here.
[223.1s] The daily goals list as well.
[225.2s] This is really useful.
[226.2s] I'd recommend building this into.
[227.6s] In any feature that you see here, that you like right, I want to add it in, or for example, let's
[232.0s] say you have an idea and you like I need to add that in, I need to add this in, etc.
[235.8s] Well, when you're building any of this out, you can go back into chatchipathy and you can say,
[240.6s] okay, great.
[241.9s] Now build out a daily journal section or now build out a daily habit tracker.
[248.5s] What are the features you have in your mind?
[251.2s] You can build in.
[253.2s] And you can ask chatchipathy to build in.
[255.4s] And if it's taken ages to reply, you can just say, okay, add a journal feature.
[262.2s] And if you type forward slash steer inside or if you add the message here, it will actually give
[269.1s] you the option to steer it in the right direction if it's in the middle of coding.
[273.7s] So that's how it works.
[274.7s] I said by said, now if you have a look, we've got the mission control template installed.
[281.2s] So this is actually the template from the AAPRF boarder.
[285.4s] If we have a look at the setup from the agenda, go S, the chatchipathy is built.
[290.8s] It looks similar, but this is the version where you're signed from scratch.
[295.6s] Now when you build an inside set by set, actually looks really nice.
[299.6s] When you build an inside set by set, actually looks really nice.
# oracle/transcripts/chunks/c04.mp3
# model: tiny | lang=en

[0.0s] I said by said, you want to test it out.
[2.6s] So let's just test, are you working?
[5.2s] And you'll see things that you want to improve.
[6.6s] You'll see things that you want to make better.
[8.2s] And actually, if you have a look at this, it totally blanks.
[11.0s] So we said, are you working?
[12.9s] And then it just didn't reply.
[14.6s] So when we fix things to the flight, this and this way,
[17.5s] it can take quite a few hours per day
[19.6s] for me to test out and improve this.
[21.7s] We need to drag that screenshot in and say, hey,
[23.9s] this isn't working.
[25.3s] Can you fix it?
[27.4s] And we'll delete that previous response of the gig.
[30.0s] For example, is not replying and also when I click enter,
[34.4s] actually just had a blank screen.
[36.1s] So just make sure you test and actually make sure it works.
[38.6s] So it's much easier to start from a template,
[41.5s] a template, why actually work, whereas, for example, this set up one.
[44.6s] And you see how this is running through each stage of the process,
[48.8s] whereas, for example, with the new version from scratch,
[52.1s] it created a nice UI, but it didn't even test it,
[55.0s] even though we previously asked it to.
[56.4s] And that's on 2.5.6, so extra high.
[59.1s] So because it's such a big project and because it was coding for so long,
[62.2s] you can't have forgotten to do what we asked it to do.
[64.5s] Because we said, make sure it works.
[67.1s] Make sure it actually, you know, actually works,
[70.6s] set by set, but it doesn't, it just created the UI.
[73.0s] So with what the UI, the user interface that's looking good,
[77.0s] we just need to make sure that the rest of it is working to.
[81.0s] So it's going to start improving that as well.
[83.4s] So if we come back to scratch, you'd be fine.
[85.5s] That's beginning to analyze it as well.
[87.8s] Now also the great thing about chatchipathy,
[89.6s] and you can use code for this, you could even use her.
[91.7s] It's for it, although I think if I had a choice out of all these,
[94.3s] I'd use chatchipathy or code at this point,
[97.8s] because it's just much more visual when you build in stuff like this,
[100.6s] you can actually fork the conversation.
[103.4s] So you can click on new over here, and inside the same folder,
[108.4s] you can have agents working in parallel on building out the H&WS.
[112.6s] Now, this is really, really useful.
[114.6s] So for example, when I'm coding out the H&WS,
[117.8s] inside, called what you see on the left hand side
[121.3s] is that I will test and run different implementations side by side at the same time.
[126.4s] What I mean, for example, is that when I was building out the video agent,
[130.4s] I was also working on the H&WS.
[132.2s] And I was having them both being built side by side
[135.0s] inside the H&WS system.
[137.5s] And so the great thing about chatchipathy as well,
[141.3s] is that you can just start a new chat here,
[143.3s] and you can say add inside the H&WS,
[149.0s] an image generator.
[152.6s] And then we'll say use G556 for generating the images
[158.3s] with the OAuth slash CLI setup,
[162.8s] and make sure it actually works, but it's allow me to save anything that I create.
[167.7s] And so what you can see here is that we now have
[171.8s] two conversations at the same time
[174.2s] for our new H&WS that were built in.
[176.6s] So every idea that you have that you think,
[179.6s] wow, this would be valuable, you can build it at the same time.
[182.6s] And this means for example, you can have like six different conversations
[186.4s] inside the same project that would be encoded outside by side.
[189.8s] So we can have, for example, a video agent,
[192.8s] an SEO agent being coded out, the lead generation being coded out,
[196.1s] and we just add in new workflows inside the H&WS system,
[199.6s] as we go along.
[200.9s] So you can see here that it's asking for your questions,
[203.8s] I'm just going to say, yeah, just figure it out and build out, right?
[207.6s] I don't need to give instructions like that to G556,
[213.7s] but this is smart enough to figure out for me,
[215.5s] and if it really is bad then, okay, no problem.
