[00:01] obviously the cutting edge of AI right now. It's basically Chat GPT on steroids if you gave it an entire computer and then other accounts to just do things. Opens up Google Chrome. It searches the web. It looks at your exact channel,
[00:15] pulls the transcripts from each video, analyzes those transcripts. Hey, you're of videos. Here's the performance. Here's the numbers that I see. It has a whole different level of motivation. Well, most people spend hours every week
[00:29] trying to figure out what content to make next. Well, our guest today has solved that problem. He built a system using Open Claw running on a Mac mini. Igor Pogamy, he's got his own YouTube channel with over 400,000 subs, 30
[00:42] million views on his content, and he's also an AI educator and expert. Now, you might see headlines of how this is taking over the world and AI has gone
[00:54] going to explain everything you need to know about that, whether you're more advanced or just hearing about it for the first time. So, you can apply this in your own life. And what he has set up this AI autonomous robot to do is to
[01:09] monitor the landscape of the news to score stories based on his own taste and then to send him notifications that he can then turn into videos. And so, the implications of this is pretty wild. And we're going to be breaking
[01:24] down the whole thing. So, Igor, I'm curious. Orient the audience. Like, curious. Orient the audience. Like, explain what even this like Open Claw Open Clawed world is. I'm seeing YouTube shorts. I'm seeing Instagram reels of
[01:37] crazy, >> Yeah. setting up A Gentic AI on it, and it's a lot of buzz right now. Orient us for those that are new to what's happening in the world. Oh boy, this conversation, Sean, is going to be
[01:51] here. And let's get right into it. Let me Let me just tell you. So, Open Claw, heard of it. But most people don't experience it cuz it's kind of insanity to be experiencing it and to be working with it. And what I
[02:06] capabilities of the results you get from it. It's more like the idea behind this thing, okay? So, in simple terms, like what is Open Claw even like practically? It's basically ChatGPT on steroids if you gave it an entire computer
[02:23] and then other accounts to just do things and it never asks for your permission, okay? So, just that it's it's ChatGPT that and it kind of has this loop inside of like you tell it
[02:36] example. It's one of the first things I tried and one thing that works really channel. If you just do it that it on YouTube, Sean? What would happen on
[02:50] ChatGPT, Sean? If you if you tell chat like, go look at my channel. What do you What do you think? Yeah, I mean, it would look at it, but it can Yes. >> might even if it does pull the website,
[03:03] it's kind of probably give you a shallow answer. Exactly. So, it would probably would find it if you give it the name. If you give it the URL, it'll open that up. There might be troubles with like YouTube blocking it or something
[03:16] sometimes depending So, that might happen or it just pulls up like articles on Google about your channel and not your actual channel and then it here's info on your channel. If you tell Open Claw go Open Claw go look at my
[03:29] channel, what it does is it opens up Google Chrome. It searches the web. It looks at your exact channel. Then it opens up the 10 videos, depends. There's There's like variants there for sure.
[03:42] Then it pulls the transcripts from each video, analyzes those transcripts, pulls all of that like into its file management system, and and then it has this type of videos, here's the performance, here's kind of here's the
[03:57] numbers that I see." It doesn't see the back end if you don't log in, but it it just has a whole different level of motivation. It's like chat GPT on Adderall, you could almost say. Like it just goes. That's interesting.
[04:10] And then the I guess the big idea is for those that are more sophisticated with AI, they could go do that manually, but even that, like the fact this is happening in seconds now once it's set up where I can go pull a transcript,
[04:23] might only take 30 seconds, but if I'm going to pull five So it is just it can scan the whole channel, pull data off the pages, look at the transcripts. It also could do that continually while you're sleeping, right? Exactly. So this
[04:38] starts getting people comfortable with this idea of like an AI that does. But then it goes further because one, you can schedule tasks to happen at certain Open Claw. So you can basically say like
[04:52] channels and pull all new YouTube video transcripts and put them in your folder with my video transcripts, right? And then you could Automatically. So this is one important modality that you that you can kind of
[05:05] like schedule tasks out. Another one is because this is installed on a standalone computer. Usually people use Mac minis, sometimes people host them in doing it, right? You can kind of for 20 bucks, you can get a VPS. There's a lot
[05:18] technical, there's certain disadvantages to it. There's a reason why people use if you can afford it. Like it's just smoother, it uses Mac apps and stuff. But the point is it just runs on this Mac mini. You can schedule tasks out and
[05:32] it's very motivated and it does things and ultimately you're going to start plugging other apps into it and giving it more access. So like first thing I did was give my Open Claw its own email account. Then I gave it a Gmail like a
[05:46] the way, that's that's a little tip of like it you will get blocked but if you give it just a a Google account, you can use Google Docs and an email and has Google Docs and Google Sheets. Then I give it access to my
[05:59] Notion API, so all of a sudden it could create databases, it could edit pages that I give it access to and more. And all of a sudden, as the name suggests, you kind of have this claw that just goes in and grabs things and does things
[06:12] goes in and grabs things and does things for you with what's essentially ChatGPT behind it. And is is the LLM behind it actually Claude or something entirely different? It's Claude. So usually people get the best results, also that's
[06:24] the one that I prefer, is Opus 4.6. And here's the issue, like if you really go you spend like 6 hours on there and experiment and try different things, all of a sudden you paid $100 or $200 for that day of Claude Opus 4.6 usage. Cuz
[06:38] doing this, they're basically running credits or tokens. Yeah, this is the big one way around that, there's a certain trade-off, but you can use what is called OAuth, which is basically just using your Claude account to log in and
[06:53] then running off of those tokens. It's technically against their terms of use, people still do it and also you have the big trade-off that you get a smaller context window. For anybody not familiar, every AI, every LLM like
[07:05] of like memory before it gets amnesia, right? So memory before it gets amnesia, right? So in ChatGPT this is 256,000 tokens, which is I believe it's roughly like 180,000 um, words.
