The 7 Levels of Claude Code & Content (From Slop to Agentic)
CChase AI
Advertising/MarketingSmall Business/StartupsComputing/SoftwareInternet Technology
Transcript
00:00:00"Let's master Clawed Code Marketing."
00:00:02Now, to be totally clear,
00:00:03this isn't gonna be like most Clawed Code Marketing videos.
00:00:07I'm not just gonna be throwing
00:00:0910 million different AI tools your way
00:00:11and saying, "Hey, add these to your stack
00:00:13and it's gonna solve all your problems."
00:00:14And there's two main reasons for that.
00:00:16Number one is if we do that,
00:00:18this video is gonna be outdated in about six hours
00:00:20because the best tool for the job
00:00:22literally feels like it's changing at that rate.
00:00:25And reason number two is because tools aren't your bottleneck.
00:00:29It's taste.
00:00:30And so because of that,
00:00:31this video is gonna be a tale of two halves.
00:00:34In the first half, we're gonna focus on taste.
00:00:36How are we able to create marketing material,
00:00:39whether that's videos, posts, images, whatever,
00:00:42that actually sound like you and not the AI slop drivel
00:00:46that's ruining every single platform out there.
00:00:48And in the second half,
00:00:49we're gonna talk about systems and scale.
00:00:52How can we take this machine that we've begun to build
00:00:55that reflects your views, your values, your voice,
00:00:58and post it not on one platform a day, once a day,
00:01:01but multiple platforms, multiple times a day
00:01:04about topics that matter to you,
00:01:06and again, retain your vision.
00:01:08We're gonna do this by traversing seven different levels
00:01:11of cloud code and marketing.
00:01:12And at each level, I'm gonna tell you what you can expect
00:01:16and more importantly, what you need to do to level up.
00:01:19In today's video, I'm gonna show you how to turn cloud code
00:01:22into a marketing machine that can pull information and data
00:01:26from any source.
00:01:28Turn that data into any form of content we want,
00:01:31whether that's audio, video, or text,
00:01:33and then push it to any social media platform we want.
00:01:37And most importantly, do it in your voice.
00:01:41Now that last point is the most important, your voice,
00:01:44because that is all about taste.
00:01:47That is how we are able to go from AI slop
00:01:50into an AI system.
00:01:51Because if you don't nail that part, the taste part,
00:01:53being able to control the feel, the vibe of what comes out
00:01:56and have it match you or your business,
00:01:59you are just going to be another AI internet tragedy
00:02:02that people see and they see your posts
00:02:04and they immediately dismiss you and not only dismiss you,
00:02:06they now associate you with garbage trash.
00:02:09And there is a ton of that out there and you don't want that.
00:02:12We want all the power of AI,
00:02:14but we don't want those downsides.
00:02:16And we can do it, but it takes a little bit of skill.
00:02:20And we are gonna build those skills
00:02:22as we go across all seven levels of cloud code marketing.
00:02:26For each level, I'm gonna explain when you know you're there,
00:02:30what skills you need to master, the traps we need to avoid,
00:02:33and most importantly, the unlocks.
00:02:35How do we actually level up and get better at this?
00:02:39But before we get started with level one,
00:02:40a quick word from your favorite sponsor, me.
00:02:43Last month, I released the Cloud Code Masterclass
00:02:45and it is the number one way to go from zero to AI dev,
00:02:48especially if you don't come from a technical background.
00:02:51I update this course every single week.
00:02:53We just came out with a bunch of modules going over
00:02:56how to make your own cloud code agentic OS.
00:02:58So if you wanna get a handle on this awesome tool,
00:03:01you need to check this out.
00:03:02There is a link to it in the pinned comment.
00:03:05Hope to see you there.
00:03:06So let's start with level one.
00:03:07So level one is the most basic of all levels.
00:03:11And this is the scourge of the internet.
00:03:13It is just you and AI,
00:03:16probably not even cloud at this point.
00:03:17You're probably still on chat GBT
00:03:19and you're just prompting it.
00:03:20And you're saying absolute nonsense
00:03:22like apparently everyone is.
00:03:23Just write me a tweet or write my next LinkedIn article
00:03:27or write my email for me or write my blog.
00:03:29And that's pretty much it.
00:03:30Maybe you go a step further and you say something
00:03:33like don't sound like AI, no em dashes.
00:03:36Don't even get me started on em dashes
00:03:37as someone who's used em dashes since the dawn of time.
00:03:40It's very upsetting that they've now been associated
00:03:43with AI slot, but that's only the conversation.
00:03:45This is sadly where most people live.
00:03:47This is how we actually get AI slot.
00:03:49And you see this everywhere, everywhere.
00:03:51And this level is why AI gets a bad rap
00:03:53because it's kind of a survivorship bias sort of thing
00:03:56where all you notice is the bad AI writing
00:03:58and it is very, very obvious.
00:03:59It's the things where we're talking about
00:04:01like the chat GBT isms.
00:04:02Like it's not X, it's Y, you know, the em dashes,
00:04:05all that sort of stuff.
00:04:06Now, if you've begun to already recognize
00:04:08these sort of AI isms,
00:04:09well, you've already started to master some of these skills.
00:04:12One of them is recognizing AI default phrasing.
00:04:14We can't cut this out of our system
00:04:17if we don't even know how to recognize it.