[217.8s] I can go back and fix it later.
[220.2s] The other cool thing about using this is like,
[222.1s] whether you're using, for example, called or using chatGpc,
[225.0s] you can plug in your existing subscriptions into this system.
[229.3s] So if you use Claude Woll, ask it to build,
[232.2s] Claude CLI into the H&WS, and that way you don't need to use API,
[236.4s] which means that you don't blast through loads of tokens,
[238.7s] which means that it doesn't cost you anything extra,
[241.2s] to number one build the H&WS and number two use the H&WS.
[245.7s] Because for example, we have a look inside the system,
[248.9s] we have an agent who goes with Claude with Hermits,
[252.6s] with Vantigravity with Codex.
[254.2s] Now Codex already subscribed to chatGpc before I build this,
[257.2s] and so it didn't cost me anything extra.
[259.0s] Claude already subscribed to it before I build it,
[261.5s] didn't cost me anything extra.
[262.7s] Hermits free open source projects, we plug on the Rue sub-brain into it,
[266.6s] doesn't cost me anything.
[267.8s] So you see how you can build these agente cooperatives systems for free,
[271.1s] that's very important to note as well.
[273.7s] And now this is beginning to code step by step.
[278.4s] It's adding the image generator inside the existing storming home,
[281.4s] this is one.
[282.6s] And by the way, for the actual agent or system that uses the template,
[287.4s] that was all set up and configured here.
[289.3s] So it says the new system has separate configuration
[292.3s] and what I would do if you are installing this template from scratch on the upper floor,
[296.2s] it's just tell it, listen, test everything out, anything that doesn't.
# oracle/transcripts/chunks/c05.mp3
# model: tiny | lang=en

[0.0s] work, tell me and make sure that you authenticate it or if you can't authenticate it,
[5.4s] that you ask me so that I can set out.
[7.8s] So the template set up is pretty much finished at that point.
[10.4s] That's way, way easier.
[11.7s] Whereas for example, be on an agent or a search, you know, you could be here for a few hours
[16.0s] or you know, you might see something that breaks tomorrow and then you have to go and fix it
[19.4s] as well.
[21.6s] Now, when we were talking about the memory system as well, that obsidian system, you can
[26.3s] go back into the agent or else here.
[29.7s] And you could say, also add the memory system from obsidian, so I have obsidian installed
[35.1s] locally and install that into the agent or else as well.
[38.5s] So it's automatically synced and also make sure that the chat activity and hermys are synced
[45.0s] to my memory system.
[46.5s] So when I use them, you update the obsidian as an immatarver.
[51.7s] But also when I ask you questions, you learn from the obsidian vault that we have locally
[57.4s] and you use that as context.
[60.0s] That way everything is personalized inside the H&O S.
[64.0s] So this is the memory system and how you can install that.
[67.6s] And this is pretty cool because now we have the front end plus functionality that's being tested
[75.2s] of it.
[76.2s] We have the custom workflow that we're building out here and we have the memory system that's
[82.8s] being built over here.
[84.4s] So we're adding three different layers as we go along.
[88.2s] We've done that in a space of 27 minutes.
[92.9s] There's going to be some workflows as well over the other ideas for and you're like, there's
[96.2s] no way it could do that.
[97.6s] I would genuinely just ask because most of the time, G556 can handle anything that you've
[103.1s] got out of here.
[104.1s] I've not seen G556 be unable to build this stuff out.
[108.5s] It's pretty unbelievable, everything that you can use here.
[113.1s] So I'm going to say figure this out.
[115.8s] If I had more time if I was here for a few hours or whatever, then what I would actually do is give
[121.4s] it more instructions and be more in-depth and just take a little bit longer to answer these
[126.5s] questions, but I'm trying to save you time and I'm trying to show you in real time how it works
[130.5s] about it being boring, honestly.
[132.6s] So if we have a look over here, we've now set up G556 so we've in-miss generation, we've set up
[138.9s] the front and functionality and we have the dashboard here.
[142.0s] We can also improve it and test out side by side inside chatchubity.
[146.8s] So this is the offering system of built and this is the dashboard that we're built.
[151.8s] So if we go inside the Hermiss agent here, we're just going to say working and make sure Hermiss
[157.5s] agent actually replies to us, boom, and it says yes, I'm working what do you need?
[161.9s] Now this is amazing because now we have Hermiss agent, we have G556 plugged in which is like
[168.0s] one of the best frontier models in the world.
[169.9s] We have chatchubity codex and that's creating some Missist stuff as well.
[175.6s] So if we go to Hermiss agent here, we can use it identically and we can use it inside our H&WS.
[180.8s] Now we've got the inside chatchubity codex and now we have the mission control as well.
[184.7s] If we click on any of these, it takes a straight to the commissation.
[188.0s] So UI looks nice and looks clean and looks simple, which is great as well.
[191.8s] And then over here we have the live activity, so we can see what's being done.
[197.0s] We have the work that moves with recent operations, we have both of the agents that we've built,
[201.4s] and we have everything else ready to go.
[204.6s] Unbelievable how easy it was to set this up.
[207.3s] Now of course we do have the emiss generator that needs to be tested and we'll see how that forms
[211.9s] as well.
[212.9s] And then we've also got the memory system that's being built out as well.
[217.8s] There's a lot of layers do you need to add to this as well.