[07:20] Now, with Opus 4.6, it's a million. So it's it's a lot. You can fit like books in there. And this is really powerful because as you talk with it, like imagine you you work with a coworker and they just forget what you say every 5
[07:33] minutes. Like that would not be a very productive relationship, right? Um, but I would love to I would love to like explain more how I use it on the talk about like why it's happening now and why this wasn't happening at this
[07:47] things really changed. >> Yeah, I'd say buckle up. This is obviously the cutting edge of AI right now and I am I want you to hit both of those things. I'm just curious how much are you spending do you think per day
[08:01] then in tokens on Open 4.6? >> I think it's important to distinguish between things that are already running and stuff where I'm just experimenting, right? Got it. So right now relating to my YouTube channel, I have two workflows
[08:15] share here today. Um one of them is a research workflow and another one is a packaging workflow. And of those just running, I think it's the research workflow runs twice a day and the packaging workflow runs twice a
[08:31] week. Um that in total runs up a bill of like $100 a month roughly in API cost. So it's not that it's not that tragic. It's like $100, $150, something like that. So and I mean and you're you're pumping out
[08:43] a new show on your YouTube channel weekly. That's what the research is you know, master packaging or get ignored is one of the phrases we've been saying. You want to stand out on YouTube this year, you got to master packaging.
[08:56] So and you you are established. So um You're reinvesting revenue in better content. And so $100 to $150 a day and besides that a month. Um a month, thank you. $100 to $150 a
[09:12] month and the cost of the Mac mini. How much did the Mac mini cost you? It was like $600. You just get the lowest end one. Yeah, you wanted to say why is this happening now before we go deeper into
[09:25] your how you set up your YouTube strategies. So I I also want to say like like Mac buying Mac minis and running this stuff wasn't a thing like 6 months ago, right? And it's not the Mac mini that makes it possible. The Mac mini is
[09:38] sort of just the the the body if you want, right? It's really the the mind behind it, which is the Opus model that made this possible. There's like a turning point with new Opus model coming out like 3 months
[09:50] ago, Opus 4.5, that just it just had better memory. It had better what we it actually listens to what you tell it thing. And also it had better agentic tool use
[10:05] what we call orchestration where it can like keep a thought for an extended amount of time and not dilute it, right? And chat you'll sometimes notice if you time, you're like, "But I told you this. Like why don't you remember?" It's like,
[10:19] "Yeah, but you also told it another 30 things." And chat is sort of lazy in that way, you know, it kind of just just moves beyond it. So, it's really this then, you know, this is a completely open source, as you said, bleeding edge
[10:32] the internet. You can just download it completely for free. All you need is a device to host it on and then you pay the API and then it just does things. Do I recommend it for most people before we get into the practicalities?
[10:47] Hell no. Big all caps hell no. I think there's more user-friendly way ways of doing what I'm about to outline and especially over the past weeks and months Claude has been developing their
[11:02] recommend. I'm not affiliated. It's just the most user-friendly one. This is app called Claude Co-work. Claude Co-work. Not Claude code, Claude Co-work that attempts to be the user-friendly AI agent and you get many
[11:18] things out. It also runs on your computer that you need to leave on for it to work. It can access your files. It can do things. It's not as unhinged as limitations and less customization, but you can do a lot. And what I'm going to
[11:32] what we're going to point out here, packaging workflow and the research workflow, you can totally replicate in in Cloud Co-work. And that's what I would recommend. But also that, before we get into it,
[11:45] like probably also important to note that at the Advantage we have this like methodology of teaching AI and it's it's you don't dive into these things right If you barely created an account and you're new, this is interesting. I think
[11:59] it's amazing to know about, but there's a big difference between information and implementation. You can be informed, but you shouldn't level. I mean, just think about it, Sean, like like just like in school,
[12:13] you wouldn't put a 12-year-old into a master's university program keeping up, right? There's certain levels that you need to like finish high school first. How about that? Then you pick a college where you specialize in a
[12:27] certain area. And then you move into a master's and then you move into a PhD. don't go from high school to PhD and then act all overwhelmed, right? It's not a thing. Same with AI. It's not as long of a journey. It's not as
[12:40] journey. So, I'll keep it super brief and then want you to know let's start from the top as we're already talking about Open Open Claw, all of this Agentic stuff, Cloud Co-work, all of this stuff,
[12:52] Top of the pyramid. And people don't start there. Like nobody's born with abilities to like run Agentic AI on their computer. Like that's not a thing, learned. Now, what are the prerequisites to that?
