So right here you can see that I’ve got Jeff Bezos, Paul Graham, Alex Rosie, Gary Vee, Dan Martell, Gino Wickman, and Warren Buffett all discussing a business question that I had. You basically shoot off a question with different options and all of these different council members will kind of come together, roast your idea, and help you think about it in a different way because each one of these seats has one specific question that they’re focusing on. And now you can see what’s happening is they’re actually debating with each other. So Jeff Bezos was debating with Gary Vee. Same thing with Paul Graham and Gary Vee. Now I’ve got Alex Herozi going after Gary Vee too. And wow, everyone’s deciding to pick on Gary Vee in this roast room. This is crazy. There we go. Warren Buffett goes after Hermoszi. But the point is, like I said, everyone has their own question they focus on. So Warren Buffett’s question is, what does it do to cash risk and long-term value? Alex Herozy’s question is, will people pay? How much? And how do we close? And Gino Wickman’s question is, can the business run this without the founder holding it all together? Now, after all three of those rounds happen, we can see that we have a final verdict, which was option C. And then we can also download a PDF, which allows us to read this, and we can sort of look at this more in depth. And we could also share this with our team if we wanted to. Now, this is really just a visual layer on top of what actually goes on in Cloud Code, which is basically just running a bunch of sub aents based on these different council members that I have created. So, you don’t need this local host open in order to run and use this, but it’s just a nice visualization of what’s going on. And this is completely customizable, meaning you can switch out different council members and you can build other ones, too. In fact, inside of this repo that I’m going to give you guys, there already are a few other ones, too. Hey Claude, what are the other council members inside of our roast room? And so that’s Glido actually shooting off a prompt right now to Claude Code, as you can see. And then it’ll show us who the other council members are. Cloud code comes back and says there are 11 seats. We have the default seven that you’re seeing here. And then we also have Naval, Cody Sanchez, Sarah Blakeley, and Leila Hermosi. Of course, you can build as many other ones as you want. So, that’s a new feature of GLO where we’re able to actually orchestrate different chats between different coding agents. If you want to try it out, you can get a free month using the link in the description. But anyways, just wanted to show this off to you guys and let me break down how this actually works. So, this is a repo called the roast room as you guys can see. And basically, we have these seven different advisors as well as four extras. And I’m probably going to keep adding more in the mix as well. But what I wanted to show you is the advisers. What are these council members? What do they actually live? They’re actually just a bunch of sub aents. So if I go into the claude, actually let me zoom in a little bit more. You can see that we have agents and in here we have one markdown file for each of these sub aents. So Bezos, Blakeley, Buffett, Gary Vee, Graham and all the way through. And if we open one up so you can see what it is, it’s literally just a markdown file that holds information about the sub aent. So for Jeff Bezos, for example, how he thinks, what are the sources that he references, how he makes decisions, all that sort of stuff. So that when I actually shoot off a prompt to the roast room, it’s going to go through the different council members and it sort of adopts their personality of how they think and how they make decisions, what they, you know, and it’s able to pull from things like their actual conversations, their lectures, their books, their principles, things that they put out there publicly, articles, stuff like that. So what you’ll notice is they will all have their own agent file, which isn’t really that long. It’s some things like how you debate, how you talk, your lens, but the really interesting stuff here is the brain right here. So we can see all the sub aents, but if I actually go into the skills in this repo, we can see that there is a roast room folder that has brains. And once again, there’s a folder for every single council member that you want to spin up. So if we, for example, open up Leila Hermosy’s brain, we can see that she has hot books, frameworks, positions, principles, sources, stories, and voice. And it’s pretty much the same thing that all of these other council members have as well, but they all obviously look different. The content is different. So if we open up Ila’s hot, what happens is we can see her pillar is can your team and your leadership carry this and who has to change. It shows us what this was built from. So for example, Laya was built from 48 sources, about half a million words from 2021 to 2026. 14 interviews where she’s the main guest and 34 videos of her own on leadership and things like that. Every person here has eight rules. They decide by frameworks, decision loops, things like that. Obviously, any books that they’ve written, frameworks that they talk about frequently and that they quote, their positions on certain things, pricing, discounting, raising money, debt, their principles, different sources that it actually pulled from, specific stories that they might talk about in videos, interviews, things like that. And every single one of these brains for the individual people is built the exact same way. If we open up Dan Martell, we can obviously see that he’s got books in here as well. We can see that he’s got frameworks that he’s using all the time. You can see the buyback principle, the buyback rate, time and energy audit, the drip matrix. All of this stuff should sound familiar to you guys if you have followed, you know, Alexi or Dan Martell, value equation, grand slam offer, all of these things that they actually talk about all the time. A quick break to tell you about the sponsor of today’s video, Fish Audio. Now, anyone who’s tried AI voiceovers knows that the words come out