0:00 The future of AI assistance is here and it's not from Anthropic, it's not from OpenAI, it's not even from Google, it's from a little open source app called Claudebot which has taken the internet by storm. It is by far one of the most powerful AI tools I've ever seen and so in this video we're going to be breaking down what it is, how it works, the dangers and risks that you need to be aware of before diving in and most importantly the top three ways to start printing money with this before everyone else catches on. Plus I'm going to be giving away my free guide on how to get Claudebot set up for yourself the right way in just a few minutes so stick around around for that all right mark we're now in the javas era of ai and uh we're going to break it down here for you guys not only what claude bot is how significant this is and how it's really an entirely new market opening up for making money and building businesses which we'll get into but more importantly like how to set this thing
0:45 up how it actually works um so you guys can figure it out and start moving on it now because it has really taken off so mark um thanks for coming on mate what are we going to be going through here all right so I'm going to walk through how CloudBot works and what it is. So you have CloudCode, which the world, including myself, I've gone pretty insane over. CloudBot is like a brain that's not necessarily on a terminal. So you don't have to interface with CloudBot on that boring terminal. And you don't have to use something like Coork either. So you can use whatever input you want. It could be a Telegram message, a WhatsApp message, Slack, Discord. You have so many options on how you interface with it. And even though it's called Claude bot, you can actually use whatever language model you want. So it's meant to be air traffic control, where you get the message. And depending on what tools and skills it has, it can talk to you like a chat
1:33 bot, it can work like an agent. And most importantly, you can tell it to be proactive. So if you're Liam, tell me about the top 10 trends happening in AI right now, what would be the most important for my audience, then Instead of just answering your questions or doing the thing like Claude Code would do, it can do the thing, it can be proactive and do something else later or come up with something on its own later. So this is kind of freeing what you guys may be familiar with, the functionality of Claude Code. Basically like jailbreaking it, freeing it from being stuck on your desktop or playing... accessing it through a command line. We can now set up a Telegram bot that we can message directly. So whenever you're on the go, in Mark's case, he's bought a Mac Mini to set this up on. But as we're going to go into, there's other options as well. It's basically setting up your own special version of Claude Co that has access to a bunch
2:22 of files and more importantly, combining these skills in there as well that allows you to access it from wherever you want. And this is a massive leap in the right direction towards the Java's AI that we're all kind of dreaming of. And really, the reason why this is a big deal, if your audience is still saying, Okay, great, why do I care? Why do I care? Because this has gone viral, because this is the first time you actually have a personal assistant that you can truly configure and you don't need NNN, you don't need make.com, you don't need a 100 node workflow. You can literally just, as soon as you install it and you can go back and forth with any AI to help you, you can build pretty much whatever you want. So this is the beginning of the personal assistant era. This is the whole, like, the future is already here. It's just not evenly distributed. Guys, this is the future right here. And if you can make most of it now, there's this arbitrage opportunity where you were going to literally be superhuman for about six months,
3:10 maybe even 12 months before the rest of the people catch on to this. Yeah. So if people look at this part where you just send a message, goes to Claudebot, comes back with an action, you might be like, okay, well, this voice flow thing from two years ago does the same thing. But there's a big difference between this and voice flow because the TLDR is, like I said, it is the air traffic control. So any channel you want. You don't have to do anything fancy. Everything comes out of the box. And this is fully open source, meaning if you want to change what it can or can't do or add a different connector, you don't have to wait for a platform to do it for you. You can do it yourself. And in terms of the tools, it comes with a series of skills. Like everybody knows how to write Apple notes. If you give it its own desktop, it can create all kinds of 11 labs, voices and clones. And a lot of these