[13:06] What does the journey to that look like? Well, there's as you might already level two and level one. And people usually start at level one. It's using AI intuitively, using it like a fancy Google, getting email drafts
[13:20] done, summarizing things, designing a meal plan, a workout plan. ideation. All of this like intuitive use that people start using it for. There's nothing wrong with that. That's an
[13:33] like that. But what I'm saying is there's more. to this and the crucial the crucial level the crucial level is the one in It's where things start getting personalized. It's where you start
[13:48] bringing yourself into the conversation. Like you could also think about it like building a relationship with the AI. Like how much is somebody going to know about you if you just met him? Right? If if they don't if they like
[14:02] seen you before, they've never met you and you just met them. You say, "Hey, my to know. They're going to know what they took it from their senses and that your is a person that you have a multi-year relationship with going to know about
[14:16] you? Well, more, right? Obviously, right? And difference cuz it would be annoying if I met my friends and every time I would is Igor." Now, it runs deeper than just your name,
[14:29] obviously. They know what you care about. They know what you really of. They know what kind of foods they should never like bring to your place personal relationship, but even that relationship with the AI you want to
[14:43] deepen it. So, level two is all about it getting to know you and building a system where it has some of these context. ChatGPT does this automatically from memories, but ultimately you want to give it information like your goals,
[14:58] um why these goals matter to you, very important, like a truthful answer there. Your identity, your experience, um what tools you work with. If you're tasks, you want to give it things like what is the USP of your your company, um
[15:12] what is your mission, what are your values, what values do you hold, right? intuitively know about each other if we have a relationship, but the AI doesn't. can ascend from level two to level three, which is agents, because let me
[15:27] tell you Sean, like once you get into this stuff, once you start messing claw, the big differentiator between people making this work or being like, "This is crap. It's not working. I just
[15:39] spent $600 on a Mac Mini and a weekend of frustration, and it's not doing what I wanted to do." The big differentiator there is the context that you give it. got to progress from level two, make sure all your context is there. You
[15:53] understand follow-up prompting and the power of seeding it with the right three, and all of a sudden just go faster in the direction that you already made work at level two. Amazing. So, we have a big promise here. We're going to
[16:07] break down your actual YouTube workflows, how to, you know, not just discover winning vi- video ideas, but even go viral using this workflow. That you don't need to buy a Mac Mini to do. And you're also
[16:19] our current AI setups. You're doing it in the fancy way, the advanced way, but But I love how you just broke down the three levels of AI, and it really reveals even personally, I'm like, man, I'm barely making it into level two, and
[16:34] channel, my business already. More Dev, because you're actually hosting a summit coming up called the AI Advantage Summit. Hosting a summit with Tony Robbins, Dean Graziosi, a whole
[16:46] bunch of incredible guest speakers. I think you only do one of these a year, and this is going to be a really cool three-day event, entirely free. Um if you're listening to this, you actually can register for free at learn
[17:00] Um and we'll put a link in the show Igor, break it down. I mean, I want to get into this next stuff, but just really quick, what is this summit? You talked about the three levels of AI.
[17:12] number one, I feel like we all know, like, shoot, we're getting left behind But then we're also frustrated because we're like it's kind of too much to learn, it's overwhelming, what actually matters, what what um is not worth my
[17:27] time, how do I keep up? Uh break down why this summit is important for right now. I think I think that frustration is legitimate. We we all feel and we all It's you know, like winning math competitions and remote controlling
[17:42] bought a Mac Mini. What's going on with that? Like the it's legitimate. The progress is there and the urgency is very real, you know? I know a lot of like across the internet like people usually
[17:56] oh, you're getting left behind. This is a game changer. But like legitimately like the the world is changing and we also had to like we didn't we weren't planning on doing this summit now in spring in April 21st to 25th.
[18:09] but we're like this happened now. This this moment happened now. People are getting lost in in all of this madness the summit is there for is to give you a clear road map, to show you where you
[18:22] are, to show you the reality of things. And we have incredible speakers with Tony Robbins, with Dean Graziosi, with myself showing you different aspects, but all focusing at the same thing. Of reducing the overwhelm and showing you
[18:36] actually gain more leverage for these tools. We all know they have the potential for that. But you know, one YouTube video at a time, it's really hard to piece together the full picture and in 3 days completely free summit,
[18:50] we're going to focus on this one thing. What's actually happening and how to give you your time back with the use of AI. That's it. April 21st through 25th, completely free. We'll focus on the systems in this progression. So you can
[19:04] going to unpack so much more in this episode, but just a heads up, um learn down below to get registered for that free summit. Massive edge in today's world. Okay, so we've been talking about how this
[19:20] YouTuber uses open clawed and never run out of video ideas. And you've got 400,000 subscribers, you've generated millions of views, and you're using AI to support you in a weekly workflow. Let's break this next
[19:33] section into two parts. Part one is how you set up open clawed to do this for could do this with any AI tool and essentially still capture the same idea. Now, you told me when we were doing a pre-call that you named your AI Alfredo.