00:04:18So that's one of the first skills.
00:04:20The other skills we need to master
00:04:21is writing clear and specific prompts
00:04:23and then being willing to iterate,
00:04:24not just sitting on the first outcome we get from Claude
00:04:28or chat GBT or Gemini,
00:04:30but actually going in there, reading it,
00:04:32seeing what it actually wrote,
00:04:34which a lot of people don't do.
00:04:35I mean like, does this make sense?
00:04:36And going back to signs of AI writing,
00:04:38it's gotten to the point where Wikipedia
00:04:40has a very extensive article about the signs.
00:04:42And I mean, extensive.
00:04:44This is a whole breakdown of the sort of things
00:04:47you see in AI writing and examples.
00:04:50We're still, we're about at the halfway point now.
00:04:52The point being, there is a lot of things
00:04:55that are telltale signs that you're putting out slop.
00:04:59Now people try to get around this in two ways at this level.
00:05:02One, they try to prompt out
00:05:05the most egregious of those problems.
00:05:06Like no more em dashes, please don't do em dashes.
00:05:09Or they try to just brute force it and go for engagement.
00:05:13So they see posts that are doing well
00:05:14on something like LinkedIn or Twitter,
00:05:16and they just tell Claude code,
00:05:18or they tell chat GBT, get louder, get louder, get louder,
00:05:22better hook, better hook, better hook,
00:05:23as if that's gonna solve the problem.
00:05:25Because that is a trap.
00:05:27Because even if it gives you a quote unquote,
00:05:29better hook or better engagement,
00:05:31is that your hook and your engagement
00:05:33and your vision and your voice?
00:05:35Chances are the answer is no.
00:05:37And so to truly unlock level two, what do we gotta do?
00:05:41Well, we need to make it so AI
00:05:42is no longer guessing your voice.
00:05:44We need to develop some sort of taste, right?
00:05:47We need to prompt it in a way
00:05:49that we can consistently have it sound like us.
00:05:52So now we're moving into level two,
00:05:53the taste injector, the voice injector.
00:05:56This is where we're no longer just praying it gets it right.
00:05:59We are going to strategically give it our voice
00:06:03via something like a Claude.md, a voice doc file,
00:06:06literally some sort of context,
00:06:08some sort of documentation that says, this is how I write.
00:06:11Sound like this all the time.
00:06:13For example, here is the system prompt
00:06:15I use for my blog system.
00:06:17It talks about my core mission.
00:06:19It talks about my voice and tone guidelines.
00:06:21It has words and phrases to never use
00:06:23versus phrases that are on brand.
00:06:25And then it has some blog specific things.
00:06:27And we'll talk about how these should be written
00:06:29for specific platforms later.
00:06:31But as you can see, this is relatively robust.
00:06:34This isn't just me telling Claude code,
00:06:36sound like me, be very concise and professional.
00:06:40Now I also have a template of this thing
00:06:41that you can use for your own purposes.
00:06:44And all this will be inside of my community.
00:06:46Now the system prompt is just one piece
00:06:48of the brand voice doc puzzle,
00:06:51because that is the template, that's the guide.
00:06:54Those are the guardrails.
00:06:55But what we can also do is we can also load examples
00:06:59of our best posts.
00:07:00Now, if you don't have any posts yet,
00:07:03then you can't really do this.
00:07:04But what you can do is you can go out
00:07:06to whatever platform you're thinking of,
00:07:08say LinkedIn, for example,
00:07:09and you can find posts of people you like.
00:07:11And then what you do is you take this template
00:07:13I'm giving you, you copy and paste that into Claude code,
00:07:16you then copy and paste those five to 10 best posts
00:07:20from other people you like or your own.
00:07:23And you say, given those posts, fill out this template.
00:07:27And it's gonna take a look at the type of phrases you use,
00:07:30the type of phrases you don't use.
00:07:31It's gonna take a look at your general style
00:07:32and it's going to apply it to this document.
00:07:35And once we have that,
00:07:36we just give the document to Claude code.
00:07:39And we say from now on, write my blog post for me
00:07:42about whatever topic,
00:07:43make sure to reference the brand voice doc.
00:07:46Now I've either directly dropped it
00:07:48into the prop window at that point,
00:07:50or I'm operating in some sort of folder,
00:07:52some sort of directory that includes the document in there.
00:07:56Better yet, we'll talk about this a little bit later,
00:07:59instead of doing all this,
00:08:01I could just turn this into a skill.
00:08:03And I would just say something like run the blog post skill,
00:08:05talk about X topic.
00:08:07And because we set up the skill that way,
00:08:09it would already know, oh, we're doing a blog post.
00:08:11Oh, because we're doing a post,
00:08:13I need to do it with his voice.
00:08:16So I'm gonna take a look at this document
00:08:17that says exactly what it should sound like,
00:08:19what phrases to use, what phrases not to use.
00:08:21And at the end of the day,
00:08:23you will get something that sounds way more like you
00:08:25and way less like generic AI.
00:08:28This is the foundation.
00:08:30This sort of document is a non-negotiable.
00:08:33You need this if you want AI to market with you
00:08:36in your voice, in any capacity, in any sort of deliverable.
00:08:39So what skills do we need to master here at level two?
00:08:41And by the way, most people don't even get to this level.
00:08:43We're talking about seven levels.
