Loading summary
Transcript8 lines
- [00:00]
A
I give Claude code 51 of my substack posts and in five minutes it told me exactly why some of my content works, why some of it doesn't. And then it generated an infinite number of new ideas that map to my winning formula. Now, that's just one skill of a much bigger content system that I've been building over the last number of months. It's an entire AI content team built in cloud code. It's actually 11 skills across five different layers of content. It can research your audience, analyze what content works, draft posts for LinkedIn newsletter, YouTube transcripts, and then it can actually improve itself every single month. On this episode, I'm going to give you a look over my shoulder on how I use cloud code to create content, how I use my content team, and I'm going to give you one of the best skills for free. Let's get into that and more on this episode of Marketing against the Grain.
- [01:01]
B
Here's a quick word from HubSpot. HubSpot helped Tumblr solve a big problem. They needed to move fast to produce trending content. But their marketing team was stuck waiting on engineers to code every single email campaign. Now they use HubSpot's customer platform to email real time trending content to millions of users in just seconds. The impact three times more engagement, double the content creation. Want to move faster like Tumblr? Visit HubSpot.com.
- [01:41]
A
So we are going to give a look over my shoulder as I use my AI content team to create content now. It's going to be awesome. So what is this? This is really for people who want to start to see the possibility of cloud code. And if you are a system thinker, a system thinker, how you can use cloud code to be your super weapon, how you can use cloud code to supercharge anything that you're doing in work, if you can think about how to build systems with it and in it. And this system is all about content. So I'm going to give you a quick visual of the system that I'm going to bring you through today. This is over on my substack for people who want to check it out. And so this is the system we're going to go through today. That little starred one here is what I'm going to give you all for free. And so it has an orchestrator skill. So the orchestrator skill is going to actually use all the other skills, which makes it really easy to use the system. You can kind of just vibe with the Orchestrator skill and ask it to do things and it goes and uses all of the other skills for you. It makes using these systems really, really easy. You can build a content audience profile, which basically you want to figure out who you're creating the content for. Build a profile and make sure all of the content is tailored for that person. The system does that for you. You can build any number of writing styles, right? I build a writing style for me because I create a lot of content. But if you haven't created much content before, you can create a writing style around someone you like and then adjust it to your taste and, and your audience. It has really great research and so it has these content idea skills. It has a talking point extractor. It can go off and find viral talking points from trending topics, or you can give it content you like, a YouTube video, a PDF, and it will actually turn that into talking points that I can create content from. It has a lookalike content skill, my favorite and one of the ones that I'm going to give you today. You can just give it a data dump. I give it 51 of my substack posts. It creates a profile of winning content patterns and then goes and finds other content ideas that map to those patterns. Show you how it works. There's a lot going on here. It has content skills. This post enricher is so good. The post enricher is I give it an existing post and it enriches it with different parts of content and that works really well in the different platforms. Maybe a data point, maybe a story, maybe a case study, content creation. LinkedIn user and X. Now YouTube would be in here, but I only added that at midnight last night. I have no life. I need to actually really think about that. And then feedback loops. This is so important if you're a system thinker. Most people stop here. I actually have built an app. The app captures all of the content that the system has created and then actually allows me to input performance data. And then it takes that performance data and every month I run a review and it actually makes the skills better, right? System thinking. We just dropped the 11 AI skills I personally use to build my own content team. I'm giving you the full architecture. It doesn't just generate content. It studies your best work. It will research your content in real time. It will make drafts across every platform and it even updates itself on what's actually performing. Get it right now for free. Click the link in the description. I basically went into cloud code and then I say, let's start creating content. And that triggers the Orchestration skill. And so the orchestration skill allows me to see what's in here, so it picks up an audience profile. This is for the lookalike research skill. Basically this is 51 subsect posts and it's created a profile of winning patterns in my substack so I can actually go find other ideas that map to what my audience like. It's picked up my writing style. I've already created a writing style for me, but I can create an infinite number of writing styles. It's already picked up content research that it's done previously. It's picked up eight LinkedIn posts created. I'm going to show you that because then the app picks up all those posts with performance data and actually improves the system. The API keys are X, Perplexity, Fire, Crawl and OpenAI. And then in the automatic I have another agent which is Claude is managing all of this. Automatic basically is when I pick this option, the system will just work itself. So I will just come in every morning and the system will have created content using all of these different skills and I don't have to do anything. And that is what I'm building right now is an agentic team to just do it all for me. It tells me what you actually need. And a previous episode about cloud code with James, who's really great at this, the born marketer. This is really like vibing. And I was never a big fan of the kind of vibe marketing where it was workflow tools because it's not vibing, you have to actually drag and drop all the workflows together. This is not software. So I have a skill that builds this audience profile and basically says, hey, these are three things you need. But you don't need those three things because you can just iterate and refine in real time. And I didn't want the client domain, I just give it a target audience. I give it a primary content platform. I actually wanted X articles because I think X articles. I'm pretty interested in starting