3:56 things that I'm mentioning in passing were huge NN workflows just four or five months ago and now it's all natural language based. And the fact that you can even make Claude modular like essentially let's say Claude is out of reach financially either for you or your clients. You can run a local model and then have that hooked up to this Claude bot brain And Claude's just the orchestrator. Claudebot just decides what happens when, which tools are used, and which skills are needed. I mean, the quickest thing you could do to cut your costs is jump over to some of the Google models. You know, like go from using Opus 4.5 to anything in the Google suite. Far, far cheaper than the Anthropoc ones. So there's that. Or, I mean, if you're really trying to minimize cost, it would be the Google side, right? rather than having to purchase the Mac Mini and run the stuff locally. I tried to
4:44 run, I don't know, I mean, what is Ollama and get it to do a run Cloud Code for me. And I've got a pretty decent Mac. And it was just like, it took like 40 minutes to generate like 3,500 tokens or something. So yeah, there is definitely a hardware limitation. So people need to be aware of that. Yeah, it's a good segue. So the ways you can use and set up CloudBots Like we said, you don't need a Mac Mini. The TLDR of why people like myself have gotten one is if you want to have that interface where you can physically see that your agent or agents have their own workspace, their own desktop where you can build, manage files, operate its own environment fully, but you want to separate it from your Mac in your life so you don't risk hackers, viruses, all kinds
5:30 of things going wrong. That's where the Mac Mini is becoming helpful. And also, If you want to use open source models, like Liam said, my computer basically nukes itself if it tries to run something on Olama. So because it's a virgin blank slate, technically computer of its own, you can also run language models a lot more easily. But for most of your audience, the cloud is probably one of the best places to go. You can literally use Amazon Web Services. There's tons of tutorials already on this. And you can hook it up on the free tier and you can use this on the free tier. It just might not be as pretty and cute as me. What I'm about to show your audience shortly in terms of how I have a setup. So you've got what maybe 10, 20 bucks a month or are we talking like 50, 100 bucks a month to actually run this at any kind of
6:17 scale? This literally free tier. You can live on the free tier depending on how much you use it. I suppose it's if you're choosing to use the like the Gemini API, like accessing it through a hosted model, like self, not a self-hosted model, but a provider where they're already running the models for you. And then you're just paying for the cloud space to be able to run the Claudebot itself, right? Rather than running the Claudebot and also hosting the model so you want to stay with hosting the model on the cloud setup as well guys if you are already sold on claude bottom want to get this set up for yourself the right way in a few minutes i've got my full guide on how to do that in the first link in the description it's going to break it down for you so that you can set it up the right way without all the risks plus with that resource i'm also going to be including a more in-depth breakdown of each of the money making methods for you guys so that you can take action on them today
7:03 yeah and there's one more thing um let's say you don't want the mac mini but you still want to rent locally you could also use something like docker to create a container, to run a local model, to also interface with this. So it's very malleable because it's open source and unlike a company telling you how you have to use something, it's completely up to you. This is a matter of what fits the best for you. Now, in terms of where mine lives, so obviously I have the Mac mini here. Where it lives is its own laptop. So this laptop, I'm basically rendering this on my actual screen on my real laptop. It just has access to its own notes. It creates its own folders. It manages this system. So this is technically my agent's computer, which is why it's so malleable because anything that happens here is only constrained to what's happened here. So in other words, the blast
7:49 radius of anything going haywire won't affect my finances, my credit cards, or my files in general, which is a peace of mind if it's something within reach because you can manage everything and you can also manage what's being installed, how it's being used, et cetera. I'm even considering the investment myself now, just so I can tinker around over the next week or so. And once you set it up, it comes with its own local hosted command center. So this is the CloudBot command center. This gives you an overview over how long it's been connected, how many instances it has. If it has jobs that it runs over and over again in the form of what's called Cron, when you speak to it, let's say via Telegram like I do, it can decide, okay, cool. It looks like Mark wants to run this reminder to go check his YouTube competitors and look at their comments using the appify scraper every hour