00:08:44Most people don't even get here too.
00:08:46So we need to curate the best examples.
00:08:48When I said find five to 10 of the best posts
00:08:51or five to 10 of posts you like and you want to emulate,
00:08:54that's the high mark, 10, no more than 10.
00:08:59Three to five is probably the sweet spot
00:09:01because if I feed it too much information,
00:09:03it's we're kind of just overloading the system.
00:09:05And again, less is more.
00:09:07The document I have given you that template
00:09:09is really all you need.
00:09:10You don't need some 30,000 word dissertation
00:09:13about what to do and what not to do,
00:09:15which kind of leads into this idea of context rock
00:09:17because there are people who understand this idea
00:09:20of feeding examples and they say,
00:09:21"Hey, I'm gonna create an entire rag system
00:09:23with 40,000 articles that I like."
00:09:25That is stupid.
00:09:26We don't need to do that.
00:09:27Now, what's the real trap here?
00:09:29I talk about the voice doc ball
00:09:31and you taking forever to put this together.
00:09:32Well, the other trap is thinking your work here is done
00:09:35at this point.
00:09:36Just because I created this document
00:09:38and it's starting to sound like me,
00:09:40this is a living and breathing thing,
00:09:42especially if you haven't posted before.
00:09:45What you wanna continue to do is figure out
00:09:47what posts actually do well.
00:09:49And then as you figure out what posts are doing well,
00:09:52you know what you're gonna do?
00:09:53You're going to feed it back in a cloud code.
00:09:56You're going to update your document.
00:09:57You're gonna take a look at the outputs
00:09:59after you use this for a week or so.
00:10:00You're gonna go through them with a fine tooth comb
00:10:02and see what you like and don't like.
00:10:04This isn't a one-shot thing.
00:10:05This is something that if you want cloud
00:10:07to be some sort of self-improving system,
00:10:10you need to constantly be updating it.
00:10:12But simply getting this document in there
00:10:14is a huge step forward and puts you ahead
00:10:16of like 90% of the people using AI to write stuff for them.
00:10:20But we wanna do more.
00:10:21We wanna talk about level three.
00:10:23How do we get to level three?
00:10:24Well, level three is where we start thinking about strategy
00:10:29because it's one thing to have it right in our voice,
00:10:31but what are we writing about?
00:10:32We need a system in place where we can grab information
00:10:35from somewhere and create this funnel
00:10:38between information and analysis and deliverables.
00:10:41Well, that's what we're gonna talk about in the next level.
00:10:44Now, level three is where we start going into systems.
00:10:46How do we figure out
00:10:47what we're actually going to be talking about?
00:10:49Now, this might be obvious for some of you,
00:10:51but I know there's a lot of you out there
00:10:52who kinda just show up in front of your computer
00:10:54or in front of your camera
00:10:55and you just pray inspiration falls into your lap.
00:10:57But we can do better, especially when we have cloud code.
00:11:00And it is with cloud code that we are going to figure out
00:11:03how to scrape information and how to grab it
00:11:05from whatever we deem to be this sort of fountainhead
00:11:09of knowledge that feeds our content.
00:11:11For me, that looks like my daily morning report
00:11:14that cloud code creates for me.
00:11:15And this is simply a set of skills.
00:11:17This set of skills queries the web for information
00:11:20about AI and coding agents and anthropic.
00:11:23It looks at Twitter and GitHub and does the same thing.
00:11:26It then synthesizes all this into a briefing
00:11:29and then it just shows up inside of my Obsidian vault
00:11:32and I wake up and it's there.
00:11:33Based on that morning report,
00:11:35I then identify things that I think will actually make sense
00:11:38to talk about.
00:11:39And that's where something like a deep research workflow
00:11:41comes in, which is again, surprise, surprise,
00:11:44just a series of skills I created with cloud code.
00:11:47So if it's something that I think there's YouTube videos
00:11:49about, I use my YouTube pipeline skill,
00:11:51which then gets sent to notebook LM
00:11:53using the notebook LM CLI and the notebook LM skills.
00:11:56It does all the analysis for me, it synthesizes it,
00:11:59and then it creates a brief which basically says,
00:12:01here's what it's about, here's what you need to care about,
00:12:04here's some sort of content ideas.
00:12:06The point isn't that, hey, you need to do this exact thing,
00:12:08although all of these skills and setups
00:12:10are inside of my community.
00:12:12The point is you need to identify one,
00:12:16where does the info live for you?
00:12:18This is gonna change depending on your niche.
00:12:20For tech, it's pretty obvious.
00:12:21It tends to be Twitter,
00:12:22and then usually even below that is like GitHub itself.
00:12:25This is like where all the information is coming from.
00:12:27Beyond that, you have YouTube,
00:12:28which is a little bit of a laggard,
00:12:30and then even more lagging behind is stuff like Instagram,
00:12:34TikTok, LinkedIn.
00:12:35So if you're someone who's creating content
00:12:37and you kind of need to be on the cutting edge of it
00:12:39to get people to talk about it,
00:12:40well, where's that information, right?
00:12:43That's what you need to figure out.
00:12:44And there are step two, you need to figure out
00:12:45what do you want out of that information?
00:12:47What does that synthesis need to look like?
00:12:49For me, it's pretty easy.
00:12:50It's just like, what is it?