[220.7s] That might be another question they're asking.
[222.8s] So something that's really, really useful is the conversation history.
[228.4s] Now if I open this up and it just starts from blank every time, not very useful.
[233.0s] But if I can start a new chat, like you can see here, and we can switch between them, that is great.
[238.6s] Because then you're kind of using Hermiss like chatchubity.
[240.8s] So you can go back and forth inside the chat of it, but also you can use it identically because
[246.8s] you've got Hermiss agent, basically have superpowers, you can do computers, you can edit files locally.
[253.0s] And so it's kind of like you've created your own better version of chatchubity, this is chatchubity,
[257.9s] but it can't edit my local files.
[259.5s] Whereas for example if we go over to Hermiss agent, this can work directly here too.
[263.2s] So that's great.
[263.9s] Now one thing that I also add here as well is a system where you can essentially save everything
[272.8s] that you've created.
[273.8s] Now come on to that in a second, but let's have a look at this image generator.
[277.8s] And you see how it would basically build an outer and soft tool.
[281.5s] Think about how amazing that is.
[283.5s] We're at the stage now where you can build your own tools, you don't need any technical knowledge,
[288.8s] you don't need to be really technical or understand HTML or be a coder to build this stuff out.
[295.1s] You can just build your own custom workflows using our automation in a way that is
# oracle/transcripts/chunks/c06.mp3
# model: tiny | lang=en

[0.0s] custom-wise and personalized here. That is unbelievable. So if we go inside here and we're like,
[4.3s] okay, create a image of a cat eating biscuits and we'll delete this bit here, we'll change that.
[11.8s] We'll get rid of that. You see how it's got this aspect ratio as well? So I didn't even ask you to do that,
[17.0s] but code X4 ahead because that asks it to figure out. That's why I quite often prefer just
[21.4s] giving the AI to figure out because it has better ideas and it has more time to give attention to this
[26.7s] than I do. Like if I've been coding this out for hours, I'm going to get tired, I'm going to get
[32.2s] distracted, is going to be difficult to focus at times. You might get frustrated sometimes.
[37.1s] Well, where's if you give the decisions to a frontier model like GPT-5 on 6, it can do it all for
[43.2s] you and handle it as you go along. So now we can say, okay, I want to just go with a cinematic image of a
[49.6s] cat eating biscuits. We've got the prompt, we've got the aspect ratio with cinematic, we'll hit generate
[54.7s] image and that may not work, but so far it's actually looking really good, really impressive actually.
[61.5s] And so one thing that might happen here is if we click from image to video to homeization, I wonder if
[68.0s] it's going to stop? Yeah, see how we click on that and then it's stopped generating. So we need it to be
[73.0s] able to run in the background. That's another great feature. And you'll see the more you test it, the more you'll
[77.4s] see ideas from improve it and more things that will come up that are kind of like bugs by this. And so
[83.1s] what we want to do here is just say, okay, inside the custom workflow image generator and this is exactly
[89.2s] why you want to rename the conversation. So if you right click and rename the conversation, you can see exactly
[93.9s] what you're working on. It keeps makes it easy to keep track. We can go inside here and say, based on the image
[100.2s] you, if I generate an image, can you just make sure it still runs in the background whilst I'm moving around?
[107.1s] And also what we've seen here is that it's actually tested out and it's found some bugs that we didn't
[112.6s] even know about. So it's turned a 500 error and it's uncovered a real bug there which is fantastic
[119.4s] as well because it's still in the bug checking for me. But overall it's looking very, very promising
[125.0s] that's beginning to work. I mean, a bit of we go back to the app here and we're like, okay,
[131.8s] create an image of a cat and then we just hit generate image. That should actually work without us
[137.8s] switching between them. But let's see. So the mission control, we have the different agents built in.
[143.0s] We have the studio of custom workflows as well. But this is beginning to look and form something
[149.8s] that is actually amazing. And you can see here that we've got the agentic memory of all as well.
[157.0s] So this is the system that we've used and as you want to add that as a tab. So I'm going to say,
[163.0s] add the memory system as a tab. So the I can easily see the memory has been added and see what's
[170.0s] been done and also visualized the inside like a beautiful knowledge graph and we can hit into here.
[174.6s] Now it's beginning to run out of work usage as you can see because I've been running on fast mode and we've
[179.8s] created so much stuff started by side. I'm going to say figure this out. But the 8020 of building the
[185.5s] agentic operating system is being done as you can see. And so all of this is working in the background.
[190.5s] You do need to be careful with the token usage as well because you can easily run out of tokens.
[194.6s] If you're wondering how to check your token usage, you can go inside the advanced section here and you can
[200.6s] switch the effort levels and speed and everything else. As you can see how many tokens you've used side by side.
[207.4s] And you can also click on reset usage as well as an option. Now this is beginning to code out each part as well. Nice.
[217.0s] Memory system should be added in and we hit the limit. So there we go. We hit the usage limit already.
[226.2s] We can actually click on reset usage. I'll be quite interested. I know if we can and I'll. Oh yeah we can.
[231.2s] Of course you can do a weekly reset usage limit as well. I didn't know that.
[237.0s] That's pretty good. And then we'll go out weekly usage limit that resets on the 18th which is quite a few days ago.