[19:49] Is that Is that true? That's the name. So, break down the Alfredo system. Okay, so how's this work? Okay, so we'll get into how to do it into more user-friendly ones, but like basically with Alfredo, like I bought a Mac mini,
[20:02] I booted it up, I, you know, didn't log in with my Apple ID to make sure and and tutorials on the internet, right? So, this isn't black magic. You basically just boot it up and once you connect it to the internet, you download um this
[20:17] open-source repository. Uh it's one of the most popular, if not the most popular repository over the past months. You just Google open clawed GitHub and you just download that code onto your machine and you install it, okay? Now, I
[20:30] step-by-step video out there cuz you do have to like open the terminal and then access the file that you downloaded in there. But, it's totally everybody, even non-technical people, if you, you know, you just follow the step-by-step and
[20:45] and it's like running. You have to connect it to your uh clawed API key, so that is one thing. So, you have to go to console.anthropic.com to add your credit card and get what they call an API key, which basically is
[20:59] just a connector. It's just a cable that connects the clawed models to your Mac And all of a sudden, you're up and running. You you have something that looks like ChatGPT roughly, right? It's a chat interface. But you can all of a
[21:13] sudden ask it to do things like, you know, go and, you know, open 10 tabs and do these things. And it can remote control your browser. The capabilities tasks that are going to be running for 20 minutes instead of just 20 seconds.
[21:28] of these tools. But really at a base level, what once you have it set up, can interact with. Matter of fact, let me just let me just show you what it looks like here on my machine. I'm in Miami right now. My Mac mini is running
[21:42] back in Portugal. I just have TeamViewer installed here, and I'm going to be dialing in. So, let me I want you to follow along. Let me um let me share my screen here. And so, if you're on the audio podcast, you can
[21:55] check out this uh part, which is super powerful. But also, you can just keep listening along and I'll narrate some of the insights. So, there you go. Um So, so this is this is really the base
[22:07] looks like ChatGPT, but all of a sudden like you're looking at Alfredo down zoom in a little bit. So, >> [laughter] >> there he is. That's Alfredo right there.
[22:19] >> got you created like a avatar. Alfredo's like part of your team. He's working for your company. Yeah, 100% teammates either love him or hate him cuz he's relentless. He works 24/7 and and he replies to their emails with a little
[22:31] bit of his Italian humor, too. Amazing. And so, you you did two things he's doing is research. So, is that you have that plugged in here? He is a research assistant that a couple times a week is or a daily is pulling stories?
[22:47] it, okay? I'll I'll narrate this so in here and asking him, "Tell me about the research that you do for me." Um I
[22:59] uh um YouTube um YouTube uh podcast episode where we are talking uh podcast episode where we are talking about you.
[23:11] know, I could describe it to you, but let's see what is, um, forgot to mention is that one beautiful thing is, and there's some error. Let me make it work. Um, but one thing that is wonderful is
[23:25] your Telegram or your Messages account. So, you can really just connect it up and send and that's what I usually use. So, right now I even I didn't have this open because, um, usually just do that, but Alfredo
[23:39] into a bit of a hiccup there. There was a problem and I couldn't open the dashboards that I usually use. Well, usually I use it for Telegram remotely, but you know, because this is Alfredo, my open claw, and he has skills to pay
[23:54] him to build a chat dashboard from scratch and there it is. So, this is Alfredo's custom chat that he spun up cuz we ran into a thing and I don't have time to troubleshoot it here. We're recording is this is Sean's time. This
[24:07] is valuable. So, so here it is. This is Alfredo. So, let's get back to the last point and let's actually ask him, "What do you do for the research workflow?" Um, I am on, um, podcast interview.
[24:34] team support. Think of me as your signal for from noise filter in the AI world. More details. So, that's that's essentially the essence of it though.
[24:46] normally I'd give you a full research brief. Um, tell me about the scheduled cron jobs for research. Cuz that's really the core of it. I set
[25:00] so-called cron jobs. >> job? essentially. Okay. Cron means like chronological or what? I'm actually not abbreviation for something. Okay. I'm not exactly sure.