00:12:51Do you have some sort of link
00:12:53where I can look at actual documentation?
00:12:54Okay, pull out the so what for me, why should I care?
00:12:57And then figure out some sort of content ideas for me.
00:13:01Again, totally depends on your niche,
00:13:03especially if you're someone
00:13:04who's not necessarily talking about something
00:13:07that is I think as rapidly evolving as AI.
00:13:11Let's say you're someone like the fitness industry,
00:13:13like where is the info coming from?
00:13:16Well, obviously there's new studies that come out,
00:13:18but it's not like there's a new exercise
00:13:19that hits every three minutes, like a new AI tool.
00:13:22But you still need somewhere
00:13:24that you manually have been grabbing inspiration from.
00:13:27The point is, what is your normal inspo?
00:13:29Where does that come from?
00:13:30Okay, have Clod Code grab it for you by itself.
00:13:36Now, how it actually goes out and grabs that information,
00:13:39again, is sort of use case dependent.
00:13:40If it's something like YouTube,
00:13:42I have YouTube search skills,
00:13:44I have things like notebook LM,
00:13:45I have an entire app that looks on Twitter
00:13:47and finds trending things.
00:13:50But if that's not you, it's kind of just depends, right?
00:13:53This video could be seven hours long
00:13:54talking about the different ways
00:13:55to grab information from different places.
00:13:57But the way you figure out that workflow is the same,
00:13:59no matter who you are.
00:14:00And the workflow is you go into Clod Code,
00:14:04you open up this microphone
00:14:06and you just do a stream of consciousness
00:14:07where you tell Clod Code,
00:14:09here is my normal workflow when it comes to marketing.
00:14:12Here's how I normally find the information.
00:14:13Here's my normal manual flow
00:14:15for how I then turn that into a blog, a post, a video,
00:14:19an image, a carousel, whatever.
00:14:21You go to Clod Code and you explain
00:14:23in as great of detail as possible what your workflow is.
00:14:27From there, we then ask Clod Code,
00:14:30can we turn this into a skill?
00:14:33Or can we turn this into a set of skills?
00:14:36Because what it's gonna do
00:14:37is it's gonna take all of your individual tasks
00:14:39that make up your daily workflow,
00:14:41it's gonna turn them into a skill,
00:14:42which we can then automate down the line.
00:14:44But Chase, I don't know how to create skills.
00:14:46I don't know the best way to create skills.
00:14:47Doesn't matter, Clod Code knows how to create skills
00:14:50better than you or I and test them.
00:14:52It's called the Skill Creator skill.
00:14:54I've done a bunch of content on this.
00:14:56Just go inside my channel and search for skills.
00:14:58You'll find this.
00:14:59We then use the Skill Creator skill to create skills for us.
00:15:02If you don't notice, pretty much Clod Code
00:15:04does all the work for us at every step of the way.
00:15:06All you have to do is tell it what you do.
00:15:09Tell Clod Code what you do and how you work
00:15:11and it's gonna take your task, turn them into skills,
00:15:13optimize the skills themselves with the Skill Creator.
00:15:16And what have we done?
00:15:18Well, we've gone ahead and we've essentially automated
00:15:21a huge portion of our entire marketing workflow,
00:15:25which is the ideation process.
00:15:27And if you currently live in a world
00:15:29where you have no Clod Code skills that are working for you,
00:15:31the second you open up your computer
00:15:33and you're giving you at least a solid headstart
00:15:37on figuring out what it is you're gonna talk about
00:15:39from a marketing perspective each day,
00:15:41you are wrong and you're falling behind.
00:15:43Because if it wasn't obvious enough at this point,
00:15:46once we've done level two and level three,
00:15:49we're kind of like 90% of the way there
00:15:51when it comes to creating a content automation process.
00:15:55Because you now know how to make AI sound like you
00:15:58and you now understand the process
00:16:00by which you can automate the ideation process,
00:16:03where you can go out, find data and bring it back to you.
00:16:06So if I know what I wanna talk about
00:16:09and I know how to then talk about it in my voice,
00:16:11well, the next four levels, as you will see,
00:16:13are all just about refining it for specific platforms,
00:16:15how to do this from a multimodal perspective.
00:16:18But this level two and three is the moneymaker completely.
00:16:21And it's super easy to do.
00:16:23You just have to start using Claude code as a collaborator
00:16:26and bring it into the fold.
00:16:27So to wrap up level three, what are the skills to master?
00:16:29Well, it's really about creating skills within Claude code
00:16:32that makes sense for your particular workflow
00:16:35for how you find information for your marketing.
00:16:38The trap here is to go a little bit deep
00:16:40and all of a sudden we're going beyond just skills
00:16:42that I can execute very quickly inside a terminal.
00:16:45And we're trying to make it all fancy,
00:16:46more than creating 10,000 dashboards
00:16:48instead of just realizing this is just step one,
00:16:51this is just the ideate,
00:16:52we don't need to turn it into the product itself.
00:16:54And to move on to level four,
00:16:56this is where we start going multimodal.
00:16:58This is where we start talking about image generation,
00:17:00video generation, bringing in other things besides text,
00:17:03because up until now it's been very text heavy,
00:17:06but there's a lot out there in the world
00:17:08of AI content creation that can really move the needle
00:17:11on specific platforms,
00:17:12especially ones that are more video and image oriented
00:17:15like YouTube, TikTok and Instagram.