[242.6s] But yeah and also when you're being out with this. If you want to reduce the amount of tokens
[248.6s] because this can be come up problem especially if you're creating some of the things in parallel using the
[254.4s] GenteColoring system, then you've got a few options. So you can switch your fast mode. You can switch off the
[259.9s] intelligence so you can actually dial down the intelligence level. You can also change and click on advanced
[266.8s] and then you can switch between the models. So if it's just something really basic then you could switch
[272.0s] the model to like 25.5 and that would use less tokens. You can change the effort level to light or medium
[279.0s] and you can change the speed standard. Now when you're building all this out,
[283.4s] the more you reduce the amount of tokens you use by dialing down the effort dialing down the speed,
[288.7s] dialing down the model, the worse the outputs will be. As always that balance with token optimization
[294.6s] creating something that you really like also using the plan.
# oracle/transcripts/chunks/c07.mp3
# model: tiny | lang=en

[0.0s] about running out of credits.
[1.7s] Also, something else that's taken to account here
[4.2s] is that there are loads of different ways
[5.9s] you can reduce the amount of tokens and minimize them
[8.5s] even more.
[9.4s] We've actually got 10 different strategies
[11.1s] inside our token minimization playbook
[13.2s] in the app off of volume.
[14.3s] They can help you save like Nibes,
[15.7s] 95% of your tokens.
[16.9s] So if you want to get almost two to three X,
[19.6s] out of your system and be able to code for more
[23.7s] and use more tokens, then you can use this system here
[26.4s] and it will actually reduce the amount of tokens you use.
[29.2s] And that's also one of the reasons
[30.6s] that we recommend a template to set up
[32.6s] because for example, with the template to set up
[34.5s] as you can see right here, that just installed it,
[36.6s] configure it in the background,
[38.0s] and that was already done in like 10 to 20 minutes.
[40.4s] That is way better than having to build this from scratch yourself.
[44.4s] So it's your choice, but for me personally,
[47.8s] I would prefer to use like a template and set up
[49.9s] just so I can save time because, you know, times is really valuable.
[54.0s] So we have talked about how to build your own agent operating system,
[57.6s] how to add custom workflows,
[59.6s] how to set up the memory inside there,
[62.4s] how to test different works,
[63.9s] how to add new features, and you can see, for example,
[65.9s] we have the image of the cat E in this kit
[68.0s] now as well, so that custom workflow works really well.
[71.3s] One thing that I'd recommend to you all at your time,
[73.7s] so you look at where you spend your time
[75.7s] and then how can the agent OS help you?
[78.4s] So anything that you have an idea for,
[80.5s] you can build it as you see.
[81.9s] So for example, the image studio, we have that idea,
[84.2s] and we build it out.
[85.8s] Now at this point, you might be wondering, okay,
[87.9s] what should you be folks, you know,
[89.4s] what custom workflow should you add to your agent OS?
[92.1s] Well, I would look at where you spend your time
[94.3s] and then work backwards from there.
[96.0s] So, anything that you need to do every single day,
[100.2s] on a schedule that is repeatable,
[103.7s] and doesn't require you to be there, which is basically nine or seven of the stuff,
[108.0s] you should be automating, and you should be building custom workflows.
[111.2s] So let me show you some examples of what we build here.
[113.5s] So inside my agent OS, we have the content pipeline.
[116.5s] The reason that we did that is no one else was in loads of time
[119.3s] on keyword research, and I was using tools that I didn't really like.
[122.6s] So for example, I was using a trusted keyword research.
[125.0s] It wouldn't really give me new data,
[127.4s] and it would give me old suggestions that weren't helpful.
[130.3s] With this system, I can type for example,
[132.5s] agent OS or whatever I'm trying to focus on when it comes to SEO.
[135.4s] It will pull in that data, and it's custom made
[140.2s] to all of my websites and niche them in.
[143.5s] So if we keep scrolling down here, you can see all of the different keywords that we've got.
[148.5s] And so I'd be safe for me about 30 minutes a week.
[151.3s] Then from there, we can click on use topic,
[153.9s] and we can actually create the content and deploy it to our website and one single click.
[158.9s] That is super useful for me.
[160.4s] Now if you're watching this, you might be like, well, I don't do any SEO.
[163.3s] Like, why don't I want that?
[164.7s] And that's exactly why.
[166.0s] You want to go back into GV5.6,
[168.4s] based on what you work on day today, and tell it exactly what you do.
[172.7s] Based on your time audit and ask it to build aint.
[176.4s] So if you spend a lot of time doing research, build aint.
[179.1s] If you spend a lot of time doing lead generation,
[180.8s] add lead generation to all inside that.
[182.8s] That's how it works.
[184.0s] And this just loops around around.
[186.2s] So like, for example, when you start automating all this stuff,
[190.2s] I don't think that you're going to find you at less.
[192.0s] I think that you're going to find you work more productively,
[195.3s] and you achieve way more.
[196.4s] Like, I mean, for me, we can create way more content.
[200.8s] We can recreate more people.
[202.3s] We can grow way faster, because we have all these systems in place.
[206.0s] And so like, for me, I don't see it as like a way to,
[210.4s] three out of my time.
[211.4s] I see it's a way to save time so that I can use that to focus on other parts of the business.
[215.7s] And then just keep cycling through.
[217.6s] Okay, what do you work on?