8:39 or every three hours. And this is where the proactivity comes in because it basically pings its own server and it messages me whatever I'm looking for. And then a part of that picture is obviously skills. So like you have normal cloud skills. These are all open source skills. So these skills that are used things like the Apple reminders, write its own notes, watch for blogs, take a screenshot of its own environment. So it can literally be like, am I doing the right thing? As if it's your actual virtual assistant. Yeah, so a lot of these come out of the box in the setup wizard, which is helpful, right? So this is, I think, a huge leap forward. This is that these come out of the box and a lot of the integrations, like the 11 labs one, which we can showcase in a second. But instead of the clunkiness of MCPs and how bloated those can be,
9:26 these skills now seem to be forming as the new unit of integration or as a feature. And the fact that we'll be able to like browse marketplaces for these and figure out what we want to plug in very quickly to our personal like Jarvis here, I think is super exciting for non techies especially. Yeah, just the TLDR for people on difference between skills and MCP. A skill is like a constrained set of, let's say endpoints or different functionalities from a service. Let's use 11 labs as an example. If you want to clone a voice, if you want to generate something with an existing voice, Those are two different functions. So skill just creates a cheat sheet in plain English with those functions and it tells whatever LLM it is, this is how to use it and this is when to use it. And the core difference is an MCP is kind of always on, meaning it's bloating your context window
10:15 of your conversation, which might push the agent to time out. A skill is invoked just in time, meaning whenever it's needed, it'll be used. Yep, love it. I think that's future for sure. And I think even Anthropoc kind of admitted that MCB wasn't the way forward by making the transition over to skills. 100%. So we already spoke about the CloudBot environment. You can connect it to whatever you want. Discord, iMessage, Signal, whatever makes you happy. A lot of this stuff comes out of the box. You just have to hook it up. If you choose WhatsApp, I think you have to just scan a QR code and it hooks up to your personal number. So one disclaimer for your audience is... If you want to separate your life, which I recommend that you do, if you're not a cybersecurity expert yet, it's probably better to use something like Telegram, especially because the blast radius of something going wrong is probably
11:03 a lot lower. Yeah. And then when it comes to where you can interface with it, you can also do Google Chat, Microsoft Teams, et cetera. So I only allude to that because there is a business opportunity here. If you can figure out how to set this up, comma, securely and use it because most of your clients are using something like that. I mean, just like scratching the surface of how this can be set up as a service, which we'll go into in a second, is a huge opportunity here. And then behind the scenes, just in case you're still, okay, let's say you know CloudCode. So if I say the words CloudCode, then when I say CloudMD might mean something to you. CloudMD is like the brain of how CloudCode operates every session. It's kind of like the values, the holy book. Like priming. Yeah, exactly. Exactly. So the priming here is a bit different. When you onboard CloudBot, it asks
11:53 you questions about how it should behave. So it's fully going the personal assistant angle. Lily asked me, when you're wrong, Mark, do you want me to be direct until you're wrong? Or do you want me to be soft or passive about it? So it interviews you in natural language, and then it stores your preferences in its soul.markdown file. So it's literally called soul.markdown. Then it asks you what it should identify as by name, creature, et cetera. out of the box in 15 minutes. Once you have it installed, it's ready to go. It knows how to speak to you. I told her I like lowercase. I write in lowercase. Don't give me some cookie cutter stuff. Don't say I'm absolutely right. And then we were ready to go. Now, we might have this just after this, but when it comes to building in your workflows to this, so you've got your workspace set up, you might provide it with some
12:39 like market research data or sort of key components and documents about your business. How then do are you needing to create custom skills or are there like simple commands that you can set up? How do you then start to make workflows on top of this? So one tool that I like to use quite a bit, you can use whatever you want. Any form of IDE where you can do stuff is helpful. I like to use warp because it has a free tier. So what I like to do is I just go and I say, hey, can you go interface with CloudBot? I don't feel like building a skill, but can you go and figure out how to use insert X API, let's say, We want to do the HeyGen API to make a copy of ourselves. Go look at the HeyGen API, see what you need from me and see what would it take to make a skill. Then once it creates a skill, it stores it behind the scenes and it tells CloudBot its
