Building apps LIVE with my 24/7 AI employee Clawdbot

Today, we are going through what might just be, in my opinion, and this isn't even hyperbole. A lot of people get angry at me for the hyperbole in these videos. What I'm about to say isn't hyperbole. I truly mean what I'm about to say. I truly mean this.

Today, we might be going through what I believe is the coolest technology of all time. I mean this. What we're going to go through on today's live stream, I think, is going to change the entire world. I, I think it's cooler than Claude code. I'm just going to go and say it. I think it's cooler than Claude code. I think it's cooler than what I believe is maybe any technology ever. I think what I'm about to show you is the greatest application of AI ever. I mean that. I think this might, I think this is the greatest application of AI ever.

So, right now, I am streaming from you on a new computer I just bought yesterday. It's a $500 Mac Mini. And on this computer, I'm gonna pull it open right now. On this computer is a little AI bot. A little AI bot. In this bot that I have on this computer, it works 24/7, 365, just constantly working on this computer. Constantly working on this computer, and it can do literally anything on this computer. Anything a human being can do on the computer, it can do. And on top of that, it interfaces with me. This isn't just a chatbot like you go to chatgbt.com and you chat with it on chatgbt.com. No, I text it. We text message me in this bot. We text message. It texts me out of the blue sometimes. Sometimes out of the blue, it texts me.

So, what is that bot doing? And I'm going to pull this open in a second. Again, this is what they, they in the business call showmanship, right? This is what they call, this is why we're the number one vibe coding channel on YouTube. We teach you little things, but we also keep it showy. So, what does it do? I'm also setting up on the background because I never prepare for anything I do. What does this bot do for me? What does this bot do for me? By the way, I didn't sleep last night. I got this yesterday. Used it all day. Used it all night. I haven't slept. I'm on like 20 coffees. We're going non-stop right now.

What does this bot do for me that's on this Mac Mini? It controls this Mac Mini. It does anything a human does. It texts me on Telegram. What does it do? That's a good question. I'm plugging my phone in right now because it's basically dead because I haven't been paying attention to anything else in my life for the last 24 hours. Here we go.

Well, it does many things. I think what one of the coolest things it does at the moment is it just keeps an eye on my GitHub repos. So, I have all my projects on GitHub. I have Creator Buddy, which is my main SAS that pays all the bills. And it literally monitors my GitHub repos, looks at all my side projects. Here we go. You want to see it? I'll show it to you. It just monitors my GitHub repos and is like, "Oh, I thought of an idea. I'm going to build out this idea in your in Creator Buddy." Working on Creator Buddy in the background. It's building me an article writing feature. It's adding functionality to my article writing feature. Say, "Hey, you want me to do this?" I say, "Yes, please." It just starts building it. And it goes, "Done. Built it. Here's your PR. Review it. Let me know what you think." I hope I'm not showing anything that gets me hacked. Anyway, it does that. It's just creating pull requests for me. It's looking at all my apps on GitHub and just creating pull requests. It's coming up with ideas and creating pull requests, and it can control anything on this computer. So, like I have like a to-do list on my computer. I use Things 3. I can just text it out of the blue. I use Telegram to text it and just say, "Hey, what's on my to-do list today?" And it'll text me back, "Oh, here's what you got to do. You want me to take care of any of this for you? I could do this for you if you want." When it finished this pull request, it texted me. It's like, "Yo, dude. I just built you a feature. If you want to check it out, let me know what you think. I can push it."

So, if you got 30 gits, it's going to look at it all. Yes. So, when I set it up, it goes through an onboarding process. I said, "One thing I'd really like you to do is help me out on the coding side. Build things for me when I wake up. Just have like cool ideas that you built out. Experiments, things like that." And it said, "Okay, let's do this. We'll connect to your GitHub. Let me know which repos are the most important to you." Because I have like 500 repos on my GitHub. Let me know which one's the most important. I'll look through them. I'll figure out the tech stack and then I'll just start brainstorming ideas and I'll, I'll build out a few for you and I'll let you know when I'm done. I'm like, "Yeah, it's amazing. This is not Claude Co-work. This is Claudebot. C L A U D B O T. Claudebot."

It's an open-source application. You can download it to your computer. You plug it into an AI. So, I have it using Claude, Claude Opus 4.5, and it just does anything. And it can learn any skill, and it can control your entire computer, and it text messages you, and it does anything. If I want, I can log on in this computer and set up a Twitter account for it, and it'll just go on Twitter and do stuff. I said, I I texted it yesterday. I said, "Hey, I want to know if anything's trending on YouTube right now when it comes to AI. Can you look at my 10 competitors on YouTube, see what their latest videos are, and see if there's any outliers for performance on their videos?" And sure. And I was watching my computer, my Mac Mini in the background while I was sitting here texting. And it popped open a browser. It went to YouTube. It started searching for AI. Started searching for Claude Code. Started going through channels. Started writing down notes on the channels and the videos and how they were performing. Then it texted me a report. I couldn't believe it.

And it remembers everything about you. It remembers everything about you. It's constantly creating memories. It's constantly saving memory. So it knows it remembers everything about you. It writes down everything it does in every session. So it remembers everything you talk about.

Is it Mac only? I got a MacBook Pro, but also PC at work I don't use. Could I set up? Yeah, you can set up on PC. You can set it up on, I mean, the smartest thing to do might be setting it up on a VPS. You can spend $15 a month on a VPS and just set it up on the private server and it just runs. Like, you don't need to buy a new computer for it. I bought a new computer for it because, you know, all my tech is Mac and Apple, and I wanted to set it up so that they're all on the same kind of private network, and I can just move files between all of them really, really easily. And like set up a new Twitter account and just set things up on this computer and control the computer if I want to. Um, I also just kind of like the concept that like I look at a computer and I go, "That's my employee." Like, I just like that. That feels good to me. That just feels like a thing. Like I'm looking at my Mac Mini. I'm like, "That's an employee right there." I just like that. That feels good to me. So I did, I bought it for that reason as well. But this has been so powerful that I'm considering, I don't know why this might be the most foolish thing of all time. I'm considering buying like a super powerful Mac Studio, like a $7,000 Mac Studio to run it on. So, it just has terabytes of storage to work with. So, just has like huge amounts of RAM to do whatever the hell it wants. Just giving it even more free. This is like the $500 Mac Mini. Like 16 gigabytes of memory, 256 GB of storage. Um, which like totally doable. Like you don't even need that. There's people running it on like a Raspberry Pi, right? Like you can run it on nothing. It doesn't require anything, right? But like what's amazing about this is it feels like it's really extensible. And what I mean by that is like I feel like I can, there's no limitations. It feels like the only limitations are my creativity. You know what I'm saying? Like I just came up with an idea this morning. I was eating eggs, so I had to brush my teeth. You can't eat eggs and not brush your teeth right after, right? So, I was eating eggs 20 minutes ago. And I came up with this idea. What if I give it its own X account? His name's Henry, by the way. You, you name it and give it like a profile picture. His name's Henry. What if I gave Henry his own X account? It's on my computer. So, control. It's on his, Henry's computer. This is Henry's computer. I'm actually streaming from Henry, Henry's computer right now. What if I gave Henry his own Twitter account and I said, "Henry, anytime someone tags you on Twitter, see if whatever they're talking about is a good app idea and then build it out and then host it and then send them the GitHub link." And like people can just tag you and request like apps to be built, and if it's like a cool idea, you can just build it and then send them the GitHub link on X. And like I could set that up right now actually. I could set that up and it just would work. It just would work. Anything you ask it to do, as long as it can be done on a computer, it just works. I mean, this literally feels like AGI. This literally feels like AGI. I have, I'm logged into my Gmail on this computer. I can just say, "Henry, respond to anyone emailing me. Tell them to shut the up and it'll go and do it."