00:17:17Now let's go into level four, which is the creative director.
00:17:20This is where we move from text
00:17:22into stuff like images and videos.
00:17:25And the great part about this is nothing really changes
00:17:29in terms of the fundamentals,
00:17:31whether we're doing images or we're doing videos,
00:17:34the important thing is taste,
00:17:36which really means our branding, our voice.
00:17:40Now, when we talked about voice,
00:17:41it was pretty easy in terms of texts.
00:17:44You know, we have this template that says,
00:17:46do this, don't do that.
00:17:47Here's some examples, fill it out.
00:17:50Well, it's not gonna change here.
00:17:51Is the prompt gonna be a little bit different
00:17:53if we're doing images with nano, banana or GPT images too,
00:17:56than if we're creating a LinkedIn post?
00:17:58Of course, same thing with the videos,
00:18:00but it's still the same in terms of a high level overview.
00:18:04Here's what I want you to do.
00:18:05Here's what I don't want you to do.
00:18:07Here's several examples of what write looks like for me.
00:18:11And by doing that, that's how you create,
00:18:13you know, a consistent brand
00:18:15across whatever deliverables you have.
00:18:17And even in terms of finding things to create
00:18:20is still the same as everything we talked about in level three.
00:18:23I demonstrated this in my last video
00:18:26and I'll put a link to that above,
00:18:27where we used Claude code to go on to GitHub
00:18:30and find trending GitHub repos for that week.
00:18:33We then synthesize that information inside of Claude code.
00:18:36We had it figure out the so what and what the copy,
00:18:39AKA the voice of our carousel post,
00:18:43which is both half text, half imagery needed to sound like.
00:18:46So we had a consistent voice and taste
00:18:49across images and copy.
00:18:50We then sent that via the Higgsfield MCP to GPT images too,
00:18:55which created this image for our carousel
00:18:58and the follow on slides in a consistent manner
00:19:01that was pretty much the same of what we've done in the past.
00:19:05So those fundamentals of brand voice,
00:19:08AKA taste of getting the sort of images we want,
00:19:10and then level three of finding things to talk about,
00:19:13in this case, it was GitHub repos,
00:19:15was done in an automated way.
00:19:18That was, in my opinion, anti-AI slop.
00:19:21All we've done here in level four
00:19:22is we're just adding this part.
00:19:23We're just not doing text only, we're now adding images.
00:19:26Now, I will say one of the tough things
00:19:28when it comes to sort of this multimodal thing
00:19:31is the amount of tools available.
00:19:32And like I said in the intro,
00:19:33this isn't gonna become a just tool tour
00:19:36where I'm saying, use this tool, use that tool.
00:19:38The point is these tools and the best ones to use change
00:19:40literally all the time,
00:19:42which is why I did the video yesterday,
00:19:43which I showed off Hicksfield MCP,
00:19:46which is like a one-stop shop that gives you all these tools.
00:19:49The point being is if you are someone
00:19:52who's gonna be operating in a multimodal environment,
00:19:55so you are gonna do videos and you are going to do images,
00:19:58just understand you do need to be on top of like,
00:20:01what is the best one?
00:20:02'Cause this literally does change from week to week.
00:20:04But even though the tools change, the fundamentals remain.
00:20:07So if I'm able to create a system
00:20:08where I consistently create videos
00:20:11that are on-brand and on-voice for VO3,
00:20:14it's gonna be the same if I'm using Cling or CDance.
00:20:16I'm just gonna have to change the API up a little bit
00:20:18if I'm going a direct route.
00:20:20Same thing with images.
00:20:22The prompts that work in Nano Banana Pro
00:20:24are gonna work inside of GPT.
00:20:25So the real bottleneck, again, isn't the tool themselves.
00:20:29It's getting that brand and getting that voice.
00:20:32That's what matters.
00:20:35And so to do this, what we want
00:20:36is we want repeatable prompt templates
00:20:38that we can execute for images in videos.
00:20:41Specifically, if we can do them in JSON format,
00:20:43which is always nice to use with AI, that is best.
00:20:45So I have an actual skill inside of Clod Code that I use
00:20:48where I can feed it any sort of image as a reference image
00:20:52and essentially turns that reference image
00:20:54into a JSON prompt,
00:20:56which I can then have it edit with natural language.
00:20:58So if I fed it this image, it breaks it down into JSON.
00:21:01I could say, hey, change the text, change the background,
00:21:04change the actual like main character.
00:21:07It's gonna do that while keeping everything else the same.
00:21:09So I'm saying use the JSON prompt generator skill
00:21:11on this image and show me the JSON you created.
00:21:14And this is the image I gave it.
00:21:16And here's the JSON prompt it created.
00:21:20Now, what is the value of turning this into the JSON prompt?
00:21:23Well, let's say as part of my brand voice,
00:21:26I wanted to create images that were in this style
00:21:31by including this as a reference image
00:21:34and this JSON prompt.
00:21:36I can now consistently create images,
00:21:40create media that fit this brand and fit the style
00:21:43no matter what I have the copy be.
00:21:46And if I apply that same process to, for example,
00:21:49the rest of the carousel or whatever images I create,
00:21:53I have a system by which I can have consistent designs
00:21:56that look the same in my voice
00:21:58and have a certain level of taste
00:22:01when it comes to these multimodal assets, images and video.
