My Claude Code marketing stack (It just works)

I sit down with Amir, who's back on the pod, and we walk through the full stack of taking a business idea from zero to a validated, A/B-tested landing page in a single session. I use Idea Browser's new MCP integration with Claude Code to pull project context, generate a lead magnet concept, design a landing page in Paper, and then wire up analytics and live experiments through HumbleLytics — all without writing a single line of front-end code manually. We cover the tools, the workflow, and why this stack creates massive arbitrage for marketers and builders right now. Timestamps 00:00 – Intro and Episode Preview 02:30 – Building a Growth Strategy with Idea Browser 06:10 – Designing Landing Pages in Paper 08:38 – Refining Copy, Layout, and Components in Paper 20:06 – Deploying Landing Page and Adding HumbleLytics Analytics 28:38 – Running A/B Experiment on the Headline 32:44 – The Arbitrage Opportunity and Closing Thoughts Links Mentioned: Amir’s Agentic Marketing Skill: https://startup-ideas-pod.link/amir_marketing_skill Key Points Idea Browser now connects to Claude Code as an MCP, letting you pull project context, growth strategies, and skills directly into the terminal for building and iterating on business ideas. Paper replaces the traditional Figma-to-developer handoff by letting you design, iterate, and refine landing pages visually — all connected to Claude Code so changes stay in sync. HumbleLytics enables no-code A/B experiments that dynamically update page content without deploying new code, so you can test headlines, CTAs, and layouts in real time. Storing performance context (A/B results, revenue data, growth metrics) back into Idea Browser compounds your results over time because every future decision is informed by past data. This full stack — Idea Browser, Paper, Claude Code, HumbleLytics — creates a significant arbitrage opportunity right now because almost nobody is using it at this level. The #1 tool to find startup ideas/trends - https://www.ideabrowser.com LCA helps Fortune 500s and fast-growing startups build their future - from Warner Music to Fortnite to Dropbox. We turn 'what if' into reality with AI, apps, and next-gen products https://latecheckout.agency/ The Vibe Marketer - Resources for people into vibe marketing/marketing with AI: https://www.thevibemarketer.com/ FIND ME ON SOCIAL X/Twitter: https://twitter.com/gregisenberg Instagram: https://instagram.com/gregisenberg/ LinkedIn: https://www.linkedin.com/in/gisenberg/ FIND AMIR ON SOCIAL Humblytics: https://humblytics.com/?via=community X/Twitter: https://x.com/amirmxt Youtube: https://www.youtube.com/@amirmxt

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Greg Isenberg?

  Amir is back on the pod. Thank you. By the end of this episode, what are people gonna learn?

• 0:06Speaker 01

  Today, we're going to cover three aspects of building a business, using new tools to actually build out landing pages, and then using other tools like Humbletics to actually create high converting landing pages and running experiments. So we're going to cover how to actually use Idea Browser to build the right amount of context and planning to come up with an idea and build a landing page behind it, validate it, refine the design using paper design, and then going into Humbletics and then running an experiment and tracking data to then come back and keep optimizing for conversions.

• 0:36Greg Isenberg?

  And you're going to commit to giving all the sauce. You're not holding back.

• 0:40Speaker 01

  Raw. We're gonna go, gonna go through everything, all the sauce. By the end of this, you're gonna learn what it takes to take an idea, validate it, refine it, build a landing page with a nice design that's not vibe-coded, and then get the right data to make some money.

• 0:53Greg Isenberg?

  Okay, because there's a lot of tutorials that sort of promise that, right? And then you end up with like a purple vibe-coded landing page.

• 1:00Speaker 01

  Do I ever let you down, Greg?

• 1:01Greg Isenberg?

  No, you don't. That's true.

• 1:02Speaker 01

  There you go.

• 1:04Greg Isenberg?

  Let's go, let's get into it.

• 1:05Speaker 01

  All right, so.

• 1:14Speaker 01

  Idea Browser came out with something really cool. I really like, which is basically you can now directly connect Idea Browser to Claude Code as an MCP. What's really interesting about it is, you know, you guys used to focus a lot on finding the right idea and then building it. But what was missing was how do you track the natural progression of how the business is evolving over time with the right context and documents? To then come back and reference to say, okay, cool, like, this is the idea I had. Here's how I started with ICP positioning. Now I want to come back and refine this. Or help me figure out kind of what to pitch for, or who to go after as my customers. So I was playing around with Jordan earlier today. We were on a live stream and we found this really cool idea, which was an AI sparring partner for B2B sales teams.

• 1:59Speaker 01

  The idea was how do we build a tool that kind of essentially helps reps practice on prospects on real time? And get them to essentially go through some simulations of being like, you know, going through a sales call and getting feedback on what they can improve on. We put together a quick landing page, we made some designs and then got some data in there and started tracking it. So for today's episode, we're gonna build on top of this, we're gonna refine the messaging, we're gonna go through some sections on the website, refine the copy and design, show you how to use paper for that, and then also, yeah, get some experiments and A B testing in there.