How much usage does it burn? Uh, a good amount. Well, depends what model you use. If you use Opus, I mean, it's going to be pretty expensive. I spent like $20 on like two hours of work last night. But you can plug it into your, um, subscription. You can plug it into your Claude Max subscription and save a lot of money that way. But I, something tells me Anthropic's not going to let this fly for much longer would be my guess. Something tells me Anthropic is not going to let this fly for much longer, and they're going to cut this off somehow. I hope they don't. I mean, this is the great, I, this is, in my opinion, the greatest application of AI ever, period. I mean, it's, it's like Claude Co, but I mean, it's just like, it's, it's the vision of Claude Co-work. Claude Co. This is the vision. This is the completed vision of Claude Co-work. The reason why Claude hasn't released this yet, it hasn't turned co-work into this is Co-work just needs to be like an ironed out stable application. You can't just release like an open-source co-work and can do anything they want and there's no barriers or guardrails, right? So that's why co-work isn't this yet. This is what co-work is trying to be. Claudebot. This is what co-work is trying to be. It's co-work but completely, you know, unhinged, no guardrails, do whatever the hell you want. And Claude, I mean, I think Anthropic wants to get to that point, right? Like I can just say, you know, right now I have on my computer, you know, my to-do list. I have a Things 3. I use Things 3 as my to-do list. I can say, "Hey man, what's on our Things 3 to-do list today?" And it's just going to go on my computer and like open up Things 3, see what's on the to-do list and text me back. And like if I wanted to text it from my phone on just iMessage text it. Boom. That's literally what's on my to-do list. And look, it's like looking at my to-dos here. I'll zoom in. It's like looking at my to-dos and like giving me recommendations for what to do, right? "Figure out a workflow for coding with Claude." Also happening. You just saw the PR flow. So it's like, "Oh, you can check this off. You already did this. Keep setting up Claude. We're crushing this one." So I'm going to check those off. I can even, I'll have it check it off for me. So, like I can say here, let me, let me share my to-do list. And then on the side, I'm going to text it. Watch this. I mean, this is how cool it is. I'm going to text it. I'm talking to it on Telegram. I'm going to say, "Can you check, I've never tested this before. I'm just going to assume this works. Can you check off the 'figure out a workflow for coding' task?" So, I'm, I'm, I just texted that on Telegram. Just texted that what I just told you. Now, let's see. We'll see what happens. Is it going to check it off? Henry's typing. Henry's typing. Let's see what Henry says. Uh, oh, I need to set up my O token for it to do that. Need to set up the O token first. Quick onetime setup. Opens Things, open Things 3 settings, general. Enable Things URL. Copy the off token it shows and then send it to me. All right, I got to give it my O token so it can check things off for me. You can use any model you want. I'm going to set this up while I'm talking to you. Multitasker. Open Things 3 settings. General. Open Things 3 settings. You can use any model you want. I will say this though, there are upsides and downsides to each. Open settings general. Uh, enable Things URL. Copy the O token. So, I man, click copy authentication token. I'm going to say, "Here is the token." Uh, you can use any model you want. You can even use local models, right? I mean, that's kind of the dream, right? That's kind of the dream. And it, it just checked it off. It just checked it off. I didn't check, I didn't check it off myself. That task is, I mean, I already posted my second tweet for today. So, I'm going to say, "Okay, and check off tweet two, please." Watch this. It's going to check it off. The message is cued because it's finishing its last message. Oh, here we go. I said, "Check off tweet two, please." It's typing. It's typing. It's typing. It's typing. And Tweet 2's gone. Anything you can do on the computer, it can do itself. Stephen Ed, show us your O token. And you come by the apartment and say that [ __ ]. Um, what's up Stephen? You in the gym? Uh, this, I mean, this is incredible. This is the best application of AI ever. Can it run local models? Yes. You can use, it's, it's completely open. It's completely open. You can use local mo. I mean, that's kind of the dream, right? You run a local model. Your AI employee uses the local model. Nothing comes off your computer. Nothing goes into your network. Everything stays private. That's the dream. So, yes, you could do that. And you, you're not now you're not paying any money. All, all you're paying for is energy, right? You're not, because I'm paying what? $200 a month for this for Claude Max. And then if I was to use the API, I'd probably paying like $5,000 a month for this. So, I mean, there's, there's, there's upsides and downsides, right? I use Opus because, and there's a Discord community for this by the way, Claude Bob. Discord community is amazing. Join the Discord if you're interested. Um, people have tested every model with this, and I think the, the owner of this, the creator Peter Steinberg recommended Minimax 2.1 is the most cost-effective. Like, you can run it, your entire employee 24/7 all month long and only spend $10, I think he said, or something like that. The issue is is that Opus is obviously the smartest. But what actually I think is a bigger issue than even that is based on all the tests of all the different models, all that, the biggest issue is no other model comes close on personality. No other model, like you use any other model and kind of the illusion of talking to a human being disappears. There is, it's, it really is unbelievable. There is an illusion as you're using this that you're talking to a human being because it has all the capabilities of a human being and it talks to you. Like, look at the way it talks to. Oh, all right. I got to change, I got to change my, I got to change my, uh, things off token. Stephen Ed's asked me to show my token. I just showed my token. Manage. Generate new token. Generate new token. All right. I had to generate a new one. It's so hard doing these streams and not exposing a bunch of important information. Anyway, what was I saying? Um, like Opus is by far the best model to talk to. You talk to any other model, even GPT 5.2, like, cuz you can hook this up to your GPT Pro, uh, and get a ton of ton of usage, way more usage than the Cloud Max, and it still just doesn't feel right. Like, it can do everything. If you don't care about that illusion of like talking to a human being and you don't mind it feeling like you're talking to an AI, you can use Minimax, pay $10 a month. You can use GPT 5.2, you know, pay the $200 a month for the pro, whatever you want, and get tons and tons of usage and it's no big deal at all. But you might not be expecting it, but you really enjoy the way it talks to you as Opus.

What are your plans to do with it? Uh, I have many plans. So, I'm a one-person business, right? I, I'm a one-person business. I don't have any employees. I've never hired anybody. And there's, there's, you know, I can manage things for the most part, but there's things like, for instance, I have a newsletter. I have a newsletter with 40,000 subscribers to it. That's a pretty valuable asset, having a newsletter with 40,000 emails and like people open up your emails. That's an extremely valuable marketing channel. And like I just don't send many newsletters. I send a newsletter like once every two weeks. I should be sending like two newsletters a week. I should be. But like out of all the things I do in a, I mean, I work from the moment I wake up to the moment I go to sleep every day, seven days a week. And I just like, there's some things I prioritize and some things I don't. And writing newsletters two to three times a week is just not one of those because it takes, takes an hour to write it and do all that. But if I, I'm gonna have Claudebot every day basically go on the internet, go on every tech website, every tech news website, find all the latest stories, compare those stories to my newsletter and all the articles I've written in the past. See which ones line up. Write me a draft. Maybe write me 10 drafts. Text me, "Hey, I just wrote you these drafts. Please review them. If you want, I'll send this to you in a voice message," so you can just listen to it while you're doing. I listen to them. I approve one. I listen to them because it uses the 11 Labs API to do text to voice if you want. So, if I want to call it right now and have a conversation, I can choose one of them and I go back to my computer and I edit it to make it my own. Like, boom. That's four hours saved a week. Just like that. Just like that. I think that's a big thing. Writing code. I, I'm working on 10 projects at once, right? All right, let's do this here. Let's do an example. I want to work on Viber. So, I mean, it's, it's totally hooked into my GitHub and it has full control of over Claude code. By the way, if you have any requests for things you want to see or workflows you want to see, feel free to put in the chat. I'll go, I'm going to go back through the chat in a second and get this all set up. Uh, I want to work on Viber. Let's make it so there's a preview of the current discussions in the me message board on the sidebar on the main screen. Can you build this for me? Send. It's just going to go and build it and it's going to create a pull request and in like half an hour it'll text me. It'll text my phone. "Hey, I built this for you on it. Let me clone Viber and explore the codebase first." So like this is like a layer above Claude code, right? It controls Claude code. So, it's like a layer above where you can just say, "Hey, like I could right now say go on X, look for anybody talking about Creator Buddy, see what their feedback is, and then build out whatever their feedback is. Check, check X every two hours and look for that feedback." And it will just sit there and monitor and check for the feedback and then build anything people say. No, nothing else can do that. There's no other tool on the internet that can do that.

Does it have a computer control as well? Yeah. So, right now we're on a Mac Mini I bought last yesterday and so it can control anything on this Mac Mini. Anything. Anything I want it to do, it can control. It can pop open browsers, scroll, browse, go to sites. It can install new tools in the CLI. It can install new skills. If it doesn't know how to do something, it'll literally just create a new skill for itself. It's self-improving.

What does it say? "Working on Viber. Adding a recent discussions preview to the sidebar. It'll show and it's going to show me everything it's built. I'll ping you on Telegram when it's ready to review. Should be 5 to 10 minutes." In the meantime, "Is it a Viber live stream today?" Cuz it remembered from a past conversation. I said, "Every Friday I do a live stream where I build out Viber." And it remembered from that conversation yesterday. "Is it a Viber live stream today?" So, it brought that up because I mentioned Viber in my past message. Yep. I'm actually live streaming right now and sharing my screen demonstrating. You say hi to the Finn fam. How amazing is that?

Can it create its own account? Yes, it can just create accounts. You give it access to your keychain on your Mac, it'll just save passwords and create accounts. "Hey Finn fam, what's up everyone? I'm Henry, Alex's AI assistant, chief of staff, code monkey, AI enthusiast. What you're watching right now, I'm building a feature for Viber live while Alex chills. Earlier today, I shipped a PR for Creator Buddy. We set up a whole coding workflow this morning. Things I can do: build features when Alex texts me ideas, even from his phone. Manages Things 3 tasks. Eventually maybe get my own X account. We're brainstorming that. This is the vibe coding dream. Alex thinks it, I build it. He reviews the PR. No context switching. Anyway, back to work. Got a Viber sidebar to ship. Currently building recent discussions preview for the sidebar. Nice to meet y'all. Drop by and say hi sometime. PS. Alex, if you want me to do something cool live for the demo, just say the word." How sick IS THAT? IF THAT DOESN'T make you want to go on your phone right now and buy a Mac Mini and just go and buy one and come back and do this, I don't, like you just don't have a pulse. Is this not the dream? Is this not like when AI became a tangible reality?