to create more content on X because of X articles. And so I want to actually start to do that because I really like long form content. And so it's building me that profile. I ask it to actually validate the research before it builds the profile. So you can see here it's asking me is this right? And it actually was really good. It picked up on this is a growth practitioner who's really obsessed with AI and making AI work within their company super accurate. And then once you confirm it creates this audience profile. Now this is Hard to actually see. It's creating the MD file. So it's going to create that content audience MD file. That and the writing style power the entire system. Because every other skill looks to see what content profile and writing style that you want loaded to create content. All right, we're going to go in and see the HTML. I purposely create HTML for showing how to do these things. The system does not need the HTML file, but it's easier for me to show you all what it creates. This is my AI growth practitioner. This is specifically for X Practitioner. This is the industry. And you can see here it's created audience identity jobs to be done. The kind of pain points pilot. Hell, I love this, by the way, this audience profile. I have perfected this. This is really good. It took me 12 months to kind of go back and forth. So I have some really dope stuff here. So the vocabulary library I find really good where you can say what they do and don't say all the kind of the way they speak, the emotional register and like validation hooks. It's like really good. This is not like an icp, right? This is actually content they react to. And it's all based upon research and engagement data. It's pretty lengthy, it's pretty awesome. And it's specifically for X articles, this Persona on X articles. You can see the amount of context that I'm given this system. Just when I want to create one type of content for one type of Persona. Now this system allows me to have 10, 20, 30 writing styles, 10, 20, 30 content profiles. And it can load anyone I ask for. And so it's really flexible. We are back in the orchestrator skill. The next thing you would be doing during onboarding is what writing styles do I have to use? So you could create a writing style. So the way the writing style skill would be used is I basically would go create a writing style for myself or for any creator. And what it does is it goes and scrapes the web, creates a file for that person, and then creates a writing style. Now, I've already created one for myself. I have a founder's voice model after Dharmesh, one of the best writers there is. I have my own one here. So let me again show you maybe just how that looks so you can see how this is all kind of shaping up. So I have a style card, right? And so it creates me, this whole writing style. It goes and scrapes the web for my content, creates a cool writing style captured from existing content, tells you what it's sampled. Primarily platform is a substack. It does struggle with LinkedIn. And so what you could do for your own system and I have done started doing this is you could just have to go and like export your own files to upload because it tries to get LinkedIn from perplexity. Fire Crawl doesn't have a great time getting LinkedIn posts. So it has a lot of substack. Why has it not captured like X? Because I don't create much on X at the moment. But I actually want to change that because I do like the articles. So this is like primarily focused on substack. So if I was doing this, I would do a writing style per platform, right? That's how I would think about it. So this is like my substack one. So that's that. Right? So we've done content profile and so what does the talking point one even look like? So I'm gonna show you how it works and I'm gonna do the before and after because I already have some talking points created. What I'm gonna do is show you how I run it and then I'm gonna give you the example of the output. But the output won't match the ask I'm just made because I've done that. It's a prior one that I had did last night. But let's create some talking points for marketeers who are trying to adopt AI now. I could also give it existing content and it will take existing content and do the same thing. But I'm going to show you how it looks. I have talking points and content ideas in separate folders. This is your lookalike skill that we're going to show. So this is talking points. And so we're going to say, hey, we did some of these already. So you can see it has viral talking points, it dates them. So let's open this one, which was really around cloud code for marketing. And so again, look at this, right? It maps to the audience. So at this point it pulled in a VP of demand generation and then it looked across Reddit, it looked across X, it looked across Perplexity, and it pulled out these kind of talking points. And so it has here like your marketing team is building the tools it used to buy, it has why skilled beat prompts for demand gen work. This is a good potential post. If your team is still copying and pasting prompts from a notion doc, you are one model update away from losing half an hour of output quality every time Claude changes. And then it gives you core insights. Some of the reasons it Believes that's true gives you the application of this. It basically categorizes them. So these are educational posts, things where you teach someone. These are data nuggets. So James who came on the Great Vibe Marketer, he has a post over here where Greg Eisenberg talked about the ability to create agents to do a bunch of marketing talks about the action, the context, the value. These are all kind of data points. But I want to show you down here spicy take. Your job title marketer is split in two. In 2026, there are two types of people in marketing departments go to market operators who run systems and everyone else who is still waiting for engineering tickets. So that's a pretty cool one, right? Like that's a good spicy take. This is the impetus of a good post. So these are just like creative sparks for me. I like to go through this and I'm like, oh, these are, you know, interesting ones. Your outbound function is mostly the most replaceable thing, your demand gen budget. Then we have these kind of story sparks. Story sparks are like bringing to life something through a story. So you can see here, and this is kind of built around a story. At 10pm she give Claude code one directive by idiom, 34 new landing pages built live on the site. I would tell you if you use that hook on social, that will do really well. And so this is pretty cool. That's the output for you on the talking points. So that's viral talking points. Go scour the webs or take existing content and turn them into talking points. So let's go back here and we'll run the skill that I'm going to give away because it's also pretty amazing.