13:28 identity and its tools file. Make sure that when Mark says clone or avatar, we use this HeyGen skill. So that's one avenue. Another avenue is once you have it set it up well, you can actually ask Claude Bach to create the skill itself. So once in a while I'll ask it, I have a challenge for you. I don't know how you're going to do this, but I have this voice note from Liam. I want you to take this voice note and I want you to clone his voice somehow using 11 labs. I'm not going to tell you how long tell you where you have the API key, go figure it out. So it will go five, 10 minutes. try a couple things, then come back and figure it out. Okay, so if we pop in, you can see this is my beautiful assistant, Claudette. I actually asked her to make her own profile picture, and I asked her, can you tell me what Liam Otley's last three videos were about? So behind the scenes, I
14:14 gave it access to Appify, specifically the MCP, just to make it easier. It went, kind of scraped it, just turns back to the link. Pausing quickly on that, Appify is a great thing to have integrated... because of its depth of different scrapers, right? So that's just like a very easy general purpose connection you can make to give your claw bot a ton more capability, right? So we can find a LinkedIn scraper if we need to and use that. And it's just all set up through your Appify account. You're going to get pinged for those actors. Exactly. So if you go to, it's a great segue. Go to mcp.appify.com. You can see out of the box, you can just set everything up. Then you can click any one of these. This is what I did. I literally just clicked on warp. It gives me the request that I need to send. Or you could do, let's say, Cloud Code. You just give it this command. You paste that. You tell it, can you go learn how
15:02 to use the Appify MCP? And then you can add whatever actors you want. So in this case, you can just literally go select them, click Save. Then you're good to go. And then... So you manually add those in? Or there is a... Like when it uses the search function there, is it going to search from the ones you've enabled? Or is it going to search the whole... the whole database of actors. You're asking the right questions. So out of the box, I manually put ones in that I know for a fact I want always. So that's where I added them here. And then you just have to enable it to search actors and call actors. And then if it's not there, if it doesn't have, let's say a LinkedIn scraper, it can go and search for an actor that's highly rated and then add that to its stack. That's super helpful. That's how we got these YouTube videos. And then this is the part where it's super cool. I sent it a voice note. just
15:48 to better understand what's happening. And like, I asked it, can you create a clone of the following voice note that I'm about to send you? So it, I basically gave it this voice note of Liam swearing at me for being late to this pod and then listen to it. And then I said, can you use this file to clone his voice and send me a sample of the clone telling me that he loves me? Now I would love to play this, but Liam kind of muffled his own voice on purpose. So it's going to sound super goofy. You want me to play it or what? Yeah, I mean, I haven't heard this. This isn't my bad American accent. He's awful. Hey, Mark, I love you, man. This voice cloning technology is absolutely... That does sound like how I sounded in the voice line. In the future. Hey, Mark. Hey, Mark, man. That's what I sound like.
16:33 Yeah. And then I said, can you voice note me a breakdown of Liam's channel strategy from what you can see? And super interesting that I want to show your audience here is that it started blazing me immediately. Yeah, that's part two. That's part two. But part one is it ran this job that I told it to do to look for my YouTube competitors. So then it found the last few people in the last few hours that had a video around Claude. So I was just like in the middle of it. But here's the part glazing here. All right, let's break down Liam Otley's channel strategy. So Liam is at 724k subs right now. He's absolutely crushing it in the AI automation space. And his whole strategy is built around this concept he created called the AAA model, AI automation agency. We got the idea. So that's, you're asking it for a, did you ask it to give a
17:22 audio summary there or is that your preferences? Yes, send me a voice note. Oh yeah, I can send you a voice note. Break down Liam's channel strategy, but nothing else. Cool. I mean, so my, my questions here is like, there's the difference between like a general purpose assistant that has been given enough tools and access to stuff that it can just figure shit out on its own versus a bit more of a structured one. Like, uh, say like I've just been setting up a, a, uh, core code workspace for myself for like as a podcasting workspace to help me research competitors, uh, research guests, um, do a quick, like kind of snapshot of the market, see what's trending or a snapshot of my competitors and these kinds of core skills, not necessarily full skills, but just like a clawed command that has a markdown file with a couple of things and telling it how to use different parts of the
