Transcription
Alexozi is now posting 450 pieces of content every week, thanks to his brilliant AI content system. But unlike most AI content, all 450 pieces of Alex Mozzy's content are so good that they are indistinguishable from the content he made before he had AI do it for him. And we're actually going to be putting that to the test in this video.
Ammozi claims that this system allows him to create tweets seven times faster. It used to take him 2 hours to make one YouTube video. Now, he claims it takes just 15 minutes. He's even blown up his views across all of his short-form content. And so, in this video, we're going to hold hands and walk through step by step his exact AI tools, LLM, and most importantly, the big secret that Hamozi says is using the AI workflows.
So, I'm going to start with a high-level overview. Then, we're going to go through three specific examples: Twitter, YouTube, long-form, and then short-form content, showing you exactly how Alexozi has strapped a rocket engine to his content game. If you're as excited as I am, please hit that like button down below. This is a new channel, and I'm excited to grow it alongside you. And without further ado, let's get started.
So, Alex Emozi says that most people are doing AI content extremely poorly. They go to something like Chat GBT. They go to something like Claude, and they type in, "Hey, Claude, can you please make me a tweet?" "Hey, ChatGBT, can you please make me a YouTube script?" And the output of this is what we call AI slop. It's a waste of everybody's time, and honestly, you're probably better off not even using AI.
And so, we first need to understand the bird's-eye overview of what is an AI workflow. An AI workflow is four folders. You need a folder on your desktop, the parent folder called AI assets. Inside this folder, we have three other folders. A little bit of folder separation for you. One is called business context, two is called data, and three is called prompts. Inside these three folders, we're going to fill it with business context, business data, and exact, precise prompts inspired by Alex Hozi. In fact, I'm going to show you the exact prompts that he is using.
So, in the first folder, we put an entire context about your business. And this is going to ensure that every piece of content that the AI outputs is going to be aligned to your brand and aligned to the goals of your brand. Then we're going to put as much data as possible, and we're going to show you exactly where you get the data from your business that Alex Mosy says is necessary that you feed your AI. And this will ensure that you do not get AI slop as an output, but instead, you'll have incredible, unique, authentic pieces of content.
Alexoji says things only sound like AI when you don't have a baseline to train it on. AI sounds like the internet because it's been trained on the internet. But if you train it on the work you've done, it sounds like the work you've done, which is what we want. And he even gives us an example. He says, "My Mosy Minute emails are bangers because I wrote a crap ton of bangers. My tweets are good because I already wrote 6,000 tweets before he got AI to write his tweets on his behalf." 6,000. It has a huge data bank to pull from. In fact, in the example, we're going to pull all 6,000 of his tweets.
But before we get there, the last ingredient of this perfect formula of perfect content are prompts. And he says the longer the prompt and the more specific the prompt, and he has a prompt structure, which I'll get to in a second, the better the output the AI is going to have when it comes to creating content for you. He says the longer the prompt, the shorter the output. The shorter the prompt, the longer the output. You have to give more context so that it answers exactly what you want. It doesn't know what you want, so it just gives you a ton of answers, hoping that one of them matches.
And so, overall, Alex says that AI has to know what your business is about. It has to have good data that it can base its output off, saying, "This is exactly what we want it to look like when it's right." And it has clear instructions on exactly what you want. That's the prompt section. When you have all three of these things, you can put them together and make banger outputs. So, let's get into some specific examples.
So, I've actually laid this out like this for you because even though you're going to need this uh folder situation on your desktop, which I'll show you how to get in a specific example, you have to realize that the business context is always going to stay the same. And then within each sort of output, so if we want a YouTube video, you need YouTube data and a specific YouTube prompt. If you want tweet output, you need specific X data or tweet data and a specific tweet prompt. If you want shorts, reels, and clips, you need clip data and a specific clip prompt.
So, obviously, step one, and if you haven't watched my video I put out yesterday or about 2 days ago, I walk you through exactly how to get the business context, the data, and the prompts for your business from a holistic overview. In this video, we're going to go more specific just in the content setting, but we still need business context. And so, Alex Hamozi has a 12-questionnaire about your business that he thinks that you should give your AI before using it in any way to create content for your business. There are 12 questions. Again, go watch the video I put out yesterday if you want to get the exact 12 questions and answer them yourself as we go along. But I have mocked up Hozy's answers to these questions.