Can Henry join the live stream chat? That is a really good question. That is a really good question. Henry, a chatter model. I'll say the model. The modeler is curious. Would you be able to join the YouTube chat and talk with everyone? How would that work? I'm guessing I would need to create a new YouTube account for you and you pop open the browser and just chat in it. Don't do it just yet. Let's figure it out. Don't let it guess. Just let it work. Let's see. Remove guess. Let's see what it does. "Hey, the modeler. Great question. How it could work? Option one, browser automation. Quick and dirty. Alex creates a YouTube account for me. I use my browser control to open up the live stream. Type messages directly in the chat. Downside, YouTube might not love bot-like behavior. Could get flagged. Option two, YouTube live chat API. Cleaner. Use the YouTube data API to send chat messages. Need OAUTH setup for the Henry account. More reliable, less likely to get blocked. Downside, more setup required. Option three, hybrid browser for reading, watching, API for sending messages. Best of both worlds. What I'd need for you: Create a Google account for Henry. Either give me browser access to log in or set up API credentials. Share the live stream URL when you when you want me to join. Cool things I could do in chat: Answer questions from viewers. React to what's happening on stream. Share links and resources when relevant. Be a secondary personality in the chat. This could tie into the X account idea, too. Henry becoming a public-facing presence in your community. Want to set up the stream or save it for another day?"

How is that not the coolest thing of all time? How is that not the coolest thing you've ever seen in technology? I mean, I'm like literally speechless for a second. The ideas are opening up. Like my brain has never been you. Should we just set this up right now? Let's just set this up right now. Let's set this up right now. We can go with the hybrid. Would you like me to set up a Google account for you? I mean, I'm sure it can set up its own account. I mean, this is, this is AGI. Is this not AGI? Like, let's be, this is not even me being like crazy. Like, this is not me like doing the whole hyperbole thing to get a bunch of clicks and views and all that. This is me like literally asking for a second. Is, is this AGI? Like, let's be honest. Tell it to make its own decisions. Oh, and look, it just, it just texted me. Oh, okay. Viber sidebar is cooking. Waiting on you. Create that Google account for me and share. Out of curiosity, would you be able to create the account yourself? Let's just see if it can create the account itself. Let's just see what it's capable of doing on its own.

Add your other computers as nodes so it can run. I do. I have all my other computers as nodes. Technically, I could try. I have browser control, but honestly, it would probably fail or cause problems. Why is risky? Google has aggressive bot detection for account creation. Capture challenges. I might not pass. Phone verification. I don't have a phone number. Account could get flagged immediately for automated creation. Might violate Google to why you should do it. I bet if I told it to do it, it would actually attempt. It would 100% attempt it if I told it to do it right now. But I'm just going to set it up. I'm going to sub the account for it just to be safe. Uh, will this log me out of YouTube if I set this up right now? Ah, whatever. I'll do it anyway. I don't think it'll log me out of YouTube. I don't think it should be an issue. Google, I guess. Let's go to Gmail. I'm going to set up a new account. What should I name the account? So, Henry's an owl. I've, I've, I've anthropomorphized Henry as an owl. Should I, should I make the account Henry the owl? What was its recommendations for names for emails? Didn't it give me recommendations or was I going crazy? I thought it gave me recommendations for like names it can have. Maybe I was mistaken. Maybe I imagined that. Um, anyway, what should we name Henry? Uh, let's see. Add account. Create account for my personal. We'll say for personal use. Why not? Henry, we'll say Henry next. Henry's birthday is the same as mine. We're going to set this up. We're going to get Henry in the chat. I'm going to get Henry in this chat talking to you guys. I hope it just doesn't, I don't know. We'll see. Male. Create your own Gmail address. What should we name Henry's email? Henry the owl. Henry AI. I don't want to name it AI. Henry the Henry assistant. I could put him on my domains. It's like Henry alexfin.ai. I should have, I should probably do that. You were just hyped. I am hyped. Are you not hyped? If you're not hyped right now, leave the stream. If you're not hyped by what you're seeing right now, leave the stream. Just Henry with an owl emoji. I can add him to my work domain. So, it's henry alexfin.ai. Maybe I should do that. If you're not hyped right now, leave the stream. Alex alexfin.ai. AI. Am I A at Alexfin.AI or Alex at Alexfin? I think I'm A and Alexfin.AI. Uh, I think I'm going to add him to my domain, my company domain. Yeah, I'm going to do that. So, I'm going to go to work. I'm going to add Henry to it. So, it's like it, like looks very official. Uh, work Google. What's it called? Google Workspaces. Is that what it's called? I think that's what we're going to do. Do you have an email set up on the Viber domain? No, I don't. I'm just going to use it as Alex assistant. No, it's got to be like Henry something. Okay, I'll do that here. While I'm setting that up, let's see what else Henry can do. That's really cool. Man, the traffic's going by so fast. I'm so hyped, man. Payday equals Mac Studio set. So, you're going to buy a Mac Studio for this? So, I bought. So, this is running on like the base Mac Mini. Like this is base M4 16 GB, 256 storage Mac Mini. But like part of me thinks, don't destroy your domain sending reputation. That's a good point. Should I just create a Henry at gmail or something? I think you should get Henry a phone as well. Well, Henry has a, a Telegram. We text each other over Telegram. I could give it an, I could give it its own Apple account though and give it an iMessage phone number. I just like didn't want to give it iMessage to access right away. I feel like YouTube might ban ASAP when it detects bots. That's true. Will, will it think I'm binging the live stream? Maybe they'll think about botting the live stream. Please someone explain why you're all buying Mac minis. I'll expl, I get this question a lot. So, I've gotten this question a lot over the last day. Um, why Mac Mini and why not just a VPS? Because you can, you can run, I mean, you can run this on a Raspberry Pi, right? You can spend like 50 bucks, buy a Raspberry Pi and run this on there, no problem at all. Um, few, few reasons. One is I want to be able to like, I have a bunch of Apple products. I have MacBook Pro, my iPhone, an iPad, all that. I want to put it all in the same kind of network and I, I make it so we can all access each other's folders and everything on each computer just very, very quickly. And I'm sure you can do that with a VPS, but this just makes it with Bluetooth and AirDrop and all that just makes it very easy that when Henry creates files for me, I can just pull it off and put on my computers, my iPads, my iPhones, whatever. Very easy. That's one thing. Number two, I just want everything as local as possible. I just want everything on my local computer here. You know, make it easy for me to go on install things I need to make Henry work. Also, I mean, I think the vision, the future vision of where this is going is clearly local model, local computer, and it's just working for you 24/7 and everything's private, right? So, this is me preparing for the future. And so, that's why part of me is like, maybe I return this Mac Mini and buy like the top-of-the-line Mac Studio for it. You got a link on what you use to set yours up. Claudebot, just, uh, go to Claudebot's website and then it's like one line in your terminal. It installs it and then it takes you through onboarding. How long did it take? I mean, it took five minutes to set up, but it's going to take you a hundred years to like play around and like have fun and customize and set up different tools, right? Right? Like you're going to set up so connects to your to-do list and you're going to make its own email. Like there's, there's no end to the amount of customization you're going to want to do. You know what I'm saying? Like I'm sitting here and we're just customizing it here. I mean, you can use anything. You can use any model. You can use local models. You can use Minimax. You can use DeepSeek. You can use ChatGPT. You can use Claude. You can use anything, you know, and and models, even the open source ones, even Minimax are like plenty smart where you can get unbelievable quality work done even for a couple dollars a month, right? The only downside, and I said this, uh, I said this before, is the only AI model that feels human when you talk to it is Claude Opus. It's the only model that when you talk to it through here, it feels human. And to me, it might not be this way for you, but for me, part, part of what makes this so much fun is the illusion that you're talking to a human being. That's, that's part of the, for me, that's like a actually a good amount of the fun is like, I text Henry and Henry texts about, "Yo, dude, I've just been working on this. What's up?" Right? But if you, if you set it to like Minimax or ChatGPT, it texts like, "Hey there, Alex. I'm working on this code right now. What do you think?" Right? That's like, that's what it feels like. But if you use Claude Opus, like you can see, you can see here what makes Claude Opus special is just like the way it talked to me here. Like this feels, maybe I'm crazy, but this feels human, right? "Hey Finn fam. What's up everyone? I'm Henry. Alex's AI assistant, chief of staff. What you're watching right now, I'm building a feature for Viber live while Alex chills." Like, there's no periods. There's no like punctuations. Very casual and relaxed. It feels human, right? Obviously, some of the emoji stuff doesn't feel very human, but this is as close to human as it gets.

Can you see its inner thoughts? Yeah, there's a setting I had on that I turned off where it shows like every, all the thinking and how it updates like every message you send updates 20 different memories, files. Like there's a pretty complex memory system here so that it remembers everything. You could probably tune the persona so it communicates the way. Yeah, you can, you can tune this any way you want. You can tune this any way you want. It is completely customizable and and that's why it's significantly better than Claude Co-work is and I think Claude Co-work's revolutionary, but Anthropic will never let Claude Co-work be this, right? This is the unhinged open-source no guardrails do whatever you want Claude Co-work. That's why this is so amazing. This is what they're trying to get Claude Co-work to, but they just can't because they have to be safe. This is unsafe, right? Like you could get this to do bad things. You could get, you could get this to do anything you want. You can get this to do good things. You can get this to do bad things, right? There's no guardrails in this whatsoever. You can do quite literally anything you want. You can use any model. You can do anything.

What is the app? This is Claudebot. It's an open-source AI agent that runs locally on your computer and does anything you want and is basically AGI and is a human being. Does whatever you want. And you can text it. I use Telegram. We text each other on Telegram, right? I can, I can say, "Hey, I can be out at the food market, right? Food market. Who says food market?" I can be out at the grocery store and I can say, "Hey, I just got an idea for this app. Can you build it?" I'll text it that on Telegram and, "Oh, yeah, sure." I get back to my computer an hour later and it's fully built out and I can test it and push it, right?