- [13:05]
B
Before we continue, let me tell you about a podcast I love. Marketing School. Marketing legends Neil Patel and Eric Su bring you daily actionable digital marketing lessons learned from years in the trenches. Marketing School delivers bite sized marketing wisdom you can implement immediately. Whether you have a new website or you're an established business. You'll learn the latest in SEO, content marketing, social media, email marketing, conversion optimization and general online marketing strategies that work. Today. They just did a great episode called 10 SEO Lessons that Still Work in the AI Era. Listen to marketing school wherever you get your podcasts.
- [13:48]
A
Now let's look at the other one that you're going to get access to, which is the Lookalike one, which is like one of my favorites. The way the Lookalike one runs is I would go and say here's a data dump of my LinkedIn post. Here's a Data dump of my substack posts. Here's a data dump of my top performant X articles. Here's a data dump of my best thought leadership posts on my blog. And I would upload that messy file to this skill that you're going to get access to. And what it does is it takes that mess and if it has performance data in there, it will look at the top 30% of your best performing posts because it wants to extract winning patterns from the best performing posts. If there is no performance data in there, it will just look across the data you have given it because it can't tell which ones perform better than others. And it first of all creates a profile because it wants to work from a profile. And so let me show you the output again. In HTML, the system all works in ND files. But for the purposes of this video I wanted to be able to visually show you how this works. So it creates my substack content profile. And so what it's broken out is it's basically said, hey, like this is the winning formula. This is the top performing clusters of content. Look how cool it is here tells me the different kind of clusters of content across my substack. It talks about the structural DNA of my substack. So like typically how long as opposed to different sections, the structural type, paragraph style, embeddable prompts work really well. If you haven't checked out my substack, you should. It's awesome. The hook formula, the emotional playbook and format that works, always really important to say like what doesn't work. And so from that messy data that you have given it, it creates that profile and then it goes to take that research and gives you a bunch of ideas. And so we're going to go back here into our content system. We have these ideas. So you can see here, I ran this last night for my substack. Look at this. How good is this? Right? These are ideas for substack again, always tailored to an audience. So it picks up a content profile and will basically say what content profile do you want to use? If you remember, the content profile is the audience. And then it will say this was the winning formula. And here are ideas again. I create content. I use a lot of the content system for ideation research, first drafts and actually I have a pretty cool skill to show you that actually helps me enrich some of my content. And so here it has one about meta prompting because when it looked at my topic clusters it could see that that was a really great topic for me is like People really still want to understand how can I get more out of these models? The AI boss era. How to run a fleet of agents across your marketing team. This is a good one. Actually your AI has been agreeing with you too much. Here's how to fix it. Pod models.
- [16:46]
B
Wow.