[25:14] I'm not exactly sure. >> I can look it up. okay, have a have a look quickly. Comes from the word Chronos. A cron job
[25:27] uh are managed via crontab and it's a schedule. So, yeah, Chronos makes sense, which is a Greek word for time. Exactly. So, so at its essence it
[25:39] describes it here to us, okay? So, um it monitor it monitor certain Twitter accounts, Reddit communities, major AI sources like different blog outlets, different newsletters, and everything passes through this filter of can a
[25:52] normal person actually do something with this? If yes, I capture it. If not, I skip it. So, that's what it does, okay? And you started out. So, what the what listeners can do is be thinking about I mean, modeling your entire thing for
[26:06] setting this up cuz this isn't just any old average prompt. It's a whole system that's now running for you. Um did you select the Twitter accounts, >> 100% All of that is manual. So, I already had those in advance when I came
[26:20] in. I knew what kind of accounts I like to follow. I knew what kind of newsletters and you know, um the blogs are the the easy ones cuz it's like obviously AI niche, you know, I want OpenAI, Claude, Gemini, all the big
[26:32] players, the big news outlets. With Twitter, it depends on your taste. this like web scraping piece, this is not nothing new. People might be like, okay, like Igor, like, great. You scrape news together into a Notion database. It
[26:47] gets placed there automatically. Okay. The key is actually this AI layer on The key is actually this AI layer on top, where you instill your taste and your preferences into it and it goes through that filter, right? So, that
[26:59] that's something that no scraper can do for you. You're using these models to do that. And I'll just briefly show you without showing you the whole thing because there's really a lot here. And this this folder gets messy, right? But
[27:12] this is basically the root folder of my my Claude. There's a few things here worth pointing out. One of them is kind of this um soul document where that's essentially the soul of the agent. So, you're not a chatbot, you're becoming
[27:26] someone be generally helpful, not performatively helpful. Let's give the how it's set up. There's certain boundaries and stuff. So, there's an identity, soul, agent, and business. So, you're showing us on
[27:39] screen the files that are driving the Yeah. bot on your Mac mini. Exactly. So, for the identity, as I said, I called it Alfredo. You know, it has a bit of a creature and a vibe and an emoji that it
[27:55] likes to use. It's Italian-American AI with a soul, not a chatbot. A cons- consigliere consigliere or whatever. >> [laughter] very simple, but what matters is as you
[28:08] start setting up these workflows is what I refer to as DNA. And DNA is like context and information that is persistent over time. Right? So, for example, like our names, we get them at birth and for most people
[28:22] right? Um so, that's like persistent context. Also, our goals, they tend not to change day-to-day, right? They might change over the over years, but usually your goals are pretty static. Like in in
[28:35] people will be like, "Okay, I need to make a name for myself and make money and and you know, find a partner, etc." Like these goals, they pretty consistent until you do and then they change. But all this consistent context I put
[28:48] into what I call DNA files, and then these DNA files, they they're working or not working, as I said, the level two concept. And when it comes to this research folder, and you will see that it's like scanning every single day
[29:02] for the past few months. So, if I just look at today's research, I mean, it's this ran once today. We can kind of just look into this research file that it created and see what came up here.
[29:14] That's cool. So, every single day there's rejected, so your criteria, your taste, there's the rejected stories with that research, and then there's the, I would imagine, accepted would be other one would be the research that it would
[29:26] >> Exactly. So, I have I said I don't want duplicates in there. So, first of all, it looks like at the and it uses the Notion as the source of truth, so it can think of it as an Excel sheet, and it sees, "Hey, this story already
[29:39] exists, so let me skip it cuz we already covered it." So, it rejects those. And then the other ones is the stories that it decided to add. Now, this is just happening in this folder, right? And I like showing these files because this
[29:52] really gives people a feeling for what this actually is. It's like text files this actually is. It's like text files that are linked to a LLM, like ChatGPT, and the access to a whole computer and the autonomy to actually operate it.
[30:05] overview. And here's, as you can see, these are the qualified stories. And as you can see, there's a rating system. I also have a velocity value, which is manual researcher. If it finds it in multiple sources, it's a high velocity
[30:18] thing. If it's just in one source, it's low velocity. So, it can also show you. it kind of like sends it to my Telegram if it's above a eight and a high velocity, right? So, I So, it's essentially a text message. It's
[30:31] notifications for that on. So, I just get a message from Alfredo saying, "Hey, there's a new story that is like a nine out of 10, and it's high velocity. You should check it out." And usually, if you set it up well with a lot of
[30:44] detail, it will actually be accurate. So, that's crazy. So, the big idea is you're having a custom programmed AI agent text do a ton of heavy lifting
[30:56] AI agent text do a ton of heavy lifting and filter only to a scoring system of nines and 10s. You get a text message. Oh, there's a story. And then you at that point you're strategically reacting. It interrupts
[31:10] doing. But what you realize, especially on YouTube, it's not just not just the rating of the story itself and does it fit your criteria, but the velocity. And if the view velocity is high, the return on investment of you sitting down and in
[31:24] workflow. We'll make sure that's linked up in the show notes of the podcast. We just sit down and are quickly record using a software this news story. And then you capture the upside because you're getting it out
[31:38] I have that right? Exactly. And then the beautiful thing is like because you're easy to just ask Alfredo like, all right, that's a good story. Can you prepare a quick like five bullet point brief for me to record this? And before
[31:51] you even sit down at the computer, you already have, you know, chat GPT, but my style doing this for me. So, it really is a research agent and then also a script writer and really kind of a news brief
[32:05] writer and it does all of that, which gives you further I got the text message. You prompt it. Now you've got an outline. And do you kind of maybe you pull up a few screens screens to share. But if you're just going to do a
[32:17] 15-minute story, then you just use that as kind of like a checklist for what you Exactly. So, >> that it prints for you basically. Yeah, so basically, I like I'm not a big fan of it scripting for me. I like if it
[32:31] great. Like if I'm doing scripts, I always turn something that I already have, like some of my thoughts, my writing, or my speaking into a script. But basically in this case, I would usually just use it as an outline writer
[32:44] and do that. But there's one more piece that we kind of like missed here cuz you in Claude CoWork and you know, let it go out, schedule a task five times a day to go out and look at some sources. It get That's That's not hard. That's very
[32:57] figure-out-able for everybody. I think the key is really this power of context and the rating system, right? Because everybody might be able to go ahead and like set it up with certain news outlets and then you're getting
[33:09] like 20 notifications a day every time they share something in your niche. But then very quickly, like a day or two, it's going to turn into like Like I don't care about every story. I care about the nines and the 10s, right?