• 2:32Greg Isenberg?

  Cool.

• 2:33Speaker 01

  So what I, so what we have so far is we have this one specific file, which is like an offer definition and talk tracks. Tells you who your target customer is, the transformation, the offer, the value, competitive positioning. What's really cool is you guys also have skills in here, which you can use to then build on top of the idea. So I actually want to use the lead magnet legend here, this specific skill, to build the specific lead landing page site to kind of get someone to come and sign up and, you know, go through this process. So we're gonna open up terminal and we already have all the M C Ps connected. So if I go to M C P, you can see I have idea browser, I have paper, all that.

• 3:15Speaker 01

  We're gonna go through it. So I'm gonna say connect to idea, idea browser, M C P. Look at my A I B two B sparring partner project, pull the right context, then use the lead magnet skill to create a.

• 3:34Greg Isenberg?

  By the way, I just gotta say this because people are gonna hate on you for the comments, in the comments. You type your, your prompts, you don't use voice?

• 3:44Speaker 01

  Listen, I love Whisperflow, but tell me how I'm gonna do this while I'm doing the pod, okay? All right, just, just, you gotta bear with me here.

• 3:50Greg Isenberg?

  Okay, so when you're at home, are you using Whisperflow or are you still typing?

• 3:54Speaker 01

  I am. So I'm using, I usually use Whisperflow when I'm initially starting a project or some, some session. And it's like a long, it depends on honestly on the context of how much information I'm gonna put in the terminal. But yeah, no, I got Whisperflow, you know?

• 4:06Greg Isenberg?

  Okay.

• 4:06Speaker 01

  I'm, I'm, I'm up to date with the tools.

• 4:10Speaker 01

  So what we're doing right now is we've pulled the right project context. And so we have context right here that's built as a file. And then we're then, we also have a growth strategy that we worked on earlier. And then from there, we're going to build out a lead magnet specifically for this idea. So we have this AI B2B spying partner. What's really cool is you got to have, you guys have this like activity streak to keep building on kind of your business. And that's the biggest problem these days. Like everyone can build landing pages, everyone has ideas. How do you actually like know where to get customers? How do you actually grow it? And that's the next kind of gap that we're all trying to solve for right now.

• 4:49Greg Isenberg?

  Well, yeah, I think we've all tried the vibe coding tools. We've all built stuff.

• 4:53Speaker 01

  Yeah.

• 4:54Greg Isenberg?

  I do think that having the right idea is important. Finding the right niche is important. But you know, the other thing is just like, how do you get customers?

• 5:06Greg Isenberg?

  To honestly just give you confidence to keep going.

• 5:08Speaker 01

  Exactly.

• 5:09Greg Isenberg?

  Yeah.

• 5:09Speaker 01

  Yeah, yeah. And we're going to eventually get there. And I think you guys do a really good job of this. And I'm not like promoting, like I'm promoting Idea Browser, don't get me wrong, but I wasn't, you know, I'm actually using this tool myself now because that was kind of the biggest gap I had myself. Like with Humbletics was like, okay, cool. I have all these great ideas and I've been understanding who our customer is and trying to get customers through trial and error. And I was telling Jordan earlier, I was like, I wish I had Idea Browser sooner, right? So that I can come back and use these specific skills to, to understand, okay, cool. What is the right growth strategy? Because a lot of what you guys have has been fully refined. And we were going through this earlier today. It was interviewing me and asking me questions. And I was like, this is so impressive. I wish I had Humbletics for this specifically so that I can come back and say, okay, cool.

• 5:53Speaker 01

  Like Humbletics used to go after director of marketing in this specific field with this kind of messaging. And yeah, so kudos to you guys for building a really cool product with this. And I think a lot of people now can actually use this to build the idea and then also help build the business as well. So we going back to Claude Code, we were connected to MCP. We have built out the lead magnet, which is basically five objections that kill Fright software deals. So we actually niched down. We built this AI B2B spying partner in the Fright software industry. And it's going to say, you're going to want to create a PDF guide. Give them all, yep, this is great. Okay, cool. So we have the idea of the lead magnet and it should now actually save the file in here. Let's go back.