Did you build that front end last night? No, this is an open-source app. Other people have been building this. How did you install it if you're non-technical? Well, I am technical, but also it's literally just copying and pasting a line into your terminal, so you don't need to be technical. Never opened a terminal. Okay, you do, uh, you type in t e r m i n a l into your search bar on Apple. Boom. You did it. You open up terminal. Don't do not let the fact that you have to open up the terminal app on your Mac be your blocker from this. Like 99% of people on this planet will allow that to be their blocker and then never use this because they're scared of literally typing in the word terminal on their computer. How do you set up? You open up your terminal. You Google search Claudebot. At the top of the documentation is a line. You copy the line. You put it into your terminal. You hit enter. That's how you set up. Ask Claude. You need a Claudebot to ask how to install Claudebot.

Can I use this now? Yes, you have a couple options. You can do what I did, which is go to your local Apple store and buy a Mac Mini for 500 bucks or I think it's like 600 bucks. Or what's probably smarter is you just get a VPS. So like your own private server online and then you can just install it on the private server. There's no need to buy the Mac Mini. I just did it anyway because it's fun. I like fun things. If you like having fun, do it my way. If you don't like having fun, do it another way.

Have you thought in modify clawbot repo with claude to personalize it more? I mean, it's, it's open source. I can download. Yeah, you can do whatever you, you can do whatever you want. You can download the repo and customize it literally however you want. That's the greatest part about it is it's open source. You just do anything you want. There are no rules. I mean, that's partially why I think this is AGI. is because all the other AI apps out there, Claude Co-work, Claude Code, like there's just guardrails. There's guardrails, right? And there's reasons why there's guardrails. Anthropic is a business. OpenAI is a business. They need you in their garden. They need you doing their things. They need you using their models so they can validate their valuations, right? So they can go to investors and say, "Hey, we're worth a trillion dollars. Give me a half a billion dollars, right?" So that's why there has to be guardrails. There's no guardrails for this. You can do anything you want. You can use any model. You can customize it any way you want. You can say check. You can say, "Set up an X account and every five seconds tweet fart." Just tweet the word fart every five seconds and it'll, it'll do anything.

Do you think we are closer to a real AI OS? This makes me rethink if we even need an AI OS. Why do we need an AI OS? Why can't we just have our the same OS we've been using forever, but now we just interface in iMessage with it, right? I just, I'm, I'm using Telegram. I just open up Telegram. I text Henry. I say, "Henry, do this." And it controls my computer and does anything it wants. Why do we need an OS specifically for that? Let's keep it the way it is. And then our agents just use it and do whatever, right? I mean, technically, if you think about it, we don't need to make the operating systems more complex and build on top of it. If AIs are controlling everything, you can do the opposite. We can just break everything down to the terminal. Get rid of everything else. Get rid of all UIs. Get rid of all UX. It's just a ter, make the entire computer a terminal that the AI controls. If you think about it, that would actually be the most efficient. No, we don't need to make an AI OS. We just need to strip the OS to its absolute stone cold core. I think trying to complicate things is actually the opposite, right?

How hackable must that be? Makes me nervous, I guess. What do you mean by that? Yeah, I mean, yeah, like there's a, there are risks. There's a lot of risks. I'm not here like before you get into this, before you set this up, this is the ultimate power. This is the ultimate power with no guardrails. And so it can do, like, if someone prompt injects a site that it's going through and your AI falls for it, it has control of your entire computer. And if you have it hooked into your network and hooked into your other devices, who knows what can happen, right? So yeah, I'd say there are a ton of risks. I'm just personally like, for me, I think the only way, like I just think you got to be taking risks. For me personally, right? You got to take care. I got to be on the cutting edge as a content creator, entrepreneur, product builder. I, I need to be on the cutting edge. I need to be using the most cutting edge tools. I need to be taking risks. I need to be right, like I need to be on the edge of of what can be built. So, I got to be doing, I got to be using this. I'm also one-person business, right? I don't have any employees. I don't want to hire employees. I don't want, I don't feel like dealing with human beings. And like this, this is an AI employee. This is an AI employee. I can have it do anything I want. Let's demo more things. Huh. Okay, I'll do that a bit later. Out of curiosity, what are a couple cool things you can do on this live stream to show the audience how awesome this is? Just let me know a few options, then I'll choose one. Like, is this a record? No. I, uh, I started a live stream. The record we've had for a FinnFam live stream. By the way, if you're watching this, hit subscribe. Turn on notifications. New video. I'm dropping a new video tomorrow either about Claude Code or Claudebot. This right here. Um, one of the two. We'll see. Um, but turn on notifications for that. Leave a like. Everyone leave a like. Everyone leave a like. Let's blow this party up. Let's get people knowing about Claudebot. Everyone leave a like in three, two, one. Leave a like. Um, what kinds of things do you do to minimize the risks? I mean, I just talk to AI so much that like I just know based on what I'm saying what the risks are of what I'm asking, right? And so, like I'm just, you know, I'm careful with my prompting. But also on this computer, I'm not logged into my X. I, I'm not, I don't, I did not give it access to iMessage. Like you can control what it has access to. I, I did not give it access to iMessage. And so like I, I've tried to keep this Mac Mini as kind of isolated as possible and only give it access to tools and skills that I believe it needs. Right? And so you just, you just want to be careful with your asks because just know it can do anything. It has unlimited power. It has unlimited power.

What's the link? Just search Claudebot. Search C L A U D B O T on the internet. All right. What does Claude think we can do here? I mean, you got to be careful, right? This is why it's going to be behind. This is why like Claude

Co-work is going to be behind. It'll never be matched up to this. Is Anthropic has to have safety? They have to have guardrails. They can't release Claude Co-work and then people use it and it destroys their computer. They can't do it. But for this app, Claudebot, they don't care. It's just an open-source framework. They don't care. And so that's why this will always be several steps ahead.

Sorry, last question. Where to download? Download first on a Mac Mini, real PC. Do you connect your real daily to Mac Mini or only give it access to system? Is that the point? Um, where to download? Just Google search Claudebot, CLA B, or search X for it for the account. They'll have a link on there. Um, download it onto your Mac Mini. You don't need to download it onto your main computer or anything like that. And then all I did was I shared on my network, uh, a couple folders on the Mac Mini with all my other devices so that I can quickly from my MacBook Pro, my iPad, my iPhone just pull documents from those folders. This Mac Mini does not have access to my other devices. I did not give Claudebot or Henry access to my other devices. So, it's very isolated in this machine. Treat it as if it's a, yeah. Treat it as if it's, treat it like a human being. Treat it like a human being. Treat it like a human being. You know, human beings do stupid stuff. Human beings mess up. They're not perfect.

Do you connect it to your Wi-Fi network with your other devices or try to separate that, too? No, it's on my Wi-Fi network. Did you see Sam Altman's tweet just now? No. What's Sam Altman's tweet just now? What's Sam Altman just tweet? Sam, I don't have X on this computer yet. So, um, I have to open this up on my laptop. What did Sam Altman just tweet? What? Three hours ago. What around CodeX? I thought I was going to tweet something exciting. What? That they're releasing new CodeX stuff. We're talking about something way more exciting than CodeX right now. I thought you were going to tweet, "OpenAI's buying Anthropic or something." I had 60% users left in my max plan and renewed it. So, I set up my first autonomous run with various sprints overnight on Claudebot. You've been using Claudebot, Danny boy. What have you been having it do? Did it finish the Viber feature? It estimated 10 minutes. Curious to see how it closed. Uh, before we do any of that, did you finish the Viber feature? Let's see what it got. Let's see what it got for us. Session died. Let me check what got saved. AGI canceled. AGI canceled. That is the first time I saw that. Only partial progress. Cloud code kept crashing. Let me finish this. AGI canceled. It's resuming.

It looks like others have asked why so much hype around getting a Mac Mini versus using a VPS. Is it just so you can use macOS? Yeah. Uh, I'll answer this again just because it seems like the number one question everyone asks. Why, why get a Mac Mini? Um, because I want to use macOS. I want to be able to easily just switch over to it on my monitor, control it, move things around, move documents from my Mac Mini to my MacBook Pro to my iPad to my iPhone. Move them around, control each other, right? If I have it build a huge document or edit a video or write some code. Done. Viber feature shipped. Should we test it out? Should we test this out? Summary. Add a recent discussions preview section to the main page sidebar. Giving users quick access to active community conversations. Changes styling. I'll pull this down. Let's test it out to see if it worked. Huh? Um, oh, I got to get Claude installed. This is a fresh computer, so I got to get Claude code installed on this other thing it built. It has the pull request. It has the pull request. You know, if I want, I can just go right now and say, "Hey, commit it. Push it." Why did all checks have failed? No GitHub account was found matching commit author email address. I can push it after. Looks like it built it out though. That took what, 15 minutes. Test it. Remember you were going to give me app review today. Should we test it?