[218.5s] What do you work on?
[219.2s] What do we need to do next and next and next?
[221.6s] Also, there's some other good really cool stuff that you can build inside here too.
[226.2s] So one thing that we actually did is, we notice,
[229.0s] okay, we have so many different agencies.
[230.8s] We've got a claw to get hermons, we're anti-gravity, we have to do a codex.
[233.8s] Why don't we add those inside a system that we can orchestrate?
[238.4s] And so we actually had a paper clip inside here as well,
[240.9s] so that we can have an organization, a team of AI agents working together.
[245.0s] They can go off and build stuff 24-7.
[247.4s] We can see everything that we've created that is really useful.
[251.2s] And also a lot of people are asking us to build this in.
[253.5s] So for example, in their proper form of community,
[255.8s] when people are asked stuff, we actually build it for them.
[258.0s] Like for example, there was someone was asking this morning about how to use chatchewity in the new app.
[264.5s] So what we did is we recorded like a one-hour course for them,
[268.2s] showing them exactly how to use chatchewity.
[270.8s] And so what I really like to do is get ideas from the community,
[275.8s] figure out the problems they have and then solve those problems with the agent OS.
[280.0s] And I think that's really good for you.
[281.5s] Like you can speed to your team, you can speed to your audience.
[284.4s] You can look at the problems that you have and solve each one of them,
[288.4s] one by one, by building the right systems in.
[291.8s] The same with the group chat here.
[293.0s] So we have a group chat and these are just ideas that were close, you can build in.
[296.6s] Because we have finished building out the experience.
# oracle/transcripts/chunks/c08.mp3
# model: tiny | lang=en

[0.0s] sample project. So this example we actually create a group chat where you can message a room.
[6.4s] And then we can set a keyboard, Hermit, come up with a new game idea to build. It will read
[11.4s] our obsidian vaults of the memory that it tells that earlier. And then all our agents just
[15.5s] chip in and the speed to each other and the bounce ideas off each other. So like for example,
[19.8s] Joe and I are like, oh that's a clever twist. Clawd, let's do that. And then code x as I
[24.5s] know what this sounds really good to integrate over and over and over. Let's add that to our pipeline,
[29.8s] which brings me to the next setup. So we've got a paperclip for agent orchestration. We have
[34.8s] the AI agent group chat. We can also see the history of all our group chats here. Always good to add
[39.5s] history inside the agent or I suppose you won't be able to find stuff in your loop stuff and you're
[43.5s] great. Let's all substitute the confine later. And then finally we have the pipeline. So with the pipeline system,
[49.3s] you can drop an idea in here and you can go from get me idea to going into the human approval
[56.5s] section and just implement it. And this was really good because I found like a lot of people
[60.6s] including myself would have so many amazing ideas but it's very difficult to implement it for him.
[65.4s] So with this system, you can take all your ideas all to master this from the group chat because
[69.8s] the agents add this into the pipeline. And then anything that you like right, we need to build this today.
[74.8s] Okay great, let's just build it. It will come up with a plan. It will plan it out. Then we can
[79.8s] approve it. Then it will build it and then we can view what was actually built previously. And we actually
[85.0s] have a full gallery where we can see all of our ideas. And so we can go from idea to implementations so quickly.
[92.2s] Which is really really cool. So for like okay we need to build out this website, boom shout,
[96.0s] like we can open up that website here later. And again I think it's so important to have like a gallery
[101.4s] and a workspace and stuff that you can see and say for later. Because if you don't,
[106.2s] the biggest issue is that you will not be able to find it. It would just get lots of new downloads.
[111.8s] And that will go with the other 100 projects you're working on and then you won't be able to find it.
[115.7s] So that's why I'm going to a gallery and a pipeline is so so useful. Now what we also have over here is we have,
[122.3s] for example, Claude. And anytime I'm like Claude actually brings something useful out,
[126.9s] we add that feature in as well as a separate tab. I'd recommend that for you too. So for example,
[132.4s] if you saw the ultra-code update, this was amazing. You can have a 14 of agents working in parallel,
[139.3s] working together towards one task and you don't have to orchestrate it, Claude is your extraator
[143.9s] and it builds the subagents for you. So with this system we just added that as a separate tab. And it's like
[149.4s] you create your own custom version of Claude that you can edit, you can change your UI of any time you want,
[154.2s] you can come back to at any point. Then we have Hermes. So Hermes again, really, really powerful.
[162.0s] One of the things that I noticed is like they bring out new, there's always new models that we can plug into
[166.2s] Hermes and this way you want to focus on the system, not the model. So for example,
[171.4s] G556 comes out, no problem. We're going to build that out. It doesn't change how I work.
[176.5s] Doesn't make me have to rebuild anything. It's just like we can add the model inside the system.