• 6:38Speaker 01

  Boom. Now we have the file. So it's really cool. Like you have context of the ideas it came with and it's added here. Now we're going to come back and use the scale. I gotta come back here. We have the, where is this? The landing page architect. So using, and basically what we're going to do now is for the people that don't know, Paper is, it's a really cool tool where essentially it's connected. So let me take a step back. Typically with Figma, you had designers creating static assets in Figma, landing page designs, and then heading off to engineering teams, right? The missing gap was now we have a lot of people that are building landing pages and websites with Claude Code directly in code and refining it through there and then losing

• 7:23Speaker 01

  track of like what they're iterating on and not having the ability to be able to kind of refine this. So with Paper, what you're able to do is actually have an interface connected to Claude Code where you're ideating, creating iterations, variations directly in the design and then coming up with like what direction you want to go with and then having it in code. So it's kind of like the intermediary step that we were missing before where people were just directly building in code.

• 7:50Greg Isenberg?

  And you know, you can use Paper. Are there any other competitors that people use? Or just, or just stick to Paper?

• 7:56Speaker 01

  I mean, so Figma just recently came out with their bidirectional ACP, MCP, ACP. I'm in your world, man. Yeah, MCP, where you can both design, like you can take a design and build it to code or code to design to go bidirectional, which this is essentially what Paper already is. But I just feel like the tooling and The interface and experience around paper is a lot nicer, in my opinion, and it just works a lot better. So we're now connected to, we've captured the lead magnet. We have the landing page scale architect being added as well with the offer that we want to have. And now we want to actually build out the page using paper. So while that's going through it, I want to talk about kind of how you can refine your designs as well.

• 8:42Speaker 01

  So oftentimes you see people with kind of these vibe coded designs and they're not entirely sure how to actually go with like the right direction of building out a design system. What I typically like to do is always give Claude a reference image of an existing design that I like. So for example, I'll go into an existing site or have some bookmarks of sites that I like. And I'll take the screenshots and I'll drop it into Claude and I'll say extrapolate like the key design elements from these pages and help me create a design system. So what you're seeing right now, for example, is a design system that I've created using Claude based on reference images. So that we can kind of reference back into for feature sessions. So if I want to come back and create a new section or a new component on my landing page,

• 9:26Speaker 01

  I can say reference the design style guide as like the basis as you're creating new components so that you have that consistency. Now I want to show an example, like again, a lot of people have this misconception that vibe coded design is bad. This has all been fully vibe coded. We built all these animations, all these components using Claude. It doesn't seem that way. It seems very polished, very well refined. And the way to do that is you refine it through paper and then give example components and illustrations from other websites or libraries to build on top of. So for example, I use this tool quite a lot. It's called Tailarch, Tailarch Pro. And they have a lot of great existing blocks and illustrations that you can use.

• 10:12Speaker 01

  And you can actually install these and use it as a reference. So I'll go back into paper and I'll show you kind of how, oh, nice. Right now we're seeing Claude code, design and build directly in paper on what this lead magnet looks like, which is

• 10:25Greg Isenberg?

  That's really cool.

• 10:26Speaker 01

  Really, really cool. I love the little kind of section by section it's building out.

• 10:33Speaker 01

  So I think this is perfect, actually. This is a great segue to that. So we're gonna let it build out the landing page. Then I'm gonna show you how you can actually use Tailarch as a reference point to then improve the design of this. So it goes from something like this to something a little bit more tailored like this, for example.

• 10:51Greg Isenberg?

  By the way, there's an ant on the table just there. Do we let the ant watch your demonstration? Or do we sort of move him aside? What do you think?

• 11:05Speaker 01

  I think we'll just put him on... I mean, where do we put him?

• 11:08Greg Isenberg?

  I mean, I have nothing against the ant.

• 11:11Speaker 01

  Nothing against it. Yeah, he's chilling.

• 11:13Greg Isenberg?

  He's chilling.

• 11:13Speaker 01

  He's getting the direct sauce.

• 11:14Greg Isenberg?

  Yeah, we can give the ant some sauce, you know?

• 11:17Speaker 01

  He's a napkin, in case.

• 11:18Greg Isenberg?

  Yeah.

• 11:23Speaker 01

  Okay, cool. So we're building out the section, which is pretty cool. Now, a lot of people say, "Okay, cool. Why don't I just build this directly in code? Why do I have to actually use paper?" The idea is that you can use paper to help build out different variations. So if I want to try different layouts, I can just do it directly in here, make some refinements myself, especially for designers. I think this is super powerful for designers that want to just jump in and start changing things themselves without having to use Claude code again or existing components they have. So we'll let this finish. Then I want to show you some examples of what I have also created in paper to show you what's possible if you keep refining it and go through it. Perfect. So we're all done here. Now, what does this look like in actual, real practice if you spend time refining this? So this is all designed in paper.

• 12:09Speaker 01

  I've given the design style guide I have in HumbleEtics. I've said, "Here's the components you have, and use these components to refine it." So I actually build out some of these sections in here and then port it over to code. Or I even use it to create static assets that I want to use in ads, in thumbnails, whatever it is. This is all purely built using Claude. Now, how do you go back to actually refine the design here? So what I do typically is I'll go into Tailarch and find a section that I like. So we have this content area section. I'll go into...