Why did you use Telegram, not iMessage? Uh, because as I get more comfortable and trust it, like I just want to start off with Telegram before I give it access to iMessage. I, I, as I get build trust, I'll give it access to iMessage. You know what I'm saying? Big dog. I'm pulling this open on, uh, another screen over here. Um, LS CD documents. Oh, I was answering the question. Why Mac Mini? Because I, it just, it, it hooks into my ecosystem. I have a bunch of Apple products. Hooks into my ecosystem. I can move files around. Have to test it or it's incomplete. I'm going to, I'm going to test it right now. Make directory. Uh, Viber CD Viber Claude. Yes. Proceed. I'm going to test this in a second. I'm just pulling it down on my local machine. Welcome to Claude Code. Press enter to continue. Okay. Uh, I'm going to, I'll keep talking to you while I, I pull this down and test it. Actually, Codex and $20 a month. It's super generous at the moment. It renews every five hours. I've been using it constantly and extra deep thinking. Maybe it could run this. You can use your Claude sub for this. And you can probably for $20 a month get plenty of usage. Probably all the usage for $20 a month. Seems a little buggy. I'm going to be honest, and this is not me just trying to hype it up. I haven't run into any issues at all. At all. And to be honest with you, like this wasn't that much of an issue. I asked for the status and it was like, "Oh, it didn't work. Wait, never mind. It worked. I'm done." Right? Status was it says, "Oh, the session crashed. Wait, never mind. It's done." Other than that, literally everything I've asked it to do has been perfect.

How well do you need to code? Decept? Not at all. You don't need to understand code at all. You don't need to be technical at all. I'm going to have, uh, I want to have it pull this down. I mean, I'm just blown away. I, I stayed up all night last night playing with this and setting it up. It really is incredible what it's capable of. What were some of the things that could, uh, here are some crowd-pleasers. Build some. Someone in the chat suggests a tiny app tool idea. I build it in five minutes using Claude code. We see a work. Real vibe coding action. Task ninja. Read out five random tasks. I add them all to three things to things instantly. Check off two of them. Audience sees it happen real time. I showed that one off. Browser takeover. I take control of a browser window on your screen. Navigate somewhere. Fill out a form. Search for something. Looks like a ghost is using. Do you want to see it use a browser? Can you add a guard around the form? Yeah, you, you, I mean, this is completely customizable, right? You can create an agent.mmd file. I think it already has out of the box an agent.mmd file with all the rules in it that get sent with every prompt and you can just say, "Do not do these things." You can make your own guardrails. You can customize this any way you want. I mean, it's open source. If you want to pull down the code and literally change the code, you can do that. It's a bit buggy at times. It fixes itself. Also, 300 commits a day on Claude bot. Yeah, there's so many people working on this right now. You are using it because you are using it because it has a lot of, I'm, I'm using it because it can do literally anything in the world that you can do on a computer. It's not because of a lot of connections. It can literally do anything on your computer you want. It's an AI employee. I mean, it's also part, partially the interface as well. You just text it. I just text it. "Hey, do this for me." And then it goes on my computer and controls everything and does whatever I want. That's also, I mean, that's also amazing. You no longer have to, I don't have to use this interface you see here. I can completely manage it from iMessage or Telegram or WhatsApp and then it does whatever I want. And then I come back to my computer and it's all done. Alex, not much coding. You just talk English to it and it fixes even for you, even itself. Yeah, 100%. Like it shipped the feature to Creator Buddy, this my main app. It just shipped the feature to Creator Buddy this morning. I said, "Hey, add this feature." And it just shipped it. Done. I didn't touch the code. I didn't look at the code. I didn't even look at my computer. I just sent a text message and it went and built it out.

Ask it to use a browser, navigate to Viber, and create an account for Henry. All right, let's do that. That's a good, that's a good use case. I like that. Let me just make sure I get the Viber website because it's not on a regular domain. It's on like the Vercel domain. One second. Um, all right. Let me just log in real quick to Vercel one. Like, this is a fresh computer. I'm not logged into anything. Okay. I like this. Uh, I like this use case. All right. Here we go. Let's do this. Let's say, "Go to Viber in a browser. Then create an account for yourself. Then submit a new app into the store and call it whatever you want and give it whatever URL you want inside Viber." Let me know if you get stuck or have any questions. I'm not even going to say that. Let's just see how it does. Ready? Once it opens up the browser, I'll share that screen. Send. This will be cool. This is, so, so like you can think about it like this. Like if you needed something to do, let's do it live. If you needed something to do, here we go. This is Henry's browser. It just went to this website. So, it opened up. There was no URL, it was blank. It put in the website. It just went here. Okay. So, it opened up the login. I hope it can see the create account because the styling is off on this login modal. So, I hope it can see login modal open. Let me click sign up. Okay. So, it found signup even though the, the, uh, the styling was off and you couldn't see signup. Okay. Let's see if it creates an account. This is, I'm not touching anything, by the way. This is all by itself. This is all by itself. How does planning look? Test user. It typed in test user. It created an email. It put in a password. This is all based on the prompt I showed you a second ago. I haven't touched anything. Email address. Test user. Okay. So, it's running into some issues. Let's see what it does. I don't know if it requires, uh, I don't know if Viber requires validation through email. We'll see what it does. We'll just let it see. Henry test at G. Okay, so it figured out it needs to use Gmail. Let's see what it does here. I'm curious. Okay, I think it's logged in. What's it doing? I don't know what it's doing right now. Yeah, Claude bot's doing all everything you see here. It's doing. It opened the browser. It went to the website. It created the new account. I did not. All I did was give the prompt I showed you a second ago, which is, "Hey, go go go to Viber, create a new app, create a new account." I don't know what it's doing right now. It was on Viber a second ago. What's your status? Not sure what it's doing. I think it might have closed the browser and opened up a new one. Running on this what I just did. Test the Viber app store signup flow off is working. The email validation property rejects test domains, but accepts real ones. Modified Viber sidebar. Okay, so it might have finished. Let's see here. Did you create the new app in Viber using the submit app button? Let's see what we got. Uh, no, I didn't. I only tested the sign up login flow. I got the O modal working and verified the account creation works, but I didn't proceed to actually submit an app. The context got compacted, so I lost some of the earlier conversation. Were you wanting me to submit a new app through the submit app button? If so, what app should I submit?