So, you can see question one here is, "What does your company do in one sentence?" And this is what Alex Mosy's company does. "What industry or niche are you in?" This is what industry and niche Alex Mozy's company is in. "How long have you been in business?" Again, you're going to have your own answers in here, but we're just using Alexozi's business as an example. Then you have four questions about your customer. For example, "What transformation do your customers experience?" This is the transformation that an Alexi customer will experience. And then you have your offer and your model. You've got four questions here to bring or 12 of those questions into one single document called my business context in a text file. Again, I'm going to show you exactly how to get this. Don't worry, we're just going to go through this example together. But you can see this right here is a mockup of the business context for Alex Homozy's business. And that business context, like I said, is going to stay the same through every single prompt and output.
However, we want in this first example to do some Alex Hozi tweets. So, the next thing we need is data. And like Alex Hozi says, he's tweeted thousands and thousands of times in the past. So, we want to train the AI model on these tweets. And so, guess what I downloaded? I downloaded every single Alex Hozi tweet that he's ever made. Obviously, for demonstration purposes. I'm not actually using this for my business at all, but I just want to show you how you would do it if you were Alex Hamozi. Here are 6,630 Alex Hozi tweets, including which ones got the most views, which ones got the most replies, which ones got the most favorites, etc., etc. So when I'm working with my AI, I know which ones were the favorites and which ones to avoid. We can have some machine learning in it. So this right here is what we call uh X data or tweet data. I'm talking about not X data like a placeholder X. X as in the actual company uh Twitter.
Now, the final part of this formula is, of course, the specific prompts. And interestingly enough, I found on Alex's YouTube exact prompts that he uses in his business. It says up here 10 prompts, but it's actually 14 prompts. And each prompt is for something different. So, this prompt will create sales page copy, or what's more relevant for us, this one right here, which we're going to come back to, will create short-form style content for Alexozi to go and record. And you can see each of these prompts, as he says, the longer and the more specific it is, the better. If you want to see the exact way that he structures a prompt, once again, go watch the video I posted yesterday. We do a deep dive on exactly this.
Prompt number two from Alex Heroszy's magical prompts is called a newsletter style analyzer and a new newsletter generator because he has his Mosy Minutes, which AI is now helping him write. He says it takes like 90% of the work away, but we are trying to get some tweets. He says this prompt and the setup that we're showing you takes about 90% of the work off his plate. And you can see right here, "You're an expert copywriter and narrative strategist trained to analyze writing samples, extract voice, structure, style patterns, and reproduce them with new ideas while maintaining the same feel and reader impact." Of course, it needs to be given the actual data, which is the whole point of this. Cuz if you just give it the prompt, it's not going to know what to work with. So, it analyzes the past winning newsletters, which is the data, and then builds a reusable newsletter template based on the data. Then can generate a brand new newsletter when, and this is a big thing with Alex Hamuzzi, he never lets the AI actually decide the ideas and the topics. He has them create templates and patterns and tones, but he always feeds the AI the actual idea. And then in the output format, you can see he's got the voice, tone, pacing, signature writing behaviors, narrative, teaching patterns, and call to action tendencies, plus the template that uh he can use to build future newsletters. And then step three is building that new newsletter after Hermosi gives the AI a new idea to build it around.
In these 14 prompts, there is no actual Twitter prompt, even though Alex Mosy says that he's doing his Twitter mostly through AI at the moment. And so what I did is I took that newsletter prompt and I got AI to adapt it to be a very similar-looking but equally long and equally specific tweet prompt that we can feed our AI. And I'll put this plain text ready prompt, of course, in the description below if you're wanting to use it.
And so now we have all three files. So we're going to open our LLM. I'm going to use Claude for this. And so inside Claude, I've uploaded the business context file, which is Alex Mosy's mockup data plus his 6,630 tweets. And then I'm going to paste in the prompt right here and hit enter. And so Claude will have access to the business context, the data, and the prompts. Of course, you want to set this up as reference files in the future, but this is just an example to show you how it's going to work. And we're going to come back and see those tweet outputs once we give it some ideas to tweet about.