• 12:47Speaker 01

  Let's go into content.

• 12:50Speaker 01

  Right here. Okay, cool. I like this. I like this right here. Or maybe, yeah, we'll use this one. Cool. So I'll go back in the terminal and I'll say, "Install this Tailarch component and use it for the content section in the lead magnet." First, design it in Paper. And then we just need to pull the API key for Tailarch. Cool.

• 13:14Greg Isenberg?

  I've actually never heard of Tailarch.

• 13:16Speaker 01

  Really?

• 13:16Greg Isenberg?

  Yeah.

• 13:17Speaker 01

  Yeah, it's just another UI library. I mean, I'm not sure who's behind it. I think it's like an indie founder or something like that.

• 13:23Greg Isenberg?

  Looks clean.

• 13:24Speaker 01

  Yeah, super clean. It's the way to go, you know, kudos or shout out to whoever built this. I think it's an indie person. Whoever it is, you've done a great job. Keep doing it. Yeah, I'm a fan. I'm a big fan.

• 13:38Speaker 01

  They got taste, that's for sure.

• 13:41Speaker 01

  So what we're doing right now, we're just installing Tailarch so that we can like install the right components here. And yeah, what's special about this is that you can take those components that I showed you earlier. So if we're gonna, for example, we're gonna do the content one.

• 13:55Speaker 01

  This one specifically, and I can just drop it in as well. A screenshot to say like, "Install this. Use this component."

• 14:03Speaker 01

  And then I'll go back into Paper and say like, "Use it for this section here."

• 14:07Greg Isenberg?

  What's also interesting is, you know, showing Tailarch, showing Idea Browser, the MCP. It's sort of a glimpse into like the future of software, right? Like you're not, in the past, we would go and just use SaaS, right? Now it's like, the work is getting done in the terminal.

• 14:24Speaker 01

  Nailed it. Again, it's becoming the interface for work for everything. Yeah. If you remember our first episode, like cursor was into, yeah, like.

• 14:30Greg Isenberg?

  You were, you were right.

• 14:32Speaker 01

  Yeah.

• 14:32Greg Isenberg?

  So you said, I think your exact words were something like.

• 14:37Speaker 01

  Like, yeah, but I think the mistake I made was I said cursor.

• 14:41Greg Isenberg?

  That's what you said.

• 14:42Speaker 01

  Yeah, cursor is the interface of work or it's the future of the interface of work. I would say now Terminal, the Terminal is the interface of work. We were, you know, we were very, we were early. We were early.

• 14:51Greg Isenberg?

  We were early.

• 14:52Speaker 01

  We were early. Like, we're like, yeah, markdowns are gonna become a thing. You're gonna do all your work in Cursor, all of it in Terminal.

• 14:58Greg Isenberg?

  People thought we were crazy.

• 14:59Speaker 01

  Yeah, really. Yeah, like. Here we are today, man.

• 15:02Greg Isenberg?

  And we're showing you this and someone probably listening to this is probably like, "You guys are crazy."

• 15:08Speaker 01

  Hundred percent. You know, another thing, for example, was like you're now seeing people building out markdown pages on the websites to make it easier for agents to access, right? You look at what's happening now, people are giving agents wallets. What was going on here? They're giving out wallets, they're giving them emails, inboxes, stuff like that. So it's insane to see, um, yeah. Yeah, it's insane to see kind of the evolution of that. And we're seeing, actually, tangent.

• 15:41Speaker 01

  I have a thesis. The thesis is that more agents are gonna actually visit websites than humans.

• 15:48Greg Isenberg?

  I mean, that's, that's, I'm a hundred percent certain of that. It's not even, you know, I saw that Gartner had a research report that said 20% of commerce will, on the internet by 2030, will be agents, meaning like agents are buying things. You know, that, the, the tail, or what's the, how do I say this? Basically where it's all, where it's going is agents doing more and more things. So obviously, yeah, there's just, there's gonna be way more agents than humans. Therefore, there's gonna be way more consumption of agents of products.

• 16:23Speaker 01

  Yep. And it's like the multiplier effect too, right? Cause you have people running multiple open clause or multiple agents.

• 16:28Greg Isenberg?

  Exactly.

• 16:29Speaker 01

  On behalf of one person. And who was it? There was a report saying, I don't know if they were serious, but they're saying, "Yeah, we should put a tax on, we should put a tax on agents." Right? Because it's like, yeah, it's making us more productive. There's a whole.

• 16:45Greg Isenberg?

  Think about human beings, right? Like if you have a, if you, if you're running a company and you wanna hire someone, you have to pay payroll tax on that person, right? So of course then, you know, governments are gonna be like, "Well, it's kinda like you're hiring a person. Therefore, there should be some taxes that, you know, agents have to pay."