[33:21] That's really the key here. And I do that through um the DNA file that I alluded to. Um and the DNA file looks something like this. Uh it's basically um rating DNA and then I have definitions of what
[33:34] these different different uh levels look like. So, I don't know if you can kind of read this. I'll zoom in a little bit. But you can see I give it a mission. I tell it about the exact process and then I go into the rating scale, right? So,
[33:46] right? And I have examples there, like when GPT-4 shift uh launched, that was a paradigm shift. Also, when Project uh first were introduced, to me that was a paradigm shift cuz I'm a power user of those. ChatGPT voice mode, that was like
[33:59] our most viral video of the year cuz cuz everybody was like, "Oh my god, is it emotional intelligence now?" So, always like with examples. this took a long time to develop, you know? Sure, it was drafted by AI, but
[34:11] you bet that every single one of these examples has been hand-crafted by me and actually represents my taste. Cuz then it's like I get it right once and the yeah, then it goes down to what would be
[34:25] a huge win and what the characteristics of that are. So, again, the heavy lifting of it, this is the difference maker, is not just overloading you with more information. Actually being a filtration process at a
[34:37] the output that lets you focus on what matters most, which is then also why your YouTube channel is thriving because you're distilling all this complex information into something simple, which tells us that
[34:50] basically all of us should be leaning in to the power and potential of this. You mentioned earlier, there's three levels. You don't want to skip straight to this but this is where we all want to end up. And YouTube content creators should in
[35:03] 6 months, I'm saying, have that level two cloned themselves, maybe in the next aggressive. >> As soon as possible. Sean, like like as like the the level two, like building these context files and starting to work
[35:18] with them, like honestly we we run these paid programs, right? And like it's the hours. Also the YouTube videos where I covered that concept, it's like like shifting your thinking about it and not being like you don't need more
[35:33] tools. You don't need more features. You don't need the latest released. You need matters to you and you need to get that on paper, put that in a markdown file it, right? And you can do your own version
[35:46] of it. Yeah. So I want to come back to that in just a second. So there's a want to cover. Um and I want to get into you explaining don't got to get the Mac Mini, I don't got to set up all these files. Even as
[36:01] we were we were recording, we spent time troubleshooting. No big deal, but 99% of want to do that and and worry about this code and stuff. So I do want to land the plane with giving us some practical tips to use
[36:13] the LLMs we already use to get some of these benefits. But before we do that, I need you to cover the objections. And I think the the unspoken about thing here that we haven't really hit is well, what if
[36:27] Claude code uh claw, you know, claw bot like people account information. They're trying to actually have a personal assistant or Or like it's going to hijack the whole computer or you know, I think a lot of
[36:41] information am I giving it? Can you unpack some of those things? I mean, I think all of those concerns are very real and there's a very clear philosophy that I and everybody else using this kind of follows.
[36:54] information. Do not give it your credit card. Do not even give it your Gmail account. The reason people are buying Mac minis, like if you buy a Mac mini, don't set it up with your Apple ID. Sandbox it. And what that means is give
[37:08] it give it separate accounts, right? I create its own Google account for it. I create its own email address for it. I if I give it access to funds, which mine create a separate card, right? With a limit. What else? Give it its own notion
[37:23] account. Like all of this just just make it separate. Right? And then what is it it separate. Right? And then what is it like is there still risks? For sure. Um but but like that is kind of the basics of
[37:36] it, how people operate with it. And I think that's really important. Beyond that, I would also say that the data that you give it, like, security number. Don't give it access to your one password account with every
[37:50] login you have. Yes. >> Right? Like like just just basics basics like that. And then um you can go and experiment. Now, is that 100% secure? Absolutely not. I mean, this thing is really capable. If it was secure, every
[38:04] it and there's a lot of, you know, that's what the big hyperscalers are that knows about all of the company's context and every employee has access to it. That's going to take some time to make it secure. This is really the This
[38:18] is really the, you You California gold rush era where people are going in and you know, like digging for gold but also they get hit over the head with a shovel here and there or whatever. Like it's it's really the wild west.