18:09 readmes inside of it, any scripts as well. So where's the sort of middle ground for this between this completely freeform assistant who's just got the ability to problem solve anything that you need it to and these very structured workflows. I mean, you could even say the old AI tools we used to have where it's like input, input, pass them to a prompt and it comes out. How do you ensure consistency for some of these tasks like Do you need to provide it with an example of what you want this research report to look like in the same way that you do it for the, let's say the podcasting environment I'm setting up? How have you sort of split it down the middle there where you would kind of want the report to come back in the same structure each time? You do want certain things to be included. No, that's awesome. So... Again, this is brand new, right? But the way... I don't expect you to explain your workflow patterns that you got your
18:57 Mac like six hours ago. Well, it's more of like a discussion. Because sometimes you're going to want like, and I kind of want it this way. And I suppose it can help you to create those patterns, right? And create those templates for it to fill out. Yeah. So for me, this is very like on the go. Things that I wish my executive assistant could do without me having to over explain and kind of like direct here and here and here. Let's say like, Tell me about the latest video that came out today that went viral. Like those kinds of things, really good at it because we can abstract those skills. But for me, I have like eight cloud code command centers. So if I'm trying to send an RFP or a proposal to a client, I'm very particular. Like I usually go through three, four iterations until I get it. I don't want to wait because the thing is now that it's on Telegram, your mammalian brain is programmed to wait for like what a
19:44 minute or two before you get a response. If it takes 10 minutes, you start losing it. But when you're instead of in front of cloud code, you know that you have to go grab a coffee and come back before any magic happens. So different platforms for different parts of your workflow, at least for now. AI maxing. It's basically what it is, like plugging a certain need or scratching a certain itch with a different type of AI at every part of your daily life. So, yeah, this is super, super interesting, man, how this rolls out, whether it's like eventually it consolidates into just one thing that you're using or you still have a couple of dedicated workspaces you set up for certain things. Yeah, it's interesting because Because when I give it its own laptop to work on, and let's say there's something wrong. So
20:33 one thing that we're going to talk about shortly is a little Debbie Downers is what happens when it comes to security. So I saw a bunch of tweets saying, you know, this is insecure. This is insecure. Be careful of this. I taught it to scrape X. So then I started sending it. all the tweets saying, can you make sure we're covered on these security things? And then it made its own notes in its notes folder on like the things that were covered for what it needs to implement. So it can even make its own to do list. You don't need an obsidian. You don't need a notion. You could, it has the same access as an employee with a laptop. And that's why all the use cases that need to exist don't exist yet. Man, it's exciting, man. It's insane. But we do have to talk about the elephant in the room, the crab in the room, which is security. So if you go on X, you'll find tons of people warning, sounding the alarms
21:23 that this is actually a picture of people around the world that have spun up Claudebot on a server. So this is the blast radius of chaos. And if you don't know what you're doing, yes, it can be cheaper than buying a two grand Mac mini. But if you don't know what you're doing, There's so many vectors that an IT person or a hacker can use to get access to your personal laptop, your WhatsApp, your funds, your credit cards. So even though this is super cool, well, I'm going to like emphasize this. There are a lot of consequences if you don't separate your life. So even when I created my employee, Claudette, she has her own iCloud account. She has her own brand new email. Everything is completely separate from me. And so what's the, like, where's the danger zone? Is it if you're running this
22:11 locally on your own machine and it has access to all of your stuff on your, on your main machine? Yes. You have different tiers tier one. Let's say you're purely local. So you have a local model, it's in Docker and you're running it locally. The only blast radius is it might nuke all your files on your computer, right? If you, if it's actually on your personal computer, Number two is if you run it and it's using the cloud and you are giving it your actual WhatsApp and you have a bunch of what are called ports open for different services, if you don't secure those ports, people can take advantage of those. And once they're in one, they can find their way to tunnel into the rest of your services or files or whatever. That's number two. So that's mainly like a breach of what you thought were private files, right? Like if you've got, I don't know, API keys or you've got login and password or... I hope you're not tracking those in the MD