00:22:04Same thing goes for video.
00:22:05So if I was trying to set up
00:22:06my own sort of carousel system from scratch,
00:22:08and let's say I hadn't created those visuals
00:22:10or brand voice first,
00:22:12what all you need to do is find some inspiration.
00:22:14So if I go on somewhere like Instagram
00:22:16and I search for carousel posts,
00:22:18I'm going to then see a bunch of really good carousel posts.
00:22:23And so what am I going to do?
00:22:25Well, I'm going to find the sort of carousel post
00:22:27I like that I think makes sense for my brand.
00:22:30I'm then just going to click on it,
00:22:32take a screenshot of it, dump it into Claude code,
00:22:36run that same JSON prompt generator skill you just did.
00:22:40And now I have a base from which I can edit it
00:22:42because, hey, I don't want to copy this exactly.
00:22:45But if I have that edited JSON prompt
00:22:47and the reference image,
00:22:48and I now then send that
00:22:50to something like Nano Banana Pro or GPT images too,
00:22:53I can now, again, use this as a base
00:22:56and turn it into something that makes sense for me.
00:22:58So I'm not starting from scratch.
00:22:59I'm going to be looking at stuff
00:23:00that has actually worked for other people.
00:23:02I mean, this thing has 44,000 likes.
00:23:04So why wouldn't I go out there and take inspiration that way?
00:23:08You would then combine that image
00:23:09with the actual sort of copy voice sort of things
00:23:13we've been talking about for the last few levels.
00:23:17And with those two things together,
00:23:18you would be able to create a carousel machine.
00:23:22And the real unlock here is building a fleet
00:23:24of these reusable JSON prompt templates,
00:23:26because that was one template I showed you for one carousel.
00:23:30What if we had 30 of those?
00:23:31And like you saw before in level three,
00:23:33what if we now have an automation that grabs information,
00:23:36grabs some sort of data
00:23:37that we want to talk about in a carousel format?
00:23:39Well, it's not a stretch to then automate that process
00:23:41where every day it grabs the information.
00:23:44It then creates copy in my voice.
00:23:46It then grabs one of the random 30 carousel templates
00:23:49that we've proven to work and then creates content for me.
00:23:53It's all automated.
00:23:54It's all in my voice.
00:23:55It's consistent.
00:23:56And if you dial this in, in the way I've showed you,
00:24:00it's not AI slop.
00:24:01Yet that is a pretty robust marketing machine right there.
00:24:04Automatic carousels every single day.
00:24:07It's not a small thing.
00:24:09And these other skills that go along with it
00:24:10of picking creative direction before generating
00:24:12and quality filtering AI output for tells,
00:24:15if you've done everything you need to do
00:24:16in levels one, two, and three,
00:24:18which is getting your voice locked in,
00:24:20that's kind of just automatic.
00:24:21And on top of that,
00:24:22you're gonna avoid this trap of poor AI outputs
00:24:24because you're always building off reference images
00:24:27that you've either done yourself and already work,
00:24:30or you find images that you like and you build off of them.
00:24:34We're never just building something from scratch
00:24:35and hoping it comes out well or actually converts.
00:24:40And that lets us to begin to look at level five,
00:24:42because at this point you have the voice.
00:24:45You're able to do it in text.
00:24:46You're able to do it in images.
00:24:47You're able to do it in video.
00:24:49Now it's time for distribution.
00:24:51It's time to bring this all together
00:24:54and have some sort of chain
00:24:55that goes from ideation to distribution,
00:24:57and all of it makes sense from beginning to end.
00:24:59So at level five, we're talking about scale and repurposing.
00:25:01How can I create one piece of content in one place
00:25:04and turn that into multiple pieces of content
00:25:06on multiple platforms and even different mediums?
00:25:09For example, I have a skill called my content cascade skill,
00:25:12which will take a YouTube video that I've created
00:25:14and turn it into a LinkedIn post,
00:25:17turn it into a Twitter thread, turn it into a blog post,
00:25:21and then turn it into ideas for short form content.
00:25:24This makes scaling so much easier
00:25:26because the idea that you're gonna create
00:25:27multiple pieces of marketing material
00:25:29on multiple platforms from scratch is just unrealistic.
00:25:33You're not going to do that.
00:25:34So you need something like this to boost you.
00:25:38As an example, I told Claude Code
00:25:39to run the content cascade skill on my latest YouTube video,
00:25:43and it created the blog post for me.
00:25:46It created the Twitter post for me,
00:25:48and then it also created an outline for the short form stuff.
00:25:51It then logs it all as super base,
00:25:53and in one command, I can actually push it live.
00:25:55The point of this is you need some sort of system like this
00:25:59if you wanna scale because in no world
00:26:01are you gonna do all that manually.
00:26:02Now, the one thing we have to worry about here
00:26:04is the voice variance for each platform
00:26:06because there are some differences
00:26:08between if we're on our own personal blog
00:26:09versus LinkedIn versus Twitter or any other place.
00:26:13Now, the good news is you're already 90% of the way there.
00:26:16If you're at level five,
00:26:17you've already locked in your voice.