The second example I want to walk you through, however, is one that intrigues me because I run a couple of YouTube channels, including this new one here. And if you're enjoying this video, please hit that subscribe button below. At the moment, I think we're on 40 subscribers. So, we really appreciate everyone who's watching and subscribing. And I want to know exactly what Alex Mos is doing to blow up on YouTube with AI. And so our business context, once again, is going to be exactly the same because it's the same business. However, instead of X data, we now need YouTube data and a YouTube-specific prompt. Fortunately, Alexi has shared publicly his YouTube-specific prompt, which is this one right here. Number seven, YouTube script style extractor new script generator.
So, "Ro, you're a world-class YouTube script analyst and creator. You study video transcripts and extract tone, pacing, narrative, mechanics, emotional beats, comedic timing, if any, hooks, story arcs, call to action patterns, and signature stylistic choices." Here comes the data part right here. "I will upload one to five transcripts from YouTube videos I like." Now, you can upload videos of yours from the past or just the transcripts of like this is saying right here of YouTube videos that you like and aspire to make. The AI's job is to analyze them, extract a style, and build a reusable script template. And again, this is the key right here. AI does not come up with the idea. When I give you a new idea, generate a brand new script using that template, matching the exact feel of the original videos, but with completely new content, not ripping someone off and not coming up with the idea themselves.
So you can see this is the long uh prompt, which apparently is better than a short prompt, or of course, it's better than a short prompt. And Alex Mosy says the longer the prompt, the more specific and high-quality the outcome. The output format is they're going to have that template and the style analysis that they'll apply to a new idea when we prompt it to. So we're going to open a new Claude code session. Of course, you can use Chat JPT or just the normal Claude. Whatever AI you like using, it all works. Claude, I highly recommend. If you want more tutorials, I have them here on the channel as well for you to get my exact AI setup. That is honestly, I call it god mode.
But we're going to pull in in this example the business context. We of course are going to need the exact prompts. I'm going to come up here and get it right. I'm even going to include the title in this prompt example that we are outputting right here, all the way down to the confirmation. Command C that, come in and paste it right here. The last thing we need is the data. To the one to five transcripts of videos we like. And so for this example, I had an interesting idea. I really like the YouTuber Behea. He has 1.5 million subs. He doesn't upload often, but the style of his videos is absolutely phenomenal, and they're more action-based where Alexozi is more talking and teaching-based. And so I'm like, what would an Alexi Behea style video actually look like?
So, I came in and I found five Behea videos that I really like. And I'm going to click on the description, come down to show transcript, and literally manually go from the top right here all the way down and highlight all of the text. Now, I'm going to command C, open a new text edit file, and just paste that transcript right in like this. And Hozi says you want to separate them. So, you want to call this transcript one. And then when you come down the bottom, you want to do transcript two. And then I go back and find that next Behea video. Come down to the description, show transcript, and copy it out again. Of course, paste it back into the same area. Now, you can do this for five or 10 different videos. I'm going to keep it at two videos for now, just because I'm just giving you an example. I'm going to call it "Behea transcripts" and then come back into Claude codes project and paste it in there for me, just right there. "Behea transcripts" and send that prompt.
Now, while that loads, I'm actually going to come back and show you the tweets because that is done. You can see this is the prompt we sent it. We gave it the context files, and you can see it acknowledged everything that we uploaded. Step one, it did a style analysis like we asked it to. It says the voice profile is declarative, never hedged, no "I think," no "may," no questions as engagement bait. Stoic, working-class oracle wisdom dressed in plain words. Speak directly to you. Remember this, read 6,600 Alex Mosy tweets, and now he's analyzing his exact voice profile. Authority without preaching. The hook patterns he's using are "You can beat 99% of people if," "How to stay poor," or "How to insert bad outcome." "Rich people X, poor people Y." "If condition, imperative loses X, winners Y." These are the patterns. Uh, one-word imperative openers. "Stop crying," slash "askthegirl." Length norms. This is exactly how uh long his tweets normally are. And there's two clear bands to the tweet: a punchline and a structured list. Uh, long form is 500 plus. It's rare but disproportionately viral when used, which is very interesting to look at.