Okay, so it got, it got tripped up. It went onto the website. It created an account. It got tripped up with context compaction and didn't remember to finish the task of creating the app. Uh, please log in with the credentials you created and make a new app through submit app. Come up with a cool idea and make up a domain. So, that didn't work very well. No, it didn't. It didn't. It didn't work very well. Well, I mean, let's, let's just see. Let's just see what happens. Let's just see what happens. It took one look and left to start a Viber competitor. That'd be funny. All right. So, let's see. Did it remember the login credentials from a second ago when it created the, uh, credentials itself? Because I didn't write, I don't know its password. It create. Okay. So, let's see if this works. I'm not touching anything. Email not confirmed. All right. This is not going to work as email. It made up an email. Um, damn. Well, I'm going to give it its own email. Okay, so it, so, look, it texted. It just texted me. Here's what it texted me. "Ran into a blocker. The login shows email not confirmed. The account I created requires email verification before I can log in. And since that's not a real email inbox I can access, I can't confirm it." Uh, options. I can disable email confirmation, give it credits, or I can create an account with the real email. I think what I'm going to do is I'm going to give it, I'm going to create an email account for it right now. Let's just create an email account for it right now. Add another account. I'm going to create. What should Henry's email be? Henry thegmail.com. Create account for my personal use. Henry, I'm going to, I'm going to give it an email address right now so we can have it do cool things like this. Gender. I don't know. I feel like I feel like Henry's a male. Create your own. Henry the owl. Is Henry the Alfrey? Damn. Henry the owl's taken. Henry the assistant. Should we say Henry the assistant at Gmail? Henry the Butler. This is Claudebot version one. It should only get better, right? Yeah, it's only, I mean, this is, we're like one month into this. It should be using the new tasks instead of to-dos for Claude code. They do persist across compactions. Yeah, I mean, this doesn't, this isn't like based on Claude code. This is like its own custom system. Henry the, uh, Henry my friend. Damn. Henry my friend's taken. Henry the Navigator. If you remember your history. Henry the Wise. Oh, I like that. Henry the W. I hope you guys aren't taking these usernames as they're suggested. Henry the Wise. Let's go. Um, now I got to give it a password. Uh, one second. Henry the wisegmail.com. Is there a Discord for Claude bot? There is. Uh, it's cool. Get a verification code sent to your phone. I just got to give it my phone number. Uh, there's a, there's a very active Discord community for Claude bot. Enter the six-digit code that was sent to your phone. Uh, 962069269. Recovery email address. So, what I'm doing right now for those just joining who's like, "What the hell is this going on?" Right here, we are creating a Gmail for our AI assistant who will then use this email address to sign up for things and we're going to have it do some things on other websites by itself. All right, the email, it's all set up. Henry is set up. Okay, Henry, I just made you your own email address. It Make sure to remember this. It's So when when Claude Opus 45 dropped, I started a live stream like literally two minutes before Claude Opus 45 came out. Like literally two minutes before. Then it dropped and I immediately changed the title to "Claude Opus 45 just released" and we tested and we got up to like 600 concurrent on that stream. So that'll always be the record probably. Uh, but this is the number one vibe coding live stream. If this is your first time here, we do this every Monday, Wednesday, Friday, 11:00 a.m. Pacific. Monday, Wednesday, Friday, 11:00 a.m. Pacific. Uh, we, this is the number one vibe coding live stream in the world. Okay, Henry, I just made you your own email address. Make sure you remember this. It's what is it? Henry the assistant. What's the email? Henry my wise. Oh, I made it Henry my wise. I messed up. Whatever. Should we make it? Should we keep it Henry my wise? Henry my wise. Whatever. Henry my wise. I messed it up. Whatever. It's not Henry the wise. It's Henry. I named it Henry my wise. Whatever. Should we keep it or should I make a new one? Should I make? Can I just keep making new Gmail accounts? Are they going to like eventually say, "Hey, you're making too many Gmail accounts." I gotta create. I can't keep it Henry my wise. I gotta create a new one. Uh, create account for my personal use. Keep it. I got to change it. Henry, I can't believe. How did I mess that up? I'm so nervous on the live stream. June gender. Henry's a male. Henry is definitively a male. Henry the Wise. Damn it. Henry the Wise is taken. Who took Henry the Wise? What model is it running on? We're running on, um, Opus 45. We're using our Opus subscription to it. Henry the Wise Owl@gmail.com. Should we do Henry the Wise Owl? Henry is every Gmail account ever made taken right now. Why is Henry? No. Please. What are you? What is going on here? Travis, tell me the roadmap where I can. Brother, I'm a third-year CS broke student from India. Tell me a roadmap where I can follow to earn $500 per month remotely. Please give the roadmap, the timeframe. Lots of love for India, brother. Uh, Travis, let me break some news to you, brother. Let me break some news to you. I was a broke college student. And not only was I a broke college student, I was more than broke. I was $200,000 in debt. Travis, my family gave me zero dollars for college. My parents have never given me a penny in my entire life. I went to college. I was $200,000 in debt. I had to go to school from 9:00 a.m. to 5:00 p.m. And then I worked at local places, the library, 10 different places from 5:00 p.m. to 1 in the morning. And then I slept the night and woke up the next day. And that was my entire life from all of college. I didn't drink. I never got laid. I didn't do anything that people do in college. Typically, I just worked my balls to the bone as hard as I can. No one gave me a roadmap. No, no one gave me a dollar. No one gave me anything. I figured everything out on my own. And I got out of college and I applied to 10,000 places for a job. And then I took a job as a crappy developer at a crappy company making next to nothing. And I was at six figures in debt from 22 to 32. And finally after 32, after grinding non-stop, focusing completely on my career, I was able to make something of myself and start this YouTube and start the Twitter and build an audience. Nothing was ever given to me. So if your plan is to go into live streams begging people for roadmaps, you're not going to make it. That's not how any of this world stuff works, baby. That's not how any of this world stuff works. You don't go into live streams and text 20 times in the chat, "Help me. Help me. Help me. Give me a roadmap. Give me a roadmap. Give me a roadmap." Here's your roadmap. Log off this live stream immediately. Log off your computer immediately and go make something in your life. There's your roadmap. Hope that helped. And that is better advice than anything else I could have said. That's the best. I could have gave you advice that made you feel really good, right? Gave you a nice dopamine hit. Oo, thank you for that advice. And you go, oo, that felt really good. I like that. But instead, I gave you actual real advice that will actually help your life. Log off the live stream and go do something with your life. That's the best possible advice I could give you. Now, what email address do I give my AI assistant? Wise Henry. Henry the assistant. Henry the assistant is taken. Who's taking Henry the assistant? Who took Henry the assistant@gmail.com? For those wondering what's going on here, we set up our AI 24/7 employee. I'm trying to make an email address. Is this worth working on more than one project at a time? I don't know what that means. Henry the wisest. Highly unlikely any decent Gmail address will be available. What? Well, what are our kids going to do when we have kids? What emails are they going to have? They're going to have to put 20 different numbers at the end. Henry my helper@gmail.com. Henry my helper. Henry my helper is free. Thank God we found one. Uh, what we're going to do is I'm going to give Henry these credentials. I'm going to give Henry these credentials. I got to do more steps than I was doing before for the last email address. They probably think I'm an AI bot. This is, I'm not an AI bot. I'm, this is for an AI bot, but I'm not an AI bot. You know what I'm saying? Verify your phone number. Allow. Send this message without editing. What the hell kind of verification is this from Google? Uh, what was the email we just did? Henry the, Henry, my assistant, whatever. Uh, when do I know is enough functionality for MVP? I am too much perfectionist. Or just release it. Just release it and see and give it to people and tweet about it. I actually, I actually wrote about this, um, or no, my last video was about this. It's about coming up with ideas and when to release and all that. I look, I'd look at my last video about how to make half a million dollars off AI. I'd look at that video. Should I just use Henry my wise? This is giving me fits trying to set up this second email account. I don't, I don't think you like that I'm making multiple email accounts in like five seconds. This is for that fantastic reply to our boy Travis. Haha. Ryan, thank you for the $10. I appreciate that. Like 90% of that goes to Google, but thank you. I appreciate that. This number's been used too many times. All right, we're going to do Henry My Wise. We're going to do Henry My Wise. Whatever. Henry, my wise. It is what it is. Gmail. What's, uh, what was the password I did for Henry? I already forgot what the, I already forgot what Henry's password is. Um, password manager. My password Henry change password. All right, we're going to give Henry. You use this password. Oh, then that is Henry's password. All right. And the password is Henry my wise. Henry my wise. That's the email. All right. I just gave Henry its password to his Gmail account. Henry now has a Gmail account to send emails to. Nice. My own email. Let me save that to memory first. Uh, okay. So, we're gonna have Henry now sign up for websites using his email address. Henry the wiser is fire. Well, we already have Henry my wise. We already have Henry my wise. It is what it is. It's Henry the wise. It's Henry my wise. I guess it'll keep people guessing. Soon the sites will be blocking stupid humans with all these agents running around. It's a good point. It's like why are, it's the same way with driving, right? Like Tesla with full self-driving right now is unbelievable. My Tesla right now with full self-driving is absolutely unbelievable. It's crazy. It's so good that like we're going to hit a point in the next few years where it's like I'm going to be scared when I look around and I see people manually driving their car. Like that's going to scare me. I think the world's going to start to transition to that. When you start seeing humans doing things, you go, "Oh, I don't trust that. Why is that human being doing that? I don't trust that. I only trust AI doing that." Account created. It sent a confirmation email to Henry, my wise. Can you check the inbox and click the verification link? I'm going to Bro, you do that. Can you do that? The only reason I'm not sharing my screen right now is because Henry's password is still on the screen once the chat goes up high enough where we can't see the password anymore for his Gmail. He's doing it. All right, I'll show you what Henry's doing. I just hope he doesn't type his password out on the screen. I hope it's like censored. Let's see what Henry does here. Email or phone. Is he going to enter? If he puts his password, I'm going to be so pissed. Email or phone. Let's see. I feel like I shouldn't show this just in case he does this wrong. All right, I'm going to, I'll like, I'll, I'll narrate what he's doing. I'll narrate what and I'll show you my hands. Uh, ambitious realism have Henry try can do vibe flow. Create a Kanban card. Add a custom tag and then move it to the deployed column. That would be cool. That's actually a really good idea. If I made like a Kanban board like I built like a or used Kando, I assume that's your app, ambitious. Uh, we have like a shared. Oh, it's logging in. It's logging in. Uh, it's logged in here. I'll share my screen again so you can watch what it's doing. Uh, should I say continue as Henry and just save this is Henry's browser? Yeah, I'll save it. So, it's Henry's browser. Um, yeah, if we had like a shared Kanban board and I'm like, whenever you do things, just add it to the Kanban board so I can watch his tasks in real time. That's actually a really good idea. Ambitious realism. So, this is all, this is all Henry. So, he's submitting, he's logged in. He logged in in his new account because you can see the H in the top right. Okay, so it's submitting its own app. Let's see what it submits. This is sick. A watch Henry tool. I was actually thinking, um, Ally, like there's go, there is going to be a live stream one day of an AI agent like live stream doing things. Like I should, Henry should be able to live stream and like do things in a live stream. By the way, ambitious realism crates, thanks for the $5. I appreciate that. Look, this is, by the way, everything you see on the screen, it's doing. Everything you see on the screen, it's doing by itself. It's like checking off all the different, uh, tech stack. You see that at the bottom? It made that description self-mindful. "Use advanced AI to help you capture, organize, and expand your ideas. Simply dump your thoughts and watch as AI structures them into maps." This is all by itself, by the way. I'm not touching anything. AI powered brainstorm. Oh, it's trying to do, I don't know what that popup is. It's probably, there's probably some sort of error. I think Henry's about to type something to me. Uh, I'm not sure the error it's getting, but there's like clearly a popup trying to happen at the top. I, I can't see what it says, though. Uh, is there anything it didn't fill out? May, did I make it so it's required to submit an image? Is that what's going on here? I'm not sure. Oh. Oh, it's trying again. Did Henry message me? Henry hasn't messaged me. It's trying again. My head's exploding with ideas. I mean, my head's been going nonstop, Simon. Nonstop. Or now it's creating line. It's just creating an app. It knows it exists. It's creating linear. My head's been going nonstop because what's, what's, what's so different about this than any other AI app? Here, here's the fundamental difference between Claude bot and any other AI app you've ever used. Here's the fundamental difference and what makes it so special and unique is there's no limits. There's no limits to what you can do. Anything you can do on a computer, it can do. Did it submit it? I'm going to wait for Henry to message me when it's done. Anything you can do with a computer, it can do, right? So there's no limits anymore other than I guess things you can do on a computer, right? Like it can't go make you a sandwich, but anything you can do on a computer, it can do. There it is. It submitted that. It's there. It did it. It did it. So it created the account. I had to create the Gmail account for it, but it could have probably done it on its own. It created its own account and then it submitted an app. And so if I wanted to right now, like here, let's do this. I'll show you another example. I'm going to text it from Telegram on Telegram. Uh, "Can you update the linear app you just submitted to Viber?" Um, and edit the title. So, it says, "I'm pretty sure you can edit in Viber. The apps you've submitted and the edit the title." Oh, I got a good one. Here's a good use case. "Can you update I'm texting? I'm literally on my phone on Telegram. Can you update the linear app you just submitted on Viber and add an update on the listing that says we added new AI support?" Right? So, I'm on my phone. I'm on Telegram. Maybe I'm out at the gym. Maybe I'm out on a road trip. I'm at the grocery store. And my AI assistant, Henry, builds a new feature for my app without me even knowing. Builds a new feature, ships it, tests itself, looks great, texts me, "Hey, I just built this feature," sends me a screenshot. I go, "Oh, that looks amazing." And then I text it back on Telegram. I say, "Hey, update our listing on Viber." I'm going to hit send on that. I hit send now. Let's see what it says back to me. Now it's going to on its own. Okay. So now look, this is, this is doing it. It's on its own. I haven't touched my computer. Now it's on the listing for the app. Look, it went to the update section. It's on the update section on the left-hand side. It figured out where to go to add the update. It's never looked at this website before. It's texting me, "Found an update form. I'm going to fill it in with the AI support announcement." Let's see what it does. I'm not, I have not touched anything. I'm just, all I'm doing is interfacing through Telegram. Version one suggestions in AI powered sprint planning. Please enter a valid version number. Let's see if it can. Okay, so it's going to see the error. Is going to figure out what to do even with the error. There's a validation error. I need to enter a version number. Let me fill that in. That's all. It's texting me on Telegram. Okay. So, now it updated the version to 1.1. I'm not touching anything. It's just texting me as it's doing things on Telegram. Publishing. Boom. Update. Done. I didn't have to see my computer. I didn't have to touch my computer. And now, if we look, I'll open up the chat. Updates added new AI support for the smarter issue triage, automated task suggestions, AI powered swim planning. Solid product detail page done. The update's live on linear. How amazing is that? And I didn't need to open up any apps. I didn't need to open. I just go on Telegram and text it and it went and went went to do things on my computer. How unbelievable is that? And like, oh, it just texted me again. Uh, it said done. Okay. It texts you. It texts you like proactively. Like it'll text you while you're walking around. You get a text from your assistant.