• 17:06Speaker 01

  There's an arbitrage opportunity right there, right? Like you, you as a single human now have an army of agents working for you on your behalf. There are multiplier effect of what your income is or whatever it is that you're doing, more productive.

• 17:18Speaker 01

  Yeah, whereas like, you know, you're making more money. Yeah, I wouldn't be surprised at some point, we have some agent tax on your taxes, like list out the agents you have. But um, but anyways, uh, back to kind of talking about the future of web. Yeah, like we're also seeing right now.

• 17:35Speaker 01

  More websites being built to be agent friendly, right? Fire crawl came, Firecall, uh, Cloudflare came out with their um, endpoint to be able to crawl websites, to give access to agents, to be able to get the data they need. And, you know, we actually had our website built in Webflow, then we migrated to Framer, and now we've fully built in custom code. Why? Because we wanted to be able to use the agent as our CMS. So use Claude Code to directly update our CMS. And the ability to give access to other tools to like quickly ship changes, run tests, give access to our agents and MCPs, which is you can now kind of do with Webphone Framer, but it's not as open as you want it to be essentially.

• 18:16Greg Isenberg?

  Makes sense.

• 18:17Speaker 01

  Yeah, that's kind of the, the thought process behind it. Let's go back. Okay, nice. So yeah, we're getting a much nicer refined design here.

• 18:26Speaker 01

  And then if we wanted to, we can go back and like maybe find more components to drop. So um, let's look at like maybe.

• 18:34Greg Isenberg?

  So one of the things I really liked about your website, the Humbletics website, is like the animations were really good. How do you think about, like, if we wanted to like, update the designs here to be a little more animated? How would we do that?

• 18:48Speaker 01

  Yeah, so I, so right now, we're doing it right now. So I basically went into here. So this one has a very subtle animation. I copied this component and I dropped it back into Terminal in Claude Code. I said, hey, add this section and then add a subtle animation. And the goal, like, I always say, like, you don't want to over animate things. It's interesting to see how like agents respond to constraints and guardrails, especially around like, hey, make subtle changes. Subtle animation, subtle design refinements. I'm actually very intentional with that word when I use it with agents. Because like, if you say improve the design, it's so broad and generic. Whereas you say, hey, work on refining the design. Make sure you have consistent layouts and themes and keep it subtle enough to

• 19:31Speaker 01

  make sure there's like cohesiveness across the entire page. I find that prompt works a lot better than just saying, improve the design.

• 19:37Greg Isenberg?

  Keep it subtle, stupid.

• 19:38Speaker 01

  Yeah, keep it subtle. Exactly, exactly. There you go. So, you know.

• 19:45Speaker 01

  While that's happening, I'll say, when you're done, make the changes in the code and push it. And tell me the URL route. And add a section on the homepage to the lead magnet. Cool. We already have the site right now. And we're going to now add that lead magnet section. So we were actually earlier today with Jordan and I in the live stream. We're building out the home section, the hero section, and we started working on the steps below it. But now what we're going to do is build a lead magnet page using what we just covered in paper and the designs. And then, yeah, we're going to push it live to the website and then run some, get some analytics in there and run some experiments to see what drives more conversions.

• 20:29Speaker 01

  Cool. Yeah. And then like for the audience, keep in mind, like we're doing this really, really quickly. We're speed running through this, you know? Obviously, like if you look at this design, it's a lot more refined than what it will be. Typically, I would spend a couple of hours going through this, going through specifically like different illustrations or blocks or sections. I just want to cover like the key principles of how I approach it. And then you can come back and like use these different components to refine each individual section. Any questions, thoughts?

• 20:57Greg Isenberg?

  Yeah, I think, I was just, we're just thinking about this, like we all have these, this, you know, expectations that you're going to like one prompt, something two prompt, something three prompt. So I'm happy you said that it's probably going to take a couple of hours if you really do want to get it dialed. So I just, I don't want people to go through this whole process and be like, well, when Amir did it, it looks so much, so much better. Well, Amir, Amir, first of all, you had the library, so that was helping. But also, you know, it takes time, right?

• 21:27Speaker 01

  It takes a lot of time, like we, yeah, we.

• 21:30Greg Isenberg?

  I mean, it does take a lot of time if you think about it. It just takes like.

• 21:34Speaker 01

  Yeah.

• 21:35Greg Isenberg?

  About, you know, if Amir from 2017 saw this.

• 21:39Speaker 01

  Oh, like he knew.

• 21:40Greg Isenberg?

  And heard that, and heard that you said it took a lot of time. You'd be like, what is Amir's on?

• 21:47Speaker 01

  No, yeah, no.

• 21:48Greg Isenberg?