right, but the delivery doesn’t always because it might read a joke or a warning and an apology in the exact same tone of voice. So, Fish Audio has a new model that is their most emotive yet, and it’s called drama 3. You direct it the way that you would direct an actual voice actor, which means that you can write the direction in plain English inside the text, like slowly realizing something’s wrong, and everything after that tag will follow that direction. So, one line can start calm and it can end in a panic. Thanks for calling. Your order’s on its way. Wait, that’s not the right address. Who approved this workflow? And the tags work in other languages, too. It also generates full conversations. You can give each speaker a voice. You can mark who’s talking. And every character keeps their own voice across the entire scene. Just run it one more time. Of course, that’ll be attempt 47. So, if you’re putting it on video, the streaming endpoint sends wordle timestamps back with the audio, so your captions line up exactly with the speech. Now, on the build side, you set the model to drama 3 in the fish audio API. They also have an MCP server too, so you can connect it to cloud code or codecs. And they also have an official NADN node. So if you’re building something that talks like a voice agent or a game character or just narrating your videos, then try drama 3 on the fish audio API with the link in the description. Huge thanks to Fish Audio for sponsoring this part of the video. Now, let’s get back to it. And then I added this new skill called New Roaster, which basically just lets you build new roers. So for example, if I came in here and I was like, “Hey, new roster.” and we wanted to do, you know, Elon Musk and we shot that off. This will basically create Elon Musk’s brain and his agent file and all of his sources. It will build him out and also build him a little cartoon character that will show up in the actual roast room itself exactly the same way that all of these other Roast Room council members have been built out. So, you can keep scaling this thing up. And obviously, I’m going to let this thing build out Elon and then I’m going to add that into the repo so you guys can all use Elon as well. So, while this is building out Elon, as you can see, let us do a new session. So, open up the roast room. That’s pretty much all we have to do is ask for it. And I’m just going to switch this back to Opus here. And it’s going to open up the roast room. We’ll open it up in the local browser inside of the desktop app. And then we will shoot off a prompt so you guys can see how this actually works. Now, once again, you don’t need this visual interface opened up. It will obviously run the sub aent roast room just the same, but it’s nice to see. But one thing that I will say about actually spinning up a roast room is typically you want to put in a decision. You want to put in options. So two or three options. You want to give it context so that they can actually reason through it and make a better call. Right? So let me just shoot off a quick example. We are looking at hosting another in-person event in the summer of 2027 and we’re trying to figure out where to do it. We’ve been, you know, analyzing different options like San Francisco, Salt Lake City, Austin, Chicago. We obviously want to be able to get good sponsors there, local incentives, get a lot of people there, and you know, so airports should be important. Um, public transportation, good restaurants, good activities to do outside of just having the event, things like that. So, help me figure out where this should be and do this on the roast room council. So this natural language request should be just fine. But you also couldn’t come in here and do a slash roast room. And then you could just invoke this like that. But we will go ahead and just get rid of that for now and shoot that off and it should be good enough. One thing that this will do is it will help you figure out what would be the best council members to have for this specific debate. As you can see, it’s going to load the skill. It’s going to gather event and business context. So it will ask you questions if it doesn’t feel like it yet has enough. And then it will actually go ahead and run the debate for you. So, you can see it is going to ask me a couple questions here. It has four questions that it’s asking me. And by the way, we already did book this event in Salt Lake City, June 16 through 20. So, if you guys want to attend, then definitely check it out. But for the sake of the example, I’m just going to pretend that none of this has happened yet. So, I’m just going to say not signed, open choice. It says, “What size and budget should the council plan around?” And I accidentally just skipped through those questions, but the last one says, “Who sits at the table?” And it gave me a recommendation. And it also lets you play with it if you wanted to put different people in. So that’s part of it is thinking about who should actually sit on the council for your specific debate. But now that’s going to be it. It’s basically going to run. It’s going to process this. Even if this was closed, like I said, we would have sub agents that would be going through and doing this no matter what. And the reason I decided to do this with sub agents, because when I actually started doing these roasts, I started doing this with agent teams. But the reason I’m doing sub agents now is because I found after doing lots of testing that agent teams was more expensive for a minimal amount of actual value ad. So doing sub agents works pretty well because they start to actually talk to each other and stuff like that either way. So it works just fine. Now in this case you can see that we took out Dan Martell because his is more about like protecting the founders’s time and we put in Naval and his question is does this build leverage and equity you own or does it just rent out your time? So you can see it was great that we brought in Nal for this specific type of roast room. And obviously last time you guys saw the PDF that it creates which is pretty high level but all of this stuff is stored locally and you’re able to dig in. You can ask more questions about it. You can