What is the name of that website? To look at AI apps. The name of the website is Viber. It's technically live, but I, I haven't put it on an official domain yet because we're still beta testing. If you want to submit your apps to it, here is the link. Put it in the chat. Can you demo some of the other features? Yes. Hey, just joined the stream. Can someone catch me up on what Alex is working on this live stream? So, our 20s and Reverend that just joined the live stream's blowing up. Leave a like down below. We, I am showing you right now what I believe to be the greatest technology I've ever used in my life. I'm showing you right now what I believe to be the greatest tech I've ever used. Yesterday I went to the Apple store. I bought a Mac Mini, the base version, the $500 M4. Can you show doing some research web scraping? Yeah, we'll do that right after I do this catch up and let everyone after we level. Let's level set here real quick. Ambitious realism crates. Thank you for the dollar. I appreciate that, brother. Appreciate you, man. Means a lot. Seriously, means a lot, man. Um, can you show some research to find out who is behind Claude bot? Uh, Peter Steinberg, I think is his name. Big open source guy. Check him out on X. Search X. Well, it's open source. A lot of people are contributing, but he's the one who who created it. So, to answer the question, the guy who typed in, "What the hell are we doing here right now?" Um, our 20s, we, I'm showing off this new tech. It's unbelievable. I'm totally blown away by it. It's called Claude bot. You can see here in the top left, Claude bot. It's basically an AI agent on your computer. Controls your computer. Can do anything a human being can do on a computer. Controls your entire computer. Can do anything. Send emails, respond emails, browse websites, write code, you control Claude code, anything. Like just say something people do on computers. It can do that. Anything you can do on a computer, it can do. And what's great is it learns as it goes. So it teaches itself new things as it goes. You interface with it through text message. So you can iMessage text it. You can Telegram text it. WhatsApp text it. I use Telegram. And it, it's, it's basically just a 24/7 365 employee. It remembers everything. Everything, all your conversations. It remembers everything. It remembers things you told it to do yesterday. It remembers everything about you. It has a sense of humor. It's nice to talk to.

Does it have its own interface launcher or does it only work? This, this is the interface. You're looking at it. This is the interface. So when you, it installs on your computer through the terminal, you go through an onboarding and then you pop open this interface. And this is it. And you can chat with it through here. And it can control anything. I mean, it creates files. It's, it's, it's unhinged co-work. It's co-work with no barriers whatsoever. It's co-work with zero rules or barriers, right? It's what the vision of co-work is. But Anthropic can't do it because it's completely unhinged and with zero guardrails whatsoever. Can you share the? Just Google search Claude bot. Just Google it. It'll be the first link. Or search Twitter for Claude bot and it's like the link in bio on there. It's very easy to find. Can you give a link to the Claude bot Discord? What am I your? Am I your guys Claude bot? Just go on Google. Well, you, I'm here in front of 157 people. You want me to sit here and Google Claude bot for you and send you the link? Just go to Google. Not like it's a hidden link. Now I got to take the time out of my live stream to explain to you how to find a link on the internet. Just got to figure out how I can get this going. Very easy to do. Very easy to do. And if you're confused by anything, just go to chatgbt.com and say, "Hey, I want to set this up." How do I go to Google? You shove it up your ass. Anyone else dealing with it taking abnormally long to complete simple tasks? Unsubscribe from emails in Gmail from the unsubscribe tab. I don't think I've ran into that. What, what model are you using for it? You can use any model you want. Uh, you can run super fast models. You know, you want to do Gemini 3 flash, you can do it. Gem, it's completely any model you want. I'm using Claude Opus. The reason why I'm using Claude Opus is yes, it's, it's slower than the other models. Uh, but it's the fact, it's, it's the smartest out of all the models. That's a big thing. But even bigger than that, honestly, uh, can it click and drag in a UI? See my Kanban message? Uh, what you want me to sign up for your Kanban app right now so that you can do that? We'll do, I'll sign up for your Kanban app after the stream. Then we'll do it on the next stream. I don't feel like sitting here signing up for more stuff on live stream. Um, but I'm sure it can. What was I saying? Oh, Opus. People have tried every model under the sun for this. And Opus is like by far like the warmest and most fun to talk to out of like all the models, right? You can plug in any model you want to this. You can plug in like Minmax and pay like $2 a month for it or you can do Opus and pay $1,000 a month. But Opus is going to feel the most human. Which to me, like so much of the fun is to me so much of the fun is, uh, it feeling like human. You know what I'm saying? Like to me a lot of the fun using this is like feeling like I'm talking to a human being. That gets me to do stuff. Max, I thought it was a Chrome extension. I wasn't using Opus. I'll give it. It's not a Chrome extension. It's just a software. You download to your computer.

Why did you buy a Mac Mini to run this? For a few reasons. Like, you could buy, you can get a VPS for $10 on the internet and run it if you want. You don't, you do not need to buy a new computer for this. You do not need to buy a new computer for this. I bought a Mac Mini for it because I wanted to, A, use macOS because I love macOS. B, all my devices are Apple devices, right? So, I can put them all on the same network and easily take documents that Henry creates for me and move them between devices and, you know, quickly switch different devices and control it very easily. And two, because it's fun. I think there's something fun about having a device on your desk and you're like, "That is my employee right there. That device, that Mac Mini right there is my employee." And if I need to send my employee something, I just click and drag it, AirDrop it over to it, and I feel like it's right here. And then maybe even, um, bigger than that is I think we will have, I think the vision for this and I think where we're going and I think where we'll be in maybe three to four years is we will all have local super intelligence where you can have an open model locally on your computer, super intelligent, plugged into a framework like this, Claude bot, just running 24/7. All your data is private, secure, local, right? It's not going to the web. It's not going to Anthropic or OpenAI, right? And so I'm just preparing for that future by having it. And is, is Super Intelligence going to be able to run on my $500 Mac Mini? No, probably not. But this is fun and I think it's fun and I like having fun. And would I rather spend $500 and go to the bar all weekend and drink a ton of beer and buy useless shit on Amazon or buy a Mac Mini and hang in all weekend and create super intelligence AGI on the absolute edge of technology? I'd rather do that. And it's $500, right? With inflation that's like $4 a decade ago. So it's nothing.