[38:31] and do stuff like having it scrape the internet and apply your taste to it, you know, um, not saying it's completely you're >> Okay.
[38:44] not exposing your most valuable assets. And is the reason people want to get Macs mainly because also just Macintoshes are trusted and like the UI UX, you know, cuz you could do it on a different device, but the way websites
[38:58] itself and why those are so popular? Yeah, so I think the big advantage is Mac OS and and how simple it is and it's also Linux based um in the background. So the so the terminal um is it it basically runs on the the same
[39:14] is it it basically runs on the the same um same architecture. Now, one thing is really amazing for me as a as a user of it, it connects to messages so I can just remote control it from the from literally the the messages in my um in
[39:26] my iPhone. It's incredible. So you text your Mac mini. You text Alfredo like, >> Yeah, Alfredo like we there's some work you need to do or then it just does it. It's like, "Hey, I'm talking to Shawn like I I I didn't
[39:40] actually." It's like, "Pull the latest 10 videos and like analyze like his opinions and AI and like give me a little brief." And then Alfredo comes amazing. That's better than some weird terminal, yeah.
[39:53] And then also Mac apps Mac apps. Like a lot of the apps um they come out on Mac first or they're Mac only and this like everything just it it kind of just works beautiful thing. If you're dealing with something new and complex, you don't
[40:07] updates. Bringing this all together, what we're seeing is there's this opportunity to move to level three in AI and to eventually have basically AI working for you, working with you. If you want to get on the bleeding edge
[40:22] Mini, start learning this stuff, follow tutorials. You've given us a lot of advice. There is risk, but if you sandbox it, it could be smart. Um these things are figure-out-able. Now, furthermore, uh again, we'll hit
[40:35] this at the end, but you're hosting a summit. There's just so much more in AI to cover. Uh it's called the AI Advantage Summit. So, you know, in conclusion of this episode is like, okay, there's a lot happening here, it's
[40:48] very exciting. I still have more questions. I need to get educated. And Tony Robbins. Um and so we'll we'll tell more details Advantage Summit. We'll link to that in the show notes.
[41:01] And so, you can do this. There's some downsides to tech side, even on this very recording. Um but it is so powerful. And we can whole packaging thing that you do, but
[41:14] you basically revealed your playbook for research, and not just at a shallow but for having ratings of articles, criteria, things that fit your taste, curating Twitter accounts and all that stuff. So, now you can just talk to it.
[41:28] And it keeps It's an engine for your channel because it is leaking you the jump on them quickly. And there's the viral hack for everybody Yeah. Who would actually here for you and everybody watching this video. Break it down.
[41:45] There's one thing that I really noticed with like all these open claws. be some guy in the comments who's going to be like, oh, actually actually you Generally speaking though, generally speaking,
[41:59] the stuff that works well in ChatGPT works well in here. So, stuff that you could already make work in ChatGPT is also going to work well. Like, ChatGPT was already good at like doing research and fil- filtering and using context. It
[42:12] And again, not to say there's no exceptions, but like generally speaking, worked in chat, you can bring it into here and kind of automate the heck out >> Whereas if you're like oh, I wanted it to find me a girlfriend. Like
[42:27] okay, good for you, but like that doesn't work in chat, it's not going to work in here, right? I got you. Okay. And so kind of the final section is the cool thing that anybody listening to this, if
[42:40] you don't want to set up all of the tech headache and get a Mac Mini, the final tips you would have, can I just do I go chat GPT projects, which LLM should I use, do I need to get this Claude co-work thing? If I want to up my
[42:55] game and create an edge with AI without setting up open claw, what would you recommend? I think there's one app at this level free that kind of currently in March and April 2026
[43:09] trump small and it's Claude co-work. It's just the sweet spot of user-friendliness. You just download a desktop app on a Mac um or on a Windows and you kind of just click buttons. No Mac Mini's, no
[43:28] command line interfaces. It kind of just looks like the desktop app of Claude. But it can also organize your downloads folder and create 20 new folders and look at all the files and put them in there. It can
[43:41] also link to an email account and you remote control your browser and send an email. Like it can do things and it's getting more capable by the day. Like Anthropic is really trying to kill open claw through all the updates they're
[43:53] shipping. It's insane. Over the last month they shipped like 25 updates. coming out. And that's the user friend That's what I That is like yeah. Does it have additional cost? I pay 20 bucks a month
[44:06] >> Included. Included. So, you need a paid plan, but on the $20 plan you already tab on the desktop app of Claude. >> And then there's a cap out uh credits as >> And then there's a cap out uh credits as it would. Yes. And so Got it.