00:26:18Now, all we have to do is, in essence,
00:26:21kind of repeat the process of level two,
00:26:25but it's a little easier
00:26:26because remember, I gave you this example,
00:26:28which is the template for my blog posts,
00:26:30and I said, "Hey, use this template
00:26:32"and then feed it examples of writing you like,
00:26:35"whether it's yours or someone else's."
00:26:37Well, we're just going to do the same thing,
00:26:40except it's not gonna be the blog.
00:26:41It will be LinkedIn or Twitter or whatever,
00:26:44and so you're then gonna tell Claude Code,
00:26:46"Here's the voice sheet we've been working.
00:26:48"Here's some examples of Twitter post, LinkedIn, whatever.
00:26:52"Can you now create a skill for LinkedIn writing
00:26:57"or Twitter writing?
00:26:58"That maintains my voice, but as you know,
00:27:00"those Twitter idiosyncrasies,"
00:27:02or whatever platform we're mentioning,
00:27:04and by doing that, you're gonna create a series of skills
00:27:06that allow you to take any piece of content
00:27:08and repurpose it in other places,
00:27:09and it has a specific voice for that platform.
00:27:13Now, these voice variants can get tricky,
00:27:15so this isn't just a one-and-done thing,
00:27:17just like anything else.
00:27:18You're gonna have to tweak it over time
00:27:19and see how it does in terms of engagements
00:27:21or whatever metrics you're paying attention to,
00:27:23but doing this will let us avoid the trap
00:27:25of thinking each platform has the exact same voice,
00:27:29and we can just repeat the same thing everywhere.
00:27:31That's not the case, and if it's text,
00:27:33and this is where you're gonna get the most value out of this,
00:27:35it's very easy to do this.
00:27:36We're literally talking about creating just a couple skills,
00:27:39but above all, we wanna make sure
00:27:40we just never lose that voice,
00:27:42and we never lose our brand.
00:27:44As we try to add more and more to it
00:27:46and send it in different places,
00:27:48again, there's always the chance
00:27:49that you're gonna dilute it,
00:27:50but if you're on top of this process
00:27:52and making sure you actually go over the outputs,
00:27:55you're gonna be okay,
00:27:56but once you have that voice locked in,
00:27:57that's when we can start talking about level six,
00:28:00and level six and level seven is where we begin
00:28:03to step away a little bit from the controls,
00:28:06and we give AI even more power
00:28:08to run these things autonomously,
00:28:10because in level six, this is where we go
00:28:13into the automation piece.
00:28:15This is where we start using things like loops
00:28:17and schedules and cron jobs,
00:28:19so you don't have to execute all these skills automatically.
00:28:22We already talked about this a little bit earlier.
00:28:24Remember this idea of, hey, checking GitHub every morning,
00:28:28doing the analysis, setting it off to GPT,
00:28:30and then creating these deliverables?
00:28:32Again, we can automate all of that.
00:28:33Now, how you decide to automate that is really up to you.
00:28:36I think the easiest way to do it,
00:28:38if you wanna stay inside the Clod code ecosystem,
00:28:41is simply by scheduling it,
00:28:42and we can schedule things by simply doing forward slash
00:28:45schedule and then prompting what we want,
00:28:47so I can do forward slash schedule and say,
00:28:50let's run the carousel skill every morning at 10 a.m.
00:28:54This is also something you have the ability
00:28:55to do through the desktop app,
00:28:57so if I go on the desktop app, I can go to routines,
00:28:59select the new routine, and decide if it's either
00:29:02gonna be local or remote.
00:29:03And understand, this isn't an all or nothing thing.
00:29:06I can have some portions of it be automated.
00:29:09Maybe I just have the ideation piece automated,
00:29:11where every morning it brings in all that information to me
00:29:14from whether that's GitHub or Twitter or YouTube or whatever,
00:29:16or I automate the content creation part of it.
00:29:19The one thing you don't wanna automate,
00:29:20I think, is the posting process, which some people do.
00:29:25Some people go for a crazy volume approach.
00:29:27I have my own things to say about that,
00:29:31but I would always caution you to never let AI post for you
00:29:36if you haven't had actual eyes on
00:29:37to the actual deliverables created.
00:29:39Now, the skills to master at this level
00:29:41are pretty self-explanatory, and that's figuring out,
00:29:43hey, how am I actually going to schedule this thing?
00:29:45Is it gonna be a local task?
00:29:46Is it gonna be a remote task?
00:29:48Which portions of this process am I going to automate?
00:29:51And also having some sort of taste checkpoints in the flow.
00:29:54This is really important, I think,
00:29:55when we talk about image generation and video stuff.
00:29:57And I think carousels are a great example,
00:29:59because if I'm creating sort of the body slides
00:30:02for, say, a carousel, this is slide number one,
00:30:05and then we have slide number two.
00:30:06They're very similar, but I wouldn't wanna just send AI off
00:30:11and just have it create all of them at once.
00:30:12I wanna see this slide.
00:30:15You know, I wanna see this slide first.
00:30:17I wanna make sure this is exact
00:30:19before I go ahead and let it create slides
00:30:21two, three, four, five, and six.
00:30:23So if I was automating some portion of it,
00:30:25you do need to figure out at which place in the process
00:30:28do you need to step in and give a thumbs up
00:30:30before the rest of it kicks off.
00:30:31And again, this is very use case dependent.
00:30:33Now, the trap at this stage is you go full AI
00:30:38and you just let it run everything
00:30:39and you don't have any of those checks.