  But it takes time, it takes some amount of time.

• 21:50Speaker 01

  It takes time, taste and skill to know what you got to do and how to do it, right? It's like we've, we made it so easy to achieve the tasks. How do you give the agents the right direction and approach to say, here's how you need to do it? The taste, taste is what we know. We've been talking about a lot.

• 22:06Greg Isenberg?

  Yeah. And, and direction, right? Like this particular component needs subtle animation. Here's a reference of that.

• 22:14Speaker 01

  Exactly, exactly. Like, I built this component from scratch, not from scratch, off an existing component that was, I think, let's actually find it.

• 22:24Greg Isenberg?

  So you found a component here.

• 22:26Speaker 01

  Exactly. So I found a component here. We're gonna find it, because I know, I'll tell you where it is. So like, this is, for example, right here. Okay? So we had this component, and this is it right here. I, I ported this one over, but I said add a solid animation to switch through it, for example. Or this analytics query right here. So this one is, this one, right? Now, that specific component, I think it was, it was one of these, one of these cards, essentially. And yeah, I took it and I said, hey, based on this component, oh, right here, this one. It was this one, and I turned it into this, which is, you know?

• 23:08Greg Isenberg?

  I think it looks great.

• 23:09Speaker 01

  Yeah, yeah. So, cool. So we, we've, we've built out the design of the guide page in here. And now we're gonna see if it's live on the website. So let's push this, and then we're gonna go to guide. Boom. So now we have it, we have the sections here. Okay, so now, let's get some analytics in here, to see how many people actually could sign up for this lead magnet.

• 23:33Greg Isenberg?

  So, I just wanna say one thing, this lead magnet looks gorgeous. Like, think about, this lead magnet looks better than 99.99999% of lead magnets that exist, right? So, people listening to this, like, this is, uh, think, think about, you know, don't just create one lead magnet, right? Like, every few weeks, create new lead magnets, optimize your lead magnets, and you're gonna be able to outperform, because you're gonna know how to do and create gorgeous looking lead magnets.

• 24:03Speaker 01

  Exactly. And we did, we did this in like 30 minutes. Again, like, you get the right components, the right direction, you can do some research on the best ones, port them over as reference styles, and then make that refinement. So right now, while you're going through it, I found this kind of decent component. I was like, why don't we just try this design instead and add it? So let's see what happens. Okay, cool. It's gonna adopt it and it's gonna have that split layout. Alright, so we've got this lead magnet. We'll wait while it's refining the design. We're gonna come back and essentially get some analytics installed and look at some click tracking, form submissions, and run some experiments to see if I were to change this headline, for example, what would this look like? And what would increase more form submissions for our leads?

• 24:45Speaker 01

  And we use this approach at Humbletics, we just started actually, where we built like an autonomous CRO agent, where we have direct skills and MCPs connected to a Claude. So we directly connect to our analytics and experimentation tool, our marketing site, which is why we port it over to custom code. We wouldn't have been able to do this through Webflow, and then we have Google Ads. The idea is how can we spin up multiple sub-pages that are personalized campaigns, like personalized campaign landing pages, based on the campaigns we have? So like, if I'm running a pain point campaign on Google Ads and have five ad sets, I want each one to have its own personalized landing page, and I wanna A/B test each individual

• 25:27Speaker 01

  landing page to see what headline resonates the most, or what layout, for example. You're able to do this with the custom code site really quickly with Claude code. With Framer or Webflow, you have to create the pages, create the sections, come back and update it, you know what I mean? So that's kind of the whole thesis we were talking about earlier, where it's like the feature of websites is custom code and the agent is the CMS. And then you have it running, and you can even set up like, so what we do is we have like a cron job now directly in Claude code, because they support automated tasks, to say every week, go into, like, run this specific task. So what we have is Google Ads directly connected to Claude through the API. We have three Claude scales, a paid media manager that pulls data from Meta and Google. We have a CRO optimizer that runs A/B tests, and then we have a final report that just

• 26:12Speaker 01

  gives us a report of like, here's how many users you had signed up. It connects to Stripe, Chartmogul, all the different tools that we have. And then we have Humbletics that essentially hits our endpoint to give you recommendations, actually uses Firecall to scrape the pages, and then helps you kind of say, okay, here's what you need to optimize for, helps you run the experiment all through Claude, and then it manages the, manages the, like, the results for you.

• 26:38Greg Isenberg?

  That is just so cool that we live in a world where you can do that.

• 26:41Speaker 01

  I know. Yeah. And the best part is you can actually use this with Idea Browser to pull all that context and save it as a file back into Idea Browser. Jordan and I were talking about this. We're like, yeah, we can do this now. Like, imagine you do a weekly report. You pull all the data, you add it into Idea Browser, and now Idea Browser has context of like, how growth has been over time.