- [16:46]
A
I replace the marketing teams. This is a really good one. We were in the era of functional teams and now we are in the era of pod models. And I think you have to structurally rethink the way you think about team structure. The 6x Gap why some marketers get insane AI results and most don't. A study of 9000 AI users find a 6x productivity between the effective and surface level AI. Wow. This is super cool. Wow. Data hook. So it tells me which they are by the way. You can see it kind of frames it up for me. AI frameworks. It's a data hook. It should be a four step tutorial. And then this is the emotional driver. How to simulate your customer's reaction before you launch it. I've done that one. So that's interesting. That's picked up that one. Why your LinkedIn prompt fails in every other channel. Okay, yeah, that's a good one. Actually what it's really saying is there's like this quirk where you say create a Great post for LinkedIn to the model and the model goes and says well this is what a great post looks like on LinkedIn. And. And LinkedIn for the most part sucks. Well, not sucks, but there's a lot of bad content on there. So you're going to get a ton of bad content from that. So you can see how good these are. I'm actually reading through some of them myself. I only read through the first five. The most underused AI technique in marketing. Every AI output you've ever seen was a first draft. The AI never reviewed its own work beforehand. And I mean I kind of feel like this is better than me. It's come up with so much good stuff. Before we go to market push. I fed the entire strategy into my AI model and asked you to find every way fail. That's very good. I found six problems we haven't seen here are the exact process and then it does like a nice little summary here. But you can see here gives me the title the key pattern. Right. I always like to do pattern match and I think content is kind of modular. It's patterns tutorial and then the trend and signal. It's pretty good. And so where's the one I really like? Pods replacing teams feature of teams Framework and HubSpot story. Yeah. POD defaults for high performance. Yeah. Wow. So much great content. Too much great content. But. But as we said, we are building an agentic team and the agentic team is going to do this autonomously. And because you're subscribed to the best YouTube channel on all this stuff, you will get that first when I built it. So let's go back to our graphic. I'm going to like show you the post enricher because it's dope. This works the way you think it would work. But I will do a separate video just on content creation because too many people will use these. Cut and paste. That's not how you do that. Right. So I think I will do a standalone video just on feedback loops because I think that's pretty important. But given I want to make sure that you can get through this and there's a lot to it. We're going to go add a sequence here in terms of how the system should be used. But again, I keep talking about this. This is not like using software. It's flexible. And so when you've extracted talking points, obviously the system would then say, here's your talking points. And you can basically just point at one and say, okay, like turn this into a draft LinkedIn post. Turn this into a draft YouTube transcript. Turn this into a draft substack. Turn this into a draft X article. What I want to do is because I want to show you the post enricher. I'm going to try to see if I can just give it a talking point and ask it to enrich that post. We have data enrichment. I have built content enrichment tools and I want to show you how they work. And I think I will do a standalone episode. I have just decided to do a standalone episode on content enrichment because it's super dope. But I'm going to show you this. Let's see which one of these we like. Educational workflow redesign is a real 80% of AI adoption domain expertise, speech prompt and skill. Your job is now AI agent and management. Spicy Take roi. Gap is a coordination problem, not a measurement problem. Agencies need to rebrand proactively. Reality gap standards data nugget 91 adoption vs 6% 0 employee AI marketing team in the room where nobody uses AI. Let's see what the story spark is. Okay, so let's do a first draft of a LinkedIn post for the Story Spark. Zero employees using the LinkedIn skill and then we'll use the post enricher because I'm kind of doing that sequence. So I want to show you the content enrichment tools need a first draft to enrich and so trying to enrich just the idea is kind of hard. All right, so it's done a post. Here's your LinkedIn post. So remember this is a story spark, so it should actually feel like a story. Also this is a draft, an idea and then you make much better. He didn't hire a marketing team, he onboarded one. So it's actually taken someone's real story. Documented this in February he built a full marketing function with 0 employees. 4 AI roles fully integrated on top of each other. Ads, analysts, content, user, blah blah. Each role had to find inputs, outputs and handoffs the whole stack run. So if I was doing this, I think the cool version of this is marketing teams of one and gets like multiple examples. So for example, there's one person in Claude running a lot of their marketing. So I'm going to get my transcript that I've done here and actually start to like format this into a kind of marketing team of one story enrichment. So it does have enrichment. But I'll show you how this works. The party gets buried in the excitement. He didn't just pick four AI tools, he structured each agent like an employee in board and document a defined scope yet so it tells you that's the gap between market and running. Blah blah. All right, so like it's a pretty good first draft. Again, first draft. When you look at this, I do not use this to post content. And so these are all ideas and now it's pulled in my enrichment here. So enrichment should actually get me modules, case studies, stories, examples, enrichments. Let's make sure we use the post enrichment skill. Here we go. All right, we're back. So this is the beginnings of an enrichment tool that I'm building. So it looks for ways to enrich that content. 