00:30:41And this is really dangerous,
00:30:42especially if you're trying to build
00:30:43a certain personal brand or this is your business
00:30:45because it's all gonna be a reflection of you,
00:30:48whether AI built it or not.
00:30:49Now, level six is where you should aspire to eventually live.
00:30:52This is where you have your voice locked on,
00:30:54you got your brand locked in,
00:30:55you're automating the things you want to automate,
00:30:57and you're actually scaling it up.
00:30:59Because when we talk about level seven,
00:31:01it's essentially level six on steroids,
00:31:03but it's very easy to lose the plot.
00:31:05Because level seven is the autonomous agent,
00:31:08and this is where we're kind of letting AI drive the ship.
00:31:11And to be totally honest,
00:31:12I don't think we're really here yet
00:31:14in terms of the technology to execute this well enough.
00:31:17Now, there are people doing this,
00:31:19and a lot of people who are doing this
00:31:20have moved into the AI avatar space.
00:31:23So we're talking a lot of this like, hey, Jen stuff.
00:31:27And there are channels out there who are doing this,
00:31:29and I won't name any names, to like a pretty wild degree,
00:31:32where they are essentially doing everything
00:31:35we've talked about, but in an autonomous loop
00:31:38that's pushing out six to seven,
00:31:41sometimes I've seen up to eight,
00:31:42like long form YouTube videos a day,
00:31:46where they have an automated system
00:31:48that's constantly scraping the internet,
00:31:50it's grabbing the information,
00:31:51it's creating scripts in their voice,
00:31:54it's then recording a video with some sort of technology
00:31:57like, hey, Jen, so it sounds like them,
00:31:59it uses their voice, it creates the video,
00:32:02edits it itself, creates a thumbnail itself,
00:32:04posts it on YouTube, repeat, repeat, repeat.
00:32:07And the play here is through pure volume,
00:32:10they generate some level of value,
00:32:11because if I post 900 videos a day,
00:32:13at least a couple of them are gonna be hits,
00:32:15even if 99% is trash, right?
00:32:18You get enough hits, you get enough views,
00:32:20you eventually make money.
00:32:22I would argue this is terrible
00:32:24for your personal brand in the long run,
00:32:26yet everyone wants to do this
00:32:27because everyone and their mother apparently wants to get
00:32:29an AI avatar 'cause they're afraid
00:32:31to put their face out there.
00:32:31Just, I would advise against it.
00:32:34But I think it's important to understand
00:32:36how this works and to be able to do it yourself in theory,
00:32:40because there are a lot of skills.
00:32:42If you can do this, you know how AI works,
00:32:44you know how to do your voice, you know how to automate,
00:32:45you know how to like grab data and turn it into yours.
00:32:49But I would say the actual execution here,
00:32:53it's just not here yet.
00:32:54Now I will say the caveat to this is,
00:32:56if we're talking about pure text,
00:32:58just like text stuff, whether that's blogs or whatever,
00:33:03it actually is here because it's really easy to scale text
00:33:06to the point that like Amazon has a limit
00:33:10to the amount of books you can publish per day
00:33:11because this is what people are doing.
00:33:13They've essentially taken everything I've told you here,
00:33:15but are doing it in a literary format to an insane degree.
00:33:19So text, yeah, you can scale here.
00:33:22Video, if you're able to do this well,
00:33:23is probably the highest leverage one,
00:33:27but because the text just isn't there yet,
00:33:28I don't think you should do it.
00:33:29All's I have to say is I think you should understand
00:33:33sort of the architecture of how a system like this would work
00:33:36but I kind of advise against it
00:33:38because I think there is a law of diminishing returns
00:33:41and in the end, you kind of end up hurting
00:33:42your own personal brand, but no, it exists
00:33:46because this technology is only getting better
00:33:48and I don't think we're far away
00:33:49from being able to execute this.
00:33:52So that's where I'm gonna leave you guys today.
00:33:54I hope you were able to get something out of this,
00:33:56especially the idea of taste
00:33:59in locking on your brand voice.
00:34:01Those first couple levels that we talked about
00:34:03are the highest leverage and once you nail that,
00:34:06everything we talked about in the second half of video
00:34:08about scale and automation,
00:34:09it's gonna come somewhat naturally
00:34:12and a lot of that is the functions of the tools we use,
00:34:14but being able to have AI actually sound like you
00:34:17is by far the highest leverage thing you can master
00:34:21and once you master it in one medium,
00:34:23whatever that is for you,
00:34:24then it's just a matter of moving it to other mediums
00:34:27and being able to repurpose it
00:34:28and always relying on cloud code
00:34:30to sort of be the engine for these things
00:34:32and when I talked about just doing a stream of consciousness
00:34:35to cloud code and telling it your workflow
00:34:36is for how you do marketing today,
00:34:38how you do social media content today
00:34:41and breaking that into discrete tasks
00:34:43that we can turn into skills, which then can be automated,
00:34:46that's the other, I think, second highest leverage thing
00:34:48out of this whole video
00:34:49that I think you should implement right away.
00:34:51So as always, let me know what you thought.
00:34:54Make sure to check out Chase AI+.
00:34:56If you wanna get your hands on all the skills
00:34:58we talked about here and more,
00:35:00Besides that, I'll see you around.