23:00 files, but yeah, personal stuff that you don't want. Don't want getting out there. Yeah. And then there's even there's more tiers. Again, I'm not the full expert, so do your own due diligence. But like even with the keys that you give it, let's say you give it the 11 labs API key. Let's say you give it the I don't know, GitHub. A lot of people are using it for GitHub where they're like, you know what? Go look at my vibe coded app while I sleep. Go come up with a series of improvements and work on it until I wake up. Theoretically, beautiful. But behind the scenes, if you give it full access to all of your GitHub, and let's say you're an agency and you have actual client projects, and someone gets a handle for that GitHub, and you've enabled the key to access everything, not just one repo, it could be a disaster. Okay. So scoping those keys is important. Yeah. So just scoping that out. And I have a little bit of
23:47 a checklist here. Again, not meant to glaze over your eyes of the audience. But if you want to deploy it locally, there's a mode called sandbox mode. All I did is really asked warp to talk to cloud bot and make sure we were running in sandbox mode. If you want to self host, obviously local models is the safest way. If you have credentials and secrets, make sure that you put them into secrets managers. You don't share sensitive files. You don't just put a series of API keys and then like an open Jason that anyone can access. And then, You have all kinds of things around monitoring, making sure that you maybe have a cap on your API keys so you don't have $1,000 on Anthropic API ready to go. You've got like 20 bucks and it reloads for five bucks every single time. So there's layers of fail safes. So this is all we know now, but there's going to
24:35 be a lot more because I'm pretty sure someone's going to take advantage of this. Yeah. So Mark and I had a bit of a brainstorm on ways to make money with this. And these are obviously very early stage. A lot of it is kind of arbitrage, but it does go back to the kind of core principles of the AI agency and AI services. So the obvious one here is that with great power comes great opportunity. Oh, I literally made that up on the spot. But that makes a lot of sense. When there's new powers and there's new, there's Java center on a local machine, businesses are going to be, or some businesses, if you can position it correctly, are going to be interested in paying for this. So that would, involve you starting off just setting it up for yourself. Do the cloud host version first, get a feel for it. Once you're sure, you've kind of figured out the
25:22 best workflows for it, how to set them up quickly. You could look to do it yourself, purchase one locally, and then use that as a template to offer that to businesses as well. Small businesses in your area saying, hey, I can set this up for you. The big issue here is the education gap on what the hell this is because you guys are probably already an AI enthusiast and you're watching this to try to just like understand this as the rest of us are. So there's going to be a huge gap between where you are and communicating this value. It's likely going to be one of those cases of where don't really tell them what they're going to get out of it. Don't try to explain all the AI stuff under the hood. Just like I can do X for you. Here's how much it costs. Do you want this result? Probably a personal assistant set up for a founder or a CEO or a C-suite person is the way to go here. Someone who's
26:09 got a lot of... The time is very valuable. And like Mark said, having a personal assistant that can have the ability to access the right things and do little admin tasks for them and offering that as a service. to sellers to businesses or even to individuals, maybe to like targeting entrepreneurs who are busy and want to have more time. So setting these up for them would be a great service. I would also add in set up as a service, set up as a education product as well. You could drop a $97 course right now on how to get the most out of it and set it up. And as long as you have the means to market it, then that's an opportunity as well. So either the service-based approach and then the info approach. The service one's probably going to be, again, it's always like how effectively can you market? If you have an audience already, then it's probably easier
26:55 to sell the information. If you are starting off, it's easier to do reach outs to find one or two people who pay you a few grand for this rather than trying to get thousands of people to buy your $97 course. All right. So the VA agency, not to knock our VA friends, but it's no secret that we're probably not far from the point where a lot of normal VA tasks are genuinely fully automatable and And it's affordable and it's cheaper than the six to ten bucks an hour that you'd spend. So the opportunity here is the assistance in a box, kind of like business in a box. But imagine you have, let's say, a one time setup. That setup could be, again, on cloud. You could literally sell them with a premium, a Mac mini and install it and set it up for them and install the packages. And then you can have... I can already see the pics on,