[181.4s] The same, for example, a lot of people asking about how to have a voice activated AI agent that we
[185.8s] can talk to. And so we actually build that into this system as you can see. So what it can do is you can listen to me,
[192.2s] you can talk to me, it can build stuff out for me. And then again, we've got the full gallery here of
[197.9s] everything we've created. But we have Hermes all recall. For me, I was just find way too much time was being taken,
[205.0s] research in the latest news. And so I created this. I also think a great tip for you here if you're
[209.5s] building out your own agent or S is that you make sure it's super visual and fun to use. So for example,
[216.0s] we just added this law icon here. I think it's just a little touch. It's a little UI touch, but make it fun and easy to
[223.8s] use and very memorable. So when I see this in my archive, I know exactly what this does because I associate
[230.2s] the image with the workflow here. And if you don't like the UI of some like that, you can always customize
[235.6s] it for later too. So we have an image studio here too, but we actually have a video on a voice session as
[242.0s] well. You'll not see your previous sessions or what's been done across all of your AI agents. Which is
[246.5s] great when you see the status, we can see the skills installed into a Hermes agent, the campband, the plugins,
[252.4s] the doctor of the insights. These are all just for real cool ideas. Again, as you make it more complex,
[258.4s] more things can break. And also if you're using APIs in any part of that, just be careful because
[264.4s] you know, no more than you don't want to rinse the tokens if you're using the API XT in get expensive.
[268.5s] And number two, sometimes the API is for break or let expire and you have to be careful about too.
[274.2s] What I also like to do is let separate the outputs by the system depending on the model. And one of
[280.9s] me and by that for example is we have Hermes here. What I recommend to you is that you build out workspace
[285.9s] where you can track all your creations by the model that's created it. So when we were testing out
[291.8s] Grok 4.5 earlier this week, we have all of the creations from Grok 4.5 saved inside the workspace.
[299.4s] Really is.
# oracle/transcripts/chunks/c09.mp3
# model: tiny | lang=en

[0.1s] We have the MCP section here, we have the actual Hermes dashboard embedded inside the system.
[5.7s] Another really good tip as well is that if you see anything open source,
[10.5s] you think, wow that would be great to build in.
[14.3s] For example, if we have a look through this list, we see what we've got here like on the room.
[19.6s] This is a free, this is a free way, a free API that you can use to build and automate anything with.
[27.3s] Now why is that important for you?
[29.3s] Well, let's say for example, using the H&S and you want to reduce your amount of tokens to use.
[33.3s] Well, you could take a project like this, you can grab the GitHub link, you can go back into chatchipity here,
[39.3s] and you can see it okay, based on this, build this into the H&OS system.
[45.3s] So, I can code for free.
[49.3s] And then that's going to take the details from OmnLyRoot, install it into your H&OS,
[55.3s] and all of the sudden, you've not just got chatchipity in Hermes in there,
[59.3s] you've now got a system with a free code.
[61.3s] And if you're wondering how does that look like in reality?
[64.3s] Here's an example.
[65.3s] So, this is a custom tool we made with OmnLyRoot, where we can code out for free,
[70.3s] we have the chart over here, we have the preview over here, we have the workspace on there.
[74.3s] Really really powerful stuff.
[76.3s] We have for example H&Y3 code, and also some of this stuff will expire.
[79.3s] So for example, H&Y3 is only available for free until the 21st.
[83.3s] And so, no problem, as soon as the 21st hits, we can delete this whole section,
[87.3s] or we can hide it, and we can do something else.
[90.3s] The same for example is to kind of figure.
[93.3s] Like, to kind of figure is an API, like it's not that nice to use,
[97.3s] and it's kind of like you have to find a place to store it, and be able to come back to it.
[101.3s] So, what do we do?
[102.3s] We build out our own custom tool, we have to kind of figure,
[105.3s] where we can see everything that's created, we can chat with it,
[108.3s] we can ask it to do stuff, and everything is stored inside one system right here.
[113.3s] And then anytime something is trending or interesting,
[116.3s] I want to play with, for example, I loop, loop engineer was something big,
[119.3s] that was, you know, everyone was talking about about two weeks ago,
[123.3s] we've loaded in the loop engineering system here.
[126.3s] And then the other thing I would say here is,
[128.3s] when you're building out agent of rent systems,
[130.3s] it doesn't just have to be for business stuff.
[133.3s] So, for example, you could have a system like this,
[136.3s] where you can generate music.
[138.3s] It could be, for example, you goals.
[140.3s] If you were really into more time, you know,
[142.3s] you can have something related to that.
[144.3s] It could be like a track of a training, it's simple like that.
[146.3s] Whatever you prefer to do, you can plug it inside the system.
[150.3s] So, you could, we have a music generator here,
[152.3s] if I want to just create some music,
[154.3s] I'm interested in for particularly for like,
[156.3s] ambium music, then I can just plug this inside the system,
[159.3s] we can get files, we can come back to the AI music we've generated later,
[163.3s] and it's already to go inside there.
[165.3s] Inside there is one.
[167.3s] So, those are all the systems that we get.
[169.3s] Another orchestration system, as well,
[171.3s] and this is with Hermes, is the Kambam board.
[174.3s] Now, with the Kambam board, what we have is the system
[177.3s] where we can drop in a task, it gets automatically triage,
[180.3s] and then all of that agents,
[182.3s] and we've got many different agent profiles,
[184.3s] get signed to that task as well.
[186.3s] So, yeah, it's unbelievable,
[188.3s] it's really changed the way that I work,
[192.3s] and the other thing I would say here is it changes all your team works as well.
[195.3s] So, if you have a team,
[197.3s] like, for me I have an SEO agency,
[200.3s] you know, we've got a team who needs to be able to use AI and that sort of thing,
[204.3s] we can give this to like our team,
[206.3s] that it is, and they can use the video agent.