2022, Jeff Bose sent an internal memo that changed how Amazon built software. Every team had to expose their data and capabilities through clean surface interfaces Divine and blah blah. 20 years later, the same principles are showing up in AI marketing stacks. The agent can that run reliably in production are the ones built with precise rules. Very cool connection. The failure mode isn't AI quality, it's the same failure Bezos was solved in the 2022 undefined roles that don't scale whether the work is very cool gives you a nice little case study here. But a platform running autonomously talks about the connection and then gives you a pretty good quote. Paul writes, sir, who runs the AI Marketing Institute Next gen marketers know that in order to deliver personalization experiences, modern consumers expect marked and must become smarter. Blah blah blah. Okay, so you can see the enrichment tool and I can easily just pick one of these and it will enrich that post. All right, it's not finished. It's saving that post but I can show you a different one. So remember I was really interested in that marketer role being split. I thought that was a pretty good post. I did do a first draft of that one. The marketer job split into this year. Most people haven't noticed. Very cool. It's actually done it as a LinkedIn post. It's so good. Spicy take again. Audience profile, hook pattern, contrarian punch enrichment type. Because all my stuff gets enriched enrichment source actually tells me where it was word count talking point and so it tells me what file I pulled from. It's on the other one now as well so you can see what it looks like. It's really good by the way. Right. The marketing jobs get into this year. Most people have noticed on one side go to market operators, they build systems inside cloud code, hello outbound engines running on 11 APIs, ad creative generators producing 100 various blah blah. On the other side, marketers waiting for the next platform to do it for them. And Chris Walker put the logic plainly in 2024, instead of having 100k SDRs doing it, what if we had 200 SDRs augment 8 people with AI blah blah blah. Very cool. And so it saves in there the other ones in here as well. He didn't hire a marketing team, he onboarded one. I have a hook skill and so I can improve the hook with that skill, much like I have a content enrichment skill. So obviously I would never ship this. I actually think the more interesting one is the rise of the one person marketing team. But there's some like really good stuff in here and here's the enrichment that we did. Now the last thing I want to close this out in the last minute is where people don't understand the importance of feedback loops because they're not systems builders, but systems builders do. And so what actually happens here is the system picks up all of the content in these apps. This one sucked as one of the worst performing posts. People did not like a product position, they're not into it. And so it picks up all the posts that I've created and then I can go here and I will be able to add the LinkedIn API at some point, but I don't have it added at the moment and so I have to manually add the data in here. But once I add the data, what actually happens is then I can just go and run a skill and the skill does review analysis of all of the performance so you can see what was created from the system, it can look at the performance and it actually updates all of my skills. So it's a skill that improves my skills. So everything I showed you will get updated and improved each month based upon what's doing well, what's not doing well. And so that is Claude code content system, your free skill that you get for free which is one of my favorite actually. It's pretty like cool. The lookalike content audience and if you don't have a lot of content then you can go collect other folks content that you like and give it that as a data file and it will go find those within patterns. You don't have to have your own substack, your own LinkedIn, your own X. You can go get that content from elsewhere. So that is the content system. There's a lot in this. I'm going to do some follow ups on the feedback loop, going to do a follow up on cloud code for content creation and how you could iterate really, really fast. All of that and more on a future episode of Marketing against the Green. So make sure you have subscribed and I will see you then. Foreign.
- [26:47]
B
You know, Kieran and I have been doing the podcast for a while now. We've been at this for a couple years. We love it. We could not be happier to be doing this, but we wanted to take things to the next level. We want to level up the impact we're having with marketing. It's the grain. So the next step of our journey is something we're really, really excited about. We're going to launch the Marketing against the Grain newsletter. And Marketing against the Grain newsletter is going to be amazing. If you are a marketing leader, practitioner, you're in the trenches doing marketing every day. This is for you. We're going to deliver right to your email inbox and you're going to get all the behind the scenes frameworks, practices, tutorials from us, from the guests we have on the show, and from people even beyond the podcast that we think are going to be helpful and and really have an impact on your day to day, week to week doing marketing. You're going to love it. It is something we've been talking about for a while. We're really excited to have it out in the world. We've already got a hundred thousand marketers who are on this newsletter. Please join. It's completely free. We'd love to have you as part of the Marketing against the Grain community. And it's easy. You can click the link in the description below, or you can head to marketingagainsthegrain.com subscribe.