• 27:01Greg Isenberg?

  Which is, which is sort of the unlock, right?

• 27:04Speaker 01

  Yeah.

• 27:05Greg Isenberg?

  Yeah.

• 27:05Speaker 01

  Yeah.

• 27:06Greg Isenberg?

  You know, if you have that, you're just gonna get better results. You're gonna compound, and that's like how you could compete, right?

• 27:15Speaker 01

  Yeah, exactly. So we just added the component to the, we installed this new component, but I realized the mistake we made was we got rid of the form. So we're adding it back in. But yeah, like the, to your point, we actually, before this, we were tracking all the context directly in Obsidian as well. So we have a master, like, performance log.

• 27:36Greg Isenberg?

  Yeah.

• 27:36Speaker 01

  It says, like, here's how this A B test did. Here's how much money you made. Here's what you need to improve on. And we just come back to this context. So we have this all set up now. We have this landing page. And let's go back to the homepage and maybe run an A B experiment on this homepage to see if we can increase click-through rates. So what I'll do is I'll spin up Claude again, and then we're going to go into, so with Humbletics, you can just go directly to the API. Create an API key, get your property details and instructions to hit the Claude. But what it, you know, you get the basic analytics that you typically see, where people are visiting from, traffic, you know, the channels, everything you want. If you're looking to like run paid ad campaigns, you get full attribution directly in here. So you can see exactly what channels are making you money.

• 28:21Speaker 01

  You can have heat maps to see scroll depth, what people are clicking on, the full experience. You can even create funnels. So for example, if you want to go like, you know, from page to guide, you can even run it. It'll show you what people went through the steps. And then you can run A B experiments. So we have zero right now. We're gonna use Claude to actually run an experiment. So we're just gonna copy this.

• 28:47Speaker 01

  And then we're gonna have a, we have a dedicated skill for this, but I thought just to show people the steps. So using this approach.

• 28:54Speaker 01

  Create an A B experiment, no code type for the headline.

• 29:02Speaker 01

  So just what the headline.

• 29:04Greg Isenberg?

  So you have a skill for this?

• 29:06Speaker 01

  We do.

• 29:06Greg Isenberg?

  Yeah.

• 29:07Speaker 01

  Yeah, I'll share the link afterwards.

• 29:08Greg Isenberg?

  Okay.

• 29:09Speaker 01

  That people can use where essentially we'll help you get set up with like combo analytics, run experiments, get recommendations. But we have endpoints right now that you can hit to say, okay, like tell your agent or Claude code, hit the specific endpoint, the recommendations endpoint, and it'll actually scrape your site, pull traffic insights, pull all the existing data you have in Hubble analytics and even idea browser, and then give you a very detailed report on here's what you should like focus on for your landing page or what you should test. Yeah. Nice. So it's pulling insights based on analytics. We have a 40% average scroll, 25% bounce rate. Here's the page content. Here's what your control should be. And then here's what you should be, what should be the variant. Okay, cool. We should now go back into experiments.

• 29:51Speaker 01

  We now have an A B test running directly through here, which is crazy. It's actually insane. And we did this without even deploying a change because our script dynamically updates the content on the site without you having to go to your developers and say, hey, like, can you, can you create this new page or push it for us? So this is what we had. If I refresh this, maybe a couple of times, because it's like distributing traffic. Let's see.

• 30:15Greg Isenberg?

  Right.

• 30:16Speaker 01

  There we go.

• 30:18Greg Isenberg?

  Every lost deal started with an objection your rep wasn't ready for. Honestly, not bad.

• 30:23Speaker 01

  Pretty fire.

• 30:23Greg Isenberg?

  Yeah.

• 30:24Speaker 01

  But here's the cool thing, we just launched an experiment.

• 30:27Greg Isenberg?

  Yeah.

• 30:28Speaker 01

  And it automatically updated the page without pushing code. And now you're tracking to see how this headline will actually impact conversions.

• 30:35Greg Isenberg?

  Right.

• 30:35Speaker 01

  So if I click this, for example, you know, and then let's say we go back to control. So this is right now showing, okay, now we're on control and we go back in here. We can see exactly how conversion is doing. So it's still pretty early, so it's going to like not have any good data to show, but it's showing you even what people are clicking on directly and telling you how the test is doing. Insane.

• 30:59Greg Isenberg?

  Yeah.

• 30:59Speaker 01

  Insane.

• 31:00Greg Isenberg?

  It's also insane that like, I don't know anyone who's doing this.

• 31:04Speaker 01

  Yeah, I know. Yeah, we, we, I think this, this stack right here.

• 31:11Speaker 01

  What I showed between going from idea to refining in paper, then analyzing, optimizing for this. Like, if you're a go-to-market person, a marketing person, and you use a stack, you're unstoppable. Yeah, so we're, we're doing this at scale now. We're running a bunch of paid meta ads, Google Ads, directly connected, pulling all the context, creating personalized pages, and seeing like, what's, what's making money?