[208.3s] And if we have a look down here as well,
[210.3s] these are all reviews and testimonials from the Air for Boarding,
[213.3s] you can see here, for example, like, so many people,
[215.3s] again, awesome wins and awesome results with the agent or S like Jacob,
[219.3s] he said it up in about 10 minutes and you saw that I said it up today as well.
[222.3s] And Jacob was spending like 12 hours a day trying to build out the agent OS,
[227.3s] then he just did it in 10 minutes by using our template.
[230.3s] David was saying, like,
[232.3s] Julian is cooking hard on his agent OS and anybody else seen.
[235.3s] So, it's a great way to customize it.
[237.3s] And you can see, for example, like,
[239.3s] sheena as well, sheena actually created her own version of our agent OS,
[244.3s] and used it for lead generation.
[247.3s] Now, one thing that I guess asked all the time by people was like,
[252.3s] you know, what, if you're, for example, thinking about selling AI agent services or AI automation,
[258.3s] what's the one thing the clients want to see?
[261.3s] What is in demand right now?
[263.3s] I'm for sure at the Asian Opera Insisting, like,
[266.3s] if you can build out custom workflows that solve all the problems your customer has,
[271.3s] by creating amazing systems like you've seen today,
[274.3s] so if they need help generate images, you create the image studio.
[277.3s] If they need help generate videos, you create the videos you,
[280.3s] if they need help generate social media content, you create a social media agent,
[283.3s] and how'd you do it?
[285.3s] You just go back into codex and you test it out for later.
[288.3s] So, you can easily build out all this stuff and you just grab their idea,
[293.3s] you solve all the problems you go from there.
[296.3s] Now, as well, one thing that I'll say,
# oracle/transcripts/chunks/c10.mp3
# model: tiny | lang=en

[0.0s] He was like, you might be tempted to have Claude and Codex work on the same thing at the same time.
[7.0s] I personally wouldn't recommend that.
[9.5s] So if you are building out with an agent and you're trying to work on the agent,
[13.8s] as you're trying to improve it day by day, trying to make it better as a system,
[18.0s] the thing that I'd recommend here is that you just use one agent.
[22.0s] So for me, for example, I'm using Claude right now and I use Faber 5 to help me,
[26.0s] you might prefer using chatchipy to you might prefer using Hermits, but seek to one.
[30.0s] Because if you have multiple agents, it multiple different apps, number one that gets really messy when you're building out an agent offer in system.
[38.0s] And number two, they could like overwrite or delete something that the other agent is building.
[45.0s] So for example, Claude is building out something that's just broken.
[48.0s] Chatchipy might be like, let's remove that from the code.
[51.0s] And then all of a sudden you've got two agents kind of interrupts any chance work and that can get super messy.
[56.0s] Wouldn't recommend that.
[58.0s] One thing that I haven't tested, that would be very interesting in testing mode, is with Codex is very easy to add plugins.
[66.0s] Now, what would be very interesting here is if you could add those plugins inside the agent or S as well.
[73.0s] So that's an example that we've got email and Gmail plugged into Codex, right, into Chatchipyty work.
[80.0s] So you can send the email address, you can manage your email.
[83.0s] So you can draft replies for us and if it else.
[85.0s] Now, what would be very interesting is say to Codex, a build out this scheduled task into my agent or S, so that we have the email agent inside there.
[99.0s] I'm not going to do that right now, because I think it's a task that we'll probably take about 20 or 30 minutes to really perfect.
[104.0s] But that would be interesting to take for the plugin section from Codex, and start adding that into your agent or S, so that you can automate and build anything as you want right there.
[115.0s] Other things that I would recommend.
[118.0s] Number two is you might be tempted to put this on a website, for example, like a lot of our websites we use with NetLefite.
[127.0s] I wouldn't do that just because again, like I think you need to keep it secure in everything else.
[132.0s] Also, if you want to reduce the amount of tokens, you can also use stuff like Free APIs and you can use local models.
[140.0s] So if we go to the local section here, we can use local models from Olammer or LM Studio, and this is particularly good for like the tasks that a quite time consuming, but don't need frontier levels of effort.
[155.0s] So let's say for example, we're looking at all of the models on Olammer.
[161.0s] I'm like, okay, or NetLef, or NetLef looks awesome, right?
[165.0s] Actually, it's a really good tool. I've test it out.
[167.0s] Persona, we have the local lead board on Goldiebench.
[171.0s] We've all the local models, and you can see what we've created in T.R. web, Formula Finals.
[175.0s] And so you might be tempted to build that into your agent or S as well.
[179.0s] And I think it's actually pretty interesting to build out a local code.
[183.0s] And so what we have over here is we have this local section, and we have the preview of F and created what the works face of our previous generations, and we can use local models in SOT System 2.
[195.0s] And that's another way to use it for free, because then you are a combination of like Omniru, which is free way to use like 90 models, and it kind of orchestrates across them.
[204.0s] You have HY3, which is another free API, you can plug into the system, and then you have local models, which are basically free forever once you've installed them.
[212.0s] So between those two things, those free things, you can really reduce the amount of tokens the years as well.
[219.0s] What else can we do here? The other thing that I like to do is when I'm testing out all of the models, with the same agent, the builds out the HNOS, with all the creations I just put them on Goldiebench so that I can document the process here.