inspect all that stuff if you want to look at the debate more deeply. So it’s not lost data of course. Show me where all of the debates are held later if I want to be able to dig in and ask more questions about it. So as they’re going through round one and writing their opening positions, I can just show you real quick. If I go to the files and I go down here to projects and then we scroll down to a project called founder council right here you can see that there is different sessions in here and these sessions will show us different roast rooms that we’ve run since creating this visual interface and we’re able to actually then dig in. So if I look at this one, we can see that there was a brief, there was a round one, round two, and a verdict. And this will hold all of the session logs from each of these people because as they start to write their positions, we’ll see that all of their reasoning comes through even when they start to debate. And we can view it right in this interface as well. So now you can see they’re starting to come through with their opening positions and we’ll see below their name what they voted for. So Gary came in with voting for A and gave it an 8 out of 10 score. Gino did the same. And you can see A is Salt Lake City and B San Francisco, Austin, Chicago. This one, Alex Heroszi, also voted for Salt Lake City. So, everyone voted for Salt Lake City except for it looks like Naval went for B, San Francisco. And then, like I said, we could click in and we could look at exactly why. We could see the risk. We could see what they would do. And then we can keep seeing what everyone’s going to do. And now, what’s going to happen is they’re going to read each other’s positions. And then they’ll start to go after each other and try to poke holes in each other’s opinions. In this case, they all are in pretty much good agreement, which is nice. But if they did all completely agree, then the chair or sort of like Claude code, the facilitator of this debate, would make one of them argue for a different side. So they kind of play devil’s advocate and we get to see more perspective as well. But in this case, we didn’t have a unanimous decision. So the debate will continue as expected. Now, what I think is really fun about this is we’re able to leverage the subject matter expertise of other people. And humans are really good at their domains, right? So like for example, I could not build a world-class copywriting agent. It would be tough for me because I don’t have a copywriting background. But what I could do is go find someone who I think is the best copywriter in the world. I could extract their frameworks, examples of their copy, how they think, how they decide what to write, and then I could put them into this council. And when I have questions about copy or, you know, how to market something, I could leverage them inside of this roast room. And it doesn’t mean it’s going to be perfect, right? It doesn’t mean that whatever this roast room decides I blindly take and accept, but just getting in this mindset of thinking about who has done something that you want to do, who has done something and proven it. Now, how can you go out there and study what they did and how they got there and basically just try to follow that path? I think it’s a lot better to do that than just trusting Claude’s gut or trusting your gut sometimes. Like following someone that’s actually walked the path you want to walk is a pretty good signal, right? So, doing these councils helps me just look at things through different perspectives and, you know, think about things that I might not have thought about because Paul Graham brought up this idea or because Naval brought up that idea. And so even outside of the bounds of this roast room for example, I think that just thinking like that when you’re building out other skills or you’re trying to automate different processes, I think thinking like that is a really good way to think. But anyways, you can see they’re actually coming through now. They’re deciding on final votes. And you can see that we’re probably going to win, you know, Salt Lake City. It’s 7.9 out of 10 right now. And that is probably the verdict that we’re going to get. But of course, we could go through and we could read everyone’s key points. As you can see, here is Hermosy’s key point. Here’s Wickman’s key point. You get to see here Bezos conceded to Gary Vee and Graham about them being right on certain things but also disagreed the most with Naval. So anyways, that’s how you can sort of start to read through some things in more depth. We can see that we have the PDF now which I’ll go ahead and download real quick. And now we’re able to look at the fact that Salt Lake City won which is actually pretty cool because that is where we decided to have this first inerson event. Now anyways that is done. We could obviously do more rounds if we wanted to. We could bring in other council members. And now going back to check on Elon, you can see that it created the actual portrait for him and it created these different views so that we can see what it will look like when he’s actually talking. And then if I go to the repo, we can see that we now have Elon added in here. If I go to the brain or not the brain, the agents, we can see right here we have Musk. And then of course in the actual brain because it’s supposed to do that through the new roster skill, we have a brain for Musk. Once again, it has hot books, frameworks, positions, principles, sources, stories, and a voice.md. So, we are now set up and we are able to use Elon Musk inside of this roam council. If you guys want to grab this GitHub repo and play around with it, then just go to my free school community. This will be linked in the description. Go to the classroom and then grab it and from the AI skills or from the all YouTube resources. All of my other free GitHub repos, skills, prompts, guides, everything is in the free school community. So, anyways guys, that is going to do it for this one. I hope that you enjoyed you learned something new. And if you did, please give it a like. It helps me out a ton. And as always, I appreciate you guys making it to the end of the video. And I’ll see you on the next one.