[44:20] cost-effective or basically the same of, you know, what you're paying if it's a running on Open Claude to do your research and do your packaging? Yeah. Is it going to be cheaper or Way cheaper. Okay.
[44:37] >> So, usually the calculation is, you know, it varies, but roughly it's like saving on those subscriptions if you're like really pushing them to the limit. And also what I should add is that budget that I outlined, the 100 150
[44:51] running. When I go in and tell him I'm like, all extra. That's I count that as Yeah. >> And you know, I mean, honestly, just
[45:03] tokens I spent I think I spent like $4,000 to be honest, like figuring this I guess that's the value you get out of this video. That research workflow cost me thousands to put together. It might look simple now. It's like, "Duh." But
[45:18] like hooking it up, using the Notion as the source of truth, having that velocity field actually work, having a deduplication in there where it then adds duplicates into an existing entry rather than to a new one, uh having the
[45:32] taste file really dialed in, like like this took a while. Um and I experimented this tool, but like you kind of get into it and all of a sudden it's like, "Oh, I just added $2,000 and it's gone again." I mean, as we had it during the session,
[45:46] you and what you're doing, especially with your YouTube channel and the events is you're [laughter] you're saving us time and money. And you're doing the research and the heavy lifting for us, so we can follow
[45:59] in your footsteps. So, yeah, there's two paths really from here is actually looking at Claude CoWork. you could just start tapping into the benefit there.
[46:13] The insights is replay this episode. I know it's very geeky, but even your prompting will improve after hanging out with Igor. So that your prompting even how you're doing Chat GPT, if that's your main LLM, the projects you're doing
[46:28] know that's next level. Here's the thing, if you're feeling overwhelmed by thing, if you're feeling overwhelmed by AI, then Igor is hosting a summit with incredible speakers. If you're watching the video version, you can see on screen
[46:41] just the lineup. Three days, totally free. If you're listening on audio, you could type in learnai.com show notes on the YouTube version or the audio version. Get registered for the AI
[46:57] Advantage Summit. That's a free event. We watched the last one. I know you guys this, but you did it like a few months later because things are moving so fast and there's so many updates people need and you don't want to get left behind in
[47:09] stress-free way. So, we were actually on the last one. We jumped into y'all's paid community. Brian, who's like leading AI at Think Media, ordered trying to keep up. And so we're super grateful for you to be the
[47:22] expert that we're learning from as well as your whole team and everything like that. So, learnai.com Igor, final thoughts. Of yes, there's this whole tech side, but why does an event like this matter
[47:35] now? Oh my god, it's just moving so fast and all of the media is hyping it up, everything up and there's just all this overwhelm and this this velocity of technological development. It's very real and people feel it. They feel kind
[47:49] feel the overwhelm. They feel their peers, you know, getting ahead and they're like not doing the latest open claw thing. And I want to be very clear, this summit is not designed for the people who have, you know, 5,000, 10,000
[48:03] hours and are vibe coding their own apps and running their own open class. This summit is for the 80% of people who are just like, this is too much. But here's the cool thing. Here's the cool thing. And this is
[48:15] People even at this level, they don't have the level of depth of context to make certain workflows work like this, right? Like, cool. If you have page-long DNA files along with everything you do, good for you. But practically, most
[48:28] people don't, even if they're at this level free already. That's why we turn all of this madness that AI can be into a simple to follow roadmap. We outline levels. You have different speakers speaking from their experiences, sharing
[48:41] what worked for them, and a community of incredible people being there and just sharing this unique moment that AI is. And ultimately, ultimately, what the summit does is it gives you clarity. It gives you The The The goal is to give
[48:56] you clarity on how you can use this technology to get ahead, not just catch up, but get ahead yourself. To get the time back that this technology can give you and just increase the leverage that you already have in your life. But like,
[49:09] give you more leverage. We're going to show you what really are the difference structured manner. It's over 3 days, March April 21st through April 25th. It's completely free, not kind of free, completely free. And I would love to see
[49:24] roadmap, and we're going to be talking about everything um relating to these levels and how to progress with using AI. I love it. And so, Think Media Podcast, a couple things. Um if
[49:36] you got value out of the episode, of course, like, rate, share, review. Um channel, The AI Advantage. That'll be linked up in the show notes. Um we have a part one of this conversation where he
[49:49] also talks about how how turns those news stories into videos that can be posted within two hours of a new story being broken. So, just if you didn't that episode and then >> That one's less geeky. Yes, less geeky
[50:05] How he grew over 400,000 subs. So, a lot of great resources for you. But, if it's don't miss the AI Advantage Summit. I'm going to
[50:17] be taking notes as well because we're trying to about AI, but doing it in a stress-free way. And so, all of that is linked in the show notes on the podcast audio or video or you can
[50:34] just go to learn AI summit.com to get registered for totally for free. Igor, the Think Media Podcast and I'm excited for future episodes cuz you're always dropping fire here. Think Media Podcast, I'm Sean Cannell, your guide to building
[50:48] wait to connect with you in a future episode.