[234.0s] There's another good way to generate leads, because people are searching for interesting stuff, people aren't in no what you've built with it.
[239.0s] People are interested in your agent or SOT if you actually share what you're working on and everything else, especially if you're working on some new unique stuff.
[246.0s] So you can actually give your agent or S access to NetFies well, and then it can document the whole process, the whole building of the agent or S on what you've done.
[256.0s] Kind of like a journal, but it's not you update, and it's not you writing this agent, which is really cool to us.
[262.0s] So that's basically it. That is the whole system. How to build an agent or for an assistant?
[267.0s] How to customize it as much as you want? How to add a memory inside there? Which agent do you use to build it?
[272.0s] How to reduce the amount of tokens you've won. Also, how to create a system that actually works rather than just breaks?
[280.0s] Also, how to use it in a simplified way using our templates. Now if you want to get my full agent or for an assistant, you can get that inside there,
[287.0s] or for boarding inside the Aginal S system section right here. So we update the stately and get the set file for it and you can get the last, you see the last update date here.
[295.0s] You can get a full guide and how to use it, and then every time we build something new.
# oracle/transcripts/chunks/c11.mp3
# model: tiny | lang=en

[0.0s] is full. So for example, like we create a chat GPT 5.6.1 out course, you can get that over here as well.
[6.0s] And then also if you want more help and support on building out with the H&O S, you can post inside the community
[12.0s] and we reply to you and create video tutorial for you ever single day. So for example, like Matt was asking about Ome and the best way to use it and what you can use instead,
[20.0s] we solve that problem for him and we reply to everyone and just help each other as much as you can in the best way that it can as well.
[27.0s] Inside the classroom you get access to all of our best trainings. If you're a complete beginner, you can go from begins to expert inside this section.
[34.0s] And inside the calendar, you can jump off weekly coaching course. Now this is super useful because you can share your screen.
[40.0s] You can ask questions you can get help and support him. And you can also meet other community members who are building out this sort of stuff as well.
[48.0s] Inside the map is well, you can meet people locally who are building out with AI agents like you and the soul available inside the AI Prof.
[56.0s] or link in the comments description or just go to the AI Profible.com.
[60.0s] One other thing that has been really useful when we're building out all this stuff is that when a model, like for example, doesn't work the same way or you can't use it anymore.
[72.0s] So for example, if a great example now, because you've got a system that improves everything with day, it doesn't really matter about that sort of stuff.
[80.0s] You don't need to worry about that. So if GVT5.6 comes out, that's great. We can add in.
[85.0s] But if, for example, Fable5, I've got taken down, which you did a few weeks ago, it doesn't matter because in like two minutes you can go back into chatchipity and you can say,
[96.0s] you can make remove the integration with Fable5, add this in, etc.
[101.0s] And also one thing that's really useful is because you've named the conversations, you can come back in like two weeks and edit that conversation.
[109.0s] So for example, what I like to do is I applaud if I'm scrolling down. If I need to improve the setup that we have with Pony's now, which is another AI project, I can go back inside the chat here.
[120.0s] And I can say, hey, add this, change that. And because everything is named, it's super easy to organize and it's easy to see everyone as well.
[128.0s] So my very interesting thing is that we've set up two in the past is like computers with the agent OS. So I can go inside her, for example.
[136.0s] And I can say, hey, Hermes, can you just open up Google for me and that sort of thing?
[142.0s] You can actually do that with Apollo, which is the voice agent and that actually has access to computers as well. So it can edit local files, it can edit.
[150.0s] You can open stuff inside your browser and everything else as well. Which is really, really good.
[155.0s] And that's pretty much it when it comes to the whole system. I mean, it's so much fun to build. It's so inspiring to build. It's also great to just build something that feels like your own.
[166.0s] It's never been more fun to use than I think this is only been possible this year. Like, if you went back last year, it was great using AI, but you couldn't build in the same way, but models would just not add this level.
[179.0s] Whereas when you're building the agent offerings, it's now any custom workflow that you have, things that you've never even dreamed of creating before, you can just add inside this, you know, you can plug in like, right, when I just want to delete that, no problem at all.
[194.0s] So it's so much fun, so much fun. So more things that you can do as well is you can get skills and then you can plug them into the system.
[203.0s] So for example, like there's a skill for reducing tokens, which is called Caveman, and with Caveman, it basically reduces the amount of output tokens you plug in.
[212.0s] And you could add that inside the system as well. The final thing that would say, and this is a little feature is that you can delete stuff, give yourself the ability to delete stuff.
[221.0s] So you can go into chat, cheapety, and you just say, okay, make sure that everything that we create inside the agent offerings system, we can delete so that I can remove anything that doesn't work.
[230.0s] Because like, sometimes you might generate something with a gelinfoam to, but then it actually doesn't work very well.
[236.0s] And so how do you get round that? Well, you can just make sure that you have the ability to delete that and just add a delete icon for later.
[244.0s] So that is basically it. That is a full one hour course in how to build agent offerings system, how to automate anything, how to build your own custom workflows. If you want my full setup, you can get that inside the air, pop water and you can get the amazing community that's all using this sort of stuff.
[259.0s] You can connect me personally, you can direct messry anytime you need anything. And thanks so much for watching. I'll see you in the next one. Cheers, Baba.