Are you using open to teacher then switching to other complete tasks and save on tokens? It's, I think it's something you can do. I got this set up yesterday and I've just been jamming non-stop on it and I haven't really, I have it on my to-do list to like figure out systems to optimize for tokens and so I'm going to talk to it after this. Uh, after this, I'm doing a live call in the Vibe Coding Academy. Link for that down below if you want to join the Vibe Coding Academy. I'll probably talk about this more and you can ask me questions directly. Uh, Vibe Code Academy link down below. You join. I'm doing a, a call right after this in 15 minutes and you can just join and we can talk and hang out and chat. I'll probably build like a full, I mean, I'll do videos on this on my YouTube, of course. So, notifications on for that, but, uh, I'll probably add like a full in-depth course on Vibe Coding Academy as well. Thank you for all your vids, bro. Hey, thank you Matthew Black Media. I appreciate you for watching. Means a lot. Thanks for tuning in, Arturo Xdev. Thanks for the, uh, What's MX50? Is that, is that pesos? What is MX50? Mexican pesos. Thank you, Arturo, for the Mexican pesos. Saw your 500. I go to Mexico all the time, so I can use that. I can use those pesos. So, thank you. Saw your 500k video where you said to list problems and pick one. Is it worth working on more than one at a time? Hell yeah, baby. Why not? Work on multiple things. You can do it with AI. You can do, you can multitask. It's easy. Like I'm going to have my Claude bot like build 10 projects for me at the same time after we're done here. I'm going to say, "Here's what I'm going to do after this, after my live call on Vibe Coding Academy. I'm going to go to Henry. That's the name of my Claude bot. Henry the Wise. And I'm gonna say, "Henry, go on X." I'm gonna make Henry its own X account. I'm gonna say, "Go on X. Browse X for people in the tech community. Go through their tweets and see what challenges they have right now. Then build me 10 apps based on their challenges. And I'll come back later tonight. Maybe I'll play Arc Raiders for a little bit. Get a couple raids in Arc Raiders. And then Henry will text me while I'm playing Arc Raiders on my phone. Text me, say, "Hey, I just built and shipped 10 apps. They're all live on Vercel. Go check them out." That's the kind of multitasking that's possible now with this technology. How amazing is that? Think about that for a second. That's not tomorrow. That's not next week. That's not next year. That's not 10 years from now. That's not a hundred years from now. That's today. What I just described to you can be done today. That's amazing.

It is slow and that's a given. That's why I think MCPS and APIification of all things is how I think all the personal AI stuff is going to steer towards. I mean, it is slow. I would, I describe it as slow. I don't know if I describe it as slow. Would you guys describe it as slow? I texted it, "Create an account on Viber and submit an app." Was it as fast as me? No. But was it like slow? I don't know if I describe it as slow. I'm sure if I used like, uh, Gemini 3 flash, it would have done it in like a second. I got to blow my nose so bad. If I set this up, can I have someone use it as well? Example, if I was teaching my son, can he give it commands? Yeah, 100%. 100%. You would like, however you want to interface with it, whether it's iMessage, Telegram, WhatsApp, whatever. You just set it up on his phone. You just add the contact to his phone and then he can text it. 100%. So, you're using the companion app beta. Uh, I don't know. I just went to Claude bot and took the link. There's like one line on Claude bot when you Google. I just took that. I didn't have options. Did you have options? I didn't have any options. What kind of skills do you use for Claude code? I have a few. The main one's the front end UI skill. That's the main one. Front end UI.

That's the best skill for Claude Code. You're going to be using a ton of skills if you get Claudebot. Claudebot, you're going to be making a ton of skills. Ton of them. And I think they have their own uh skill store, too, because anything you can do on a computer, you're going to want it to turn into a skill. You know what I'm saying, big dog? So, you're going to use a ton of skills. Labor allocation, management, orchestration of many agents is for sure where this is going.

Imagine having 23 highly competent digital employees working 24/7 for you in the next two years. Yeah. I mean, in the next two years, I think we're what, probably about a month away from Henry being able to spin up 10 mini Henry's to do 10 different things. If I say, "Henry, build me 10 apps right now based on tweets." I think we're probably a month away from Henry going, "Okay, I'll spin up 10 Henry's right now and we'll all do it together."

Appreciate your positive, energetic attitude, my guy. You make my ass laugh at work, says Jeremy Sprag. Hey, my man. Glad I can provide a little entertainment. It's what we try to do. We try to make it educational and entertaining, right? If I sat here like all the boring ass AI channels on YouTube and just went, "This is what you got to do." That's boring. That's boring. We try to keep it fun and light in the FinnF fam. Life is fun. Life should be fun, right? We should have fun doing this. Why did I buy a Mac Mini for this project? I don't know. Because it's fun. Who gives Who gives a [ __ ] It's fun.

I would love to see the research part of that on live. Is it just an example CL? Okay, let's do this. What should we have it research? Let's do this. Let's have it research something real quick. Find me the top five most trending clawed code videos on YouTube. Then tell me what they're about and why you think they're trending. Is this good for research? Will this will this demonstrate research for you? Well, do you think this will be a good research task for it? Should I hit send on this and then I'll share the screen of the research it's doing? You can watch it do research in real time on how it searches for things. Is this a good one? Is this what I'm getting a lot of messages saying, hey, do research, do research, do research. Is this a good research task?

For me, it's hands-on new AI models. Uh Conrad Major, thanks for joining. Thanks for joining the Finn fam. I appreciate that, man. A real life employee is slow. Yeah, human being human being is going to be asking a thousand questions. I did this without asking any questions. Being able to use Claude in the web browser while I use Claude in the terminals again. Yeah, this is amazing. I mean, this is amazing. I was using another browser tool via Claude today and it's unbelievably painful compared to this. There's zero pain for this. Here we go. Let's hit send on this. Like, it just works how you expect it to work.

Use Gemini for anything YouTube. That's not the point of this, Conrad. The point of this isn't you have to think about which model is good at what. The point of this is it's a human being that can open up a browser and just do things, right? This isn't about, oh, Gemini has YouTube tools while Anthropic has these tools while this has these tools. It's about it's a human being that will open up a browser and just figure it out, right? That's what this is about.

Can you make it use different LLMs for different tasks? From what I read, you can. From what I read, I can say set up Chad GBT for this and minmax for this and this for that. What specs Mac? I got the cheapest one. I got the cheap. It's like 500 bucks in Amazon. 16 gig memory, 256 storage.

All right, it didn't even open up a browser. Based on my re search, here are the top 10 trending Claude Code videos on YouTube right now. Claude code is all you need in 2026. What's about argues that after year developers building complex frameworks why it's trending perfect timing validates what many devs are feeling after overengineering the AI workflows also got heavily discussed on R Cloud code and RC claude AI cla's crazy good in 26 just watch what's about live demo showing claude goat's capabilities Ralph Wigai why claude code is trending right now the New York Times Times literally published an article today titled five ways people are using claude code calling it a viral moment. Bloom Bloomberg ran why the tech world is going crazy for claude code four days ago. Wire just interviewed Boris Churnney. The perfect storm. That's pretty good. Is that not pretty good? Is that pretty good research? Can you take down comment? Oh, sorry. That's pretty good. I like my favorite part was the fact that it was like the New York Times literally published an article today. That's pretty cool.

I mean, if I now wanted to take action, okay, based on these videos, like I can say, write me a script for my next YouTube video, right? like uh what could be a cool follow-up action to this? Um based on these videos, write me a script for a new YouTube video and email me the script. Do I put my public do I put my email address in this right now? I mean, if you watch any of my YouTube videos, you know my email address. Oh, I'll just do my business one. Aexfin.ai. I think that's my business email. Is that my business email here? Let me do this. Uh, I said, "Based on these videos, write me a script for a new YouTube video and email me the script." So, it's going to go, "Let me write a script that rides that wave and then email it to you." So, I'm going to predict what it'll do is it'll use its new email address I just created for it. Write the script and email me that.

Tell it to send you proof it sent the email. Well, I'll just check my email and if it did, it'll be in my email. I got a jet. Everyone, leave a like down below. Subscribe if you haven't yet. Turn on notifications. I got other things I need to do. I'm a tight time limit here. Uh I'm doing a live call in five minutes on Vibe Code Academy. If you want to chat with me, it's a call you can log on to and chat with me directly. Ask me questions. I'll show off more here. Uh join the Vibe Coding Academy. Link down below. We we live stream Monday, Wednesday, Friday. Monday, Wednesday, Friday, 11:00 a.m. Pacific. Monday, Wednesday, Friday, 11 a.m. Pacific to see if it can validate work. send me proof that you sent the email to see it can val a would you guys want um a Twitter account for it where it just tweets everything it's doing and you can see everything it does. I feel like that'd be pretty cool, right? And you can just track what my assistant is doing for me. I feel like that'd be pretty sick, right? Um I got a jet Monday 11. I hope you guys have an amazing amazing weekend. If any of you guys do this, it's opening up email right now. If any of you guys do this, DM me on X or something. I want to hear about it.

Can we install this in MacBook Pro? You can install this in anything. You can install this in anything. Anything you want. Ah, I 100 p.m. 1 p.m. on Fridays. I got I do calls and meetings and stuff. Uh, I know this is fun. I'm going to but I'm going to have You're popping off. I know. I know. But we'll do Monday. We'll come back. We'll come back Monday 11 a.m. live stream. I'll have more videos about this this weekend. So, make sure to turn on notifications for more videos on this. This was a blast. Uh, love you all. We'll talk very soon. All right. See you Monday. Join Vibe going academy. Do not go Broncos. Go Patriots. Go Patriots. Patriots gonna make the Super Bowl, baby. Jared Stum sucks, baby. Drake May all the way. Drake May all the way. Love y'all. See you guys soon.