So you can see it's done all of this analysis on all of his past tweets. Step two, the template library. So you can see they've got different styles of Alex Hamosi tweets based on the different styles he's been using. So the skeleton of the actual template is "the superlative noun is the ability to counterintuitive verb when adverse condition." Example: "The single greatest skill you can develop is the ability to stay in a great mood in the absence of things to be in a great mood about." Or "The biggest success habit is being willing to look stupid." And you can see they've done this for different styles. So "type a verb object." "Rich people buy time. Poor people buy stuff. Ambitious people buy skills. Lazy people buy distraction." You can see that is the exact format right there. The 99% threshold template. "You can beat 99% of people if you can master the shame of rejection, the boredom of repetition, and the pain of feedback." Strong tweet. The setup. Then the reframe. "Your 9-5 isn't killing your dreams. Wasting your five to nine is." Again, really strong. The imperative stack paired inversion. You can see there's all these different templates that he's been using.
And so it says, "The templates are loaded. The voice is locked in, ready for the topic. Drop the idea, and I'll spin 10 drafts across at least four of these formats." Fantastic. The idea and topic is AI and how to get ahead. So, we're going to see these 10 tweet ideas in Alex Mosy's voice, tone, and template. Now, while that first one loads, I just came over here to the YouTube one, which has uh created the exact same template and style analysis, and I've told them that the YouTube video idea, because it doesn't make an idea until we prompt it to, is about making your first million.
On the ideas front, Alex Mosy actually says, "AI does not come up with the idea. The idea is actually the alpha. The idea is where the highest returns actually come. AI can flesh it out, but the core idea is always yours. That is the essence of it." Okay, so Claude thought for about a minute, and you can see right here, here are the 10 tweet drafts. Now, remember, these are just drafts. So we're going to read through them, and what we have to do is we tell AI which ones we like, which ones we don't, and it will machine learn and get better and better at actually making the output that we want. So let's have a look at here.
"The most valuable skill of the next decade is being willing to be a beginner in public on a tool everyone else is scared to open." Pretty good. "Lazy people use AI to skip work. Average people use AI to do the work faster. Smart people use AI to do the work they couldn't do alone. Rich people use AI to find work no one else sees." The 99% threshold one. "You can leapfrog 90% of your industry by sitting with an AI tool for one boring hour a day until this time next year." I actually believe that to be true. The imperative stack. "If you've never opened an AI tool, you have no excuse. Open it, type, fail, edit, try again. The bar has never been lower." I can hear him in that one a lot. A paired inversion. "Amateurs ask AI for answers. Pros ask AI better questions." Bit generic, but not bad. How about a definitional maxim? "The biggest moat of the next decade isn't access to AI. It's the reps you put in while everyone else waited for it to feel obvious."
Okay, now heading back over to the YouTube script. You can see Claude has done thinking. Estimated runtime, topic: making your first million dollars. The hook. "Making your first million dollars is probably the most Googled financial goal in human history, and definitely the one you've been sold the most Lamborghini telling me I can do it in 90 days. So I figured, let's actually look into this." Which is actually, this is interesting, cuz I watch a lot of Beerheads videos, and this sort of is his angle. He calls out gurus all the time, and he calls out what not to do all the time. "In this video, we're going to test some of the most popular strategies for making your first million to see what's real, see what's nonsense, and try to figure out if there's actually a clear path here, or is it just a well-marketed coin flip?" Very fascinating, and it's actually lent really strongly into that Beerhead frame of "I'm going to guinea pig and test it out for you."
So, we won't go through the whole script. I'll link it up in the description below if you want to read it and see how good or bad it is, but I wanted to touch on the last example here, the shorts, reels, and clips. So, once again, the business context stays the same. The clip data, we'll go over to the actual clip prompt, and I'll show you what the clip data also is. Number eight is short-form content style extractor a new script generator. So I'm not going to read it all the way through. Again, I'm going to make sure this script is available for you. Alex Mosy did make it public on his YouTube channel. So this is his prompt. I did not actually write this for you, but I'm just passing it along to you. And you can see it's pretty similar to the long-form YouTube one, but styled around a short-form template. And of course, it does the style analysis. And then we'll wait for you to give it an idea.
However, what is the data? Once again, upload one to five short-form transcripts and videos that I like. You can do both. Upload five short-form transcripts that you like and upload the past uh short-form content that you put out there that did really well. This video is getting a little bit long, so I'm going to actually make a dedicated new video on exactly how to do the short-form one. Go give it a shot though. I've shown you the pattern and how to do it. And hit that subscribe button below so you can see it when it comes out. Thanks so much for watching. I appreciate you. I'll see you in the next video.