• 31:31Greg Isenberg?

  I mean, there's also an opportunity to actually take this stack and sell this as a service, right? Like, how many businesses, you know, would, would pay five, 10, 15, 20 grand a month? If you can actually be doing this for them, like, probably a lot.

• 31:48Speaker 01

  Yeah, we, so we actually have a couple of customers that work with us and say, hey, like, we'll pay like 5K, 10K a month. Can you just run?

• 31:54Greg Isenberg?

  Right.

• 31:55Speaker 01

  Can you just use your software and then run it for us?

• 31:57Greg Isenberg?

  Yeah.

• 31:57Speaker 01

  Yeah, yeah, because like, you can connect meta ads directly in here, Google Ads, get full revenue attribution. And then they say, hey, like, just tell us like, where, where our top of funnel spend should be. And so it's almost like, we're seeing this now too, like managed service.

• 32:10Greg Isenberg?

  Yeah, makes sense.

• 32:11Speaker 01

  Yeah, yeah. So, yeah, we've, I feel like we've covered a lot, right? We went from an idea that we had in Idea Browser to, to like now having relevant context. So we can even come back and say like, in the terminal, say, you know, using Idea Browser MCP, add this info to the context. So we can come back and actually add this as a file, right? So we went from idea to designing something in paper, to building a new landing page, to then having analytics and experiments behind it.

• 32:42Greg Isenberg?

  Crazy. What a time to be alive, my friend.

• 32:45Speaker 01

  What a time to be alive.

• 32:46Greg Isenberg?

  All right.

• 32:47Speaker 01

  Was that saucy?

• 32:47Greg Isenberg?

  That was saucy, man. I like, I would be real with you, like, this, I've, I've never seen anyone do this, this fast with this stack.

• 32:59Speaker 01

  Yeah, yeah, yeah. If, it, sometimes I yearn to be an employee again and like go to market, on marketing, so I can just take this and just blow people's mind away. Because it's like, it's so much fun. Like, I wish we had a million dollar monthly spend, so I can just dabble with these tools.

• 33:15Greg Isenberg?

  Yeah.

• 33:15Speaker 01

  And just show them what's possible.

• 33:16Greg Isenberg?

  Yeah, what's going through my mind is like, there's a lot of arbitrage.

• 33:20Speaker 01

  Insane.

• 33:20Greg Isenberg?

  Right? Like, if you... If you can, you know, if you have good ideas, if you get, you know, a source of good ideas and you can A B test them and you can be creating beautiful websites, you know, lead magnets, that sort of thing, and you have access to 7 billion people who have, you know, internet connections and their credit cards attached to it, like,

• 33:44Greg Isenberg?

  there's arbitrage. And I remember, I remember when the Facebook ad platform first came out.

• 33:51Speaker 01

  Oh, yeah.

• 33:51Greg Isenberg?

  And, you know, the average click, I think, back then was five cents a click. So, meaning like you can create an ad and I was at the time working with these social gaming companies. So I would say like, hey, you know, as an example, Farm Bill, I would say like, would you pay me $2 for every install I got for you? And they'd be like, sure, this, you know, we monetize so much better than that. And I was just using Facebook ads because I was paying five cents a click. And there was this moment of opportunity, basically, where you can do really well. Because of this ad platform and how cheap it is. Now, fast forward to today, I don't even know what a CPC is on average. It's probably $2 or something like that.

• 34:33Speaker 01

  Yeah, it's $2, $3.

• 34:34Greg Isenberg?

  $2, $3, right? So it's, it's, you know, so much more expensive and it's, and it's just because people didn't know back then, right? And it's sort of the same thing with this, with this sort of stuff. It's like, you know, 99.999% of people don't know that this stack exists.

• 34:48Speaker 01

  Yeah.

• 34:49Greg Isenberg?

  And there's an opportunity to create offers and arbitrage right now.

• 34:54Speaker 01

  The right tooling, the right approach, you have a terminal with a million context tokens. The opportunities are endless.

• 34:59Greg Isenberg?

  Yeah.

• 35:00Speaker 01

  You know?

• 35:00Greg Isenberg?

  So I appreciate you sharing.

• 35:03Speaker 01

  Yeah.

• 35:04Greg Isenberg?

  And I'll include the link that, for that skill in the show notes in the description.

• 35:10Speaker 01

  For sure.

• 35:11Greg Isenberg?

  I'll include the link somewhere to follow Amir on X and other places. And stay saucy, my friend.

• 35:18Speaker 01

  Thank you for having me, man. Good to see you.

• 35:20Greg Isenberg?

  Good to see you. All right.