27:42 you know, how you see like the SMS or the WhatsApp bot farms and it's like, well, the phone's in it or it's like the gambling bot farms and it's all these gambling things. I can already see it, bro. It's like all the Mac minis lined up and it's like my Claude VA agency prints like two million a month or something ridiculous. I can already see it. We're going to start. They're going to come from one of these videos, for sure. Yeah, that's our contribution to the world. But yeah, like the theoretical thing is, If you can figure out how to make this genuinely useful, like everything I'm showing you here, right? I want to pull this from my other screen because I think it's useful. One year ago, almost like to the day, if I just refresh this bad boy and we take a look at this, like this was the hottest thing on YouTube, which is the
28:29 N8N Jarvis where you manually had to set up all these agents, all these nodes. These would break all the time as you wanted to change things. But every time you wanted to add something new, teach it something, you can't teach it something. You can only build it something that it can learn. So we're going from this into a world that's a lot more plentiful. And with that, you can upsell adding skills. You can make those skills across the board. You can upsell maintenance. You can upsell future proofing, all kinds of stuff. I mean, you can use these if you have an upfront setup fee, you charge them you buy the Mac Mini, you set it up and you have like obviously an SOP on, okay, we want the Appify MCP plugged in in a certain way. We want these default skills out of the box. And you've got like an SOP on how to set one of these up
29:16 and then there's a bit of a teething process to get them working on it. You're setting up the Telegram, you're giving them access to it. That's repeatable. And I mean, even thinking about how you can add more value to them on top of that, you can have a, hey, I'm going to analyze your transcripts occasionally. Whether you either create a skill that is almost like a workflow tracer and will actually look back on the conversation, figure out what it used and when, and then be able to bake that into a skill itself or build it into some sort of reusable workflow. So much to build off of here. But I think you guys can see the value on it. And before you know it, if each client is paying $2,000 set up and then a recurring fee, $500 or $1,000 a month, you can very quickly spawn that whole... So the whole room or warehouse full of Magma is running
30:04 for your clients. And I would even say, play on words a little bit, but if you remember, I talked about that soul.md file that contains behavior. What if you got the perfect behavior of the perfect assistant and you started, wait for it, selling soul.md files and being able to offer that personal assistant service, prompt ready to go. People have paid for prompts before. I made my early living on that. So I know they'll do it again. And the last one, sir. The customer integrations here, building custom skills and automations. I think this is like the new SaaS. Basically, this is the in future. We're going to be relying pretty heavily on on these these kinds of personal assistants and the adding of skills. We had, we once had MCPs, we had chat GPT plugins, like all of these are just
30:49 attempts for platforms and providers and SaaS products to be able to hook into the chat interface in an easy way and give access to the functionality that they've built, their backend functionality via the chat interface. Like the chat interface is the new, is a new access point for everything in our digital lives. And I think, we're looking at here is an incredible opportunity and i wouldn't quite say the app store moment yet because we haven't got the app store but this is like is this is this the is this the iphone moment where like the the thing is there we've now got like the open source iphone where you can kind of build things on top of because it's it's so tech pro to try and like the most tech pro thing to try like call it a the x of y but where you don't have to have if you're wanting to make a little little startup now for example The amount of people who need to be able to pass on a YouTube
31:35 URL and download the MP3 and then get the proper transcription, a high quality whisper transcription of it, not just some shitty YouTube captions thing that has no punctuation and barely knows what was said. I want a high quality transcription and I'm going to pay someone $5 a month, $9 a month. And before you know it, someone who has one of these Claudebot accounts set up, And just like there's going to be agents as you pop up to, like in the previous example, these really helpful but somewhat more difficult and naturally not free. There's some sort of like money that needs to be exchanged hands to make it worthwhile. These opportunities are going to pop up everywhere. And I think you're going to see some businesses that just explode overnight because as more people get onto this and it's like, oh, this is the go-to thing for downloading YouTube videos so that my agent has ability to do that. Yeah. You
32:24 get to strip away the need for a fancy interface and UI and front end design and all of that crap that comes with SaaS. Just focus on what is the functionality that I want to really focus on? How can I make that as valuable as possible? And just make it available to some agents like this, whether I'm getting listed on the marketplace, doing my own marketing. But there's a real, real opportunity here, finally, And I think it's massively reduced the difficulty bar for getting into SaaS and that you just focus on the backend functionality, make it available to agents rather than having to build some fancy skin around it. Yeah, absolutely. And I would say one more thing to think about is when it comes to the skills specifically, those are going to become expectations. And to me, this might not be a moment. I don't want to cap and say that. But what we could say is is this is the beginning of a brand new expectation from the consumer. Meaning once this
33:13 really gets into mainstream, it's already going viral. The expectation of what a personal assistant is, is gonna be much higher. So when Siri uses Gemini, I think this year they're gonna like work together. And next year we have laptops. People will start asking the question, why is Siri just a voice thing? Why can't she interact with everything and do X, Y, and Z? So expectations from the consumer will change from this point onwards. I mean, when you see this being rolled out into all the equivalent Apple attempting, surely if a solo dev can cook this up himself by what's the sound of what happened here, then Apple can make some sort of decent competitor. But going back to what I was saying about these sort of like little skills or functionalities that these kinds of agents are going to need, once it becomes
34:03 rolled out via something like Siri, And those same things can be integrated where the Apple is going to force you to do it through an app on their phone, or they're going to say, Hey, what, you know what, we have to allow people to go outside of the Apple ecosystem. Um, we want our Siri for our Siri to be competitive. It needs to be able to hook into, we can get permissions for them to add in their own third party things that are kind of out in the, out in the ether. I don't know how Apple will feel about that if I'm honest, but dreams are free, you know? A hundred percent. Um, so yeah, that's a bunch of ways to make money with it. I think, uh, if you guys have any sort of business sense, you'll be smelling a bit of opportunity here. Not only just for us in the near term to brush up on this and to get our own one set up and see how far we can push it to max out our productivity in this new era of AI that we've seemingly stepped
34:48 into here. But also looking at how, if you're an agency, you can roll this into your services. Just like, few days ago, I was telling everyone how teaching companies how to understand core code and set up their own workspaces as a service was a thing. This seems to be like the new frontier on that. So yeah, it's just, it's so great to be in this phase now. I think last year was a little bit slow, but I feel like we're absolutely like cooking this year on new releases. And I'm so excited for what's to come this year based off what we've had already. Yeah, it's already not even the end of 2026. And it feels like it's been a year. So especially on the education side, You could have taught cloud code alone. Then a week later, you could teach cloud code and co-work. And now you could teach cloud code, co-work and cloud bot. And then you could probably keep stacking on top of that. So the education piece is huge. And like, as
35:36 this stuff runs off further into the distance, companies are still in the same spot. And the amount of like monetizable space between where the tech is and where these companies are, the automation that was available in sort of mid 2023, that, like millions of businesses still need is still sitting there all the way up to this stuff at the very cutting edge of what's going on in the AI space now. So what a time to be alive. What a time to be in this space. So Mark, I really appreciate you coming on and sharing your source. as always and so if you guys want to check out Mark's channel he's always pumping out some great content on Claude Code well Claude Bot as well now I'm sure and anything at the cutting edge of AI so I appreciate you brother thank you for coming on thanks for having me again cheers that is
36:23 all for the pod guys like I said if you want to grab my full setup guide for Claude Bot so that you can do it the right way in just a few minutes without all the risks that would be in the first thing in the description and inside that I'm also including a breakdown of each of those ways to make money with a bit more information so that you know how to take action on it right now and if you want to learn more about how to start an AI business by selling services and products like this. You can check out my full guide on starting an AI business, which is available up here. That's all for the video, guys. Thank you so much for watching. I'll see you in the next one.
