§ 01 · The Hook
The bait, then the rug-pull.
A viral tweet with 315,000 views — and a Polymarket bet on whether it was real — brought Cody Schneider onto Alex Lieberman's show to prove it live. What followed was sixty-two minutes of screen-share where a single person built bulk ad generators, programmatic landing pages, and cold outbound pipelines from scratch, on camera, in real time.
Who's talking.
00:00hostAlex Lieberman
00:08guestCody Schneider
Where the time goes.
00:00 – 05:39
01 · The viral tweet + fraud question
Alex reads Cody's 315K-view tweet claiming a day's Claude Code output matches a Fortune 500's year. Cody defends it. Polymarket bet mentioned.
05:39 – 11:52
02 · GTM Engineering defined
History from Clay-era outbound-only (2023) to full-stack distribution. Truffle pigs of inefficiency. Same job as growth hacker — better tooling.
11:52 – 22:08
03 · Bulk Facebook Ad Generator — live demo
Cody live-runs his React + HTML-to-Canvas bulk ad generator. Perplexity for ICP pain points, Claude Code for 40 ad variations, zip download to Meta.
22:08 – 32:06
04 · Ad testing + programmatic landing pages
CPC campaign testing methodology. Strapi CMS + Claude Code API calls to bulk-generate landing pages, auto-submit to Google Search Console.
32:06 – 40:50
05 · Scale, UGC ads, multithreaded work
HeyGen bulk UGC to Veo 3 iteration to human only at the end. Alex reframes multitasking as multithreaded work. 30 agents behind you.
40:50 – 51:22
06 · LinkedIn pipeline built live
Phantom Buster to Apollo enrichment to Million Verifier to Instantly AI cold email. Built from scratch in ~20 min on camera. Deployed to Railway.
51:22 – 58:22
07 · The real bottleneck: data at scale
Generating at this velocity creates data problems no MCP can solve. Origin of Cody's company Graft — AI data analyst for GTM teams.
58:22 – 1:02:09
08 · Q&A + top 5 tools
Tool discovery (Twitter first, LinkedIn 6 months behind). Team deployment via shared GitHub repos. Top stack: Instantly + Claude Code + Graft.
Lines you could clip.
03:18
“To do this type of work, I would have had traditionally to go out and hire a team of like 20 people. In contrast, it's just me now.”
04:04
“I'm actually losing sleep. To be totally honest, I woke up at 4AM this morning.”
35:55
“You're no longer a single person joining a company. You're a person with 30 agents behind you, and you have all this personal software that you've written that you're bringing to the table.”
59:01
“Everything I talked about today, you can literally be like, I don't know how to do that. Tell me how to do this, and Claude will walk you through every part of that process.”
1:01:30
“Get obsessed with the outcomes. The tools are — it doesn't matter anymore. There's no limitation on the tools. The hardest part is knowing what should I be doing.”
Things they pointed at.
18:40toolHTML-to-Canvas
24:10toolStrapi
33:50toolVeo 3
44:00toolMillion Verifier
26:40toolKeywords Everywhere
1:00:10toolHyperTite
Where the conversation goes.
00:00 – 11:52steadyGTM Engineering history and definition
11:52 – 32:08denseBulk ad creation workflow
32:08 – 40:50denseCreative testing philosophy
40:50 – 51:22denseLinkedIn outbound pipeline
51:22 – 58:22denseData analysis at AI-generated scale
58:22 – 1:02:09steadyQ&A and tool recommendations
Word for word.
metaphor
00:00If you are a marketer in sales or in go to market and you truly want to call yourself AI native, you need to watch this episode. On this episode of Human in the Loop, the most actionable AI show on the Internet, I talked to Cody Schneider who truly is a 100 x marketer, and he shows you his systems, his process, and everything that goes into him being probably the most effective and AI native marketer I've ever met and what the future of growth marketing looks like because of AI.
00:31Let's hop into it. Okay. So we I'm gonna start with a tweet that you put out that blew up.
00:41It's what got us connected to do this. So you said, I don't think you understand what is happening in GTM engineering right now.
00:50I want to try to explain this to you. What someone can do in a day with Claude code, plus APIs, plus plus n eight n, plus railway.com, plus GitHub repo, plus Skill MD files is what a Fortune 500 would do in a year.
01:02Today, I made 40 Facebook ads, a 100 landing pages, wrote three guest blog posts for backlinks, booked myself on four podcasts with a cold email automation, wrote five help desk articles, added two vlog vlog videos, scheduled 25 tweets across accounts, wrote two pieces of scripting software to give away as LinkedIn lead maggots maggots.
01:23Magnets, and baked bread from scratch and made katsu sandos with my chief of fiance. As I write this, I still have four hours left in my day. I don't think you understand what is happening in GTM engineering right now.
01:36And as of recording this, it has 315,000 views. Now, I will say, once we schedule this and people I think had a lot of feelings around your tweet, there I I think you told me this.
01:48There was actually a a Polymarket or call sheet bet around if Yeah. I'm wondering if this was like possible to do or not basically. Which I thought Basically curious.
01:57So So for the for the for the people who read this and say, Cody has to be a fraud,
02:03what is your response to them? I'm just doing I don't know, man. Like, with everything that I'm doing, it's like, this is what I'm with and seeing work right now, and I just share it in public.
02:12It's I think for a lot of people that have never interacted with my content, they're like, oh, it's a bunch of just cap, like, full stop or like, yeah. Again, this is just what's working for me. Like, I just kind of live tweet, like, here are the things that are being successful based on my go to market strategies, like, with my company and with what I'm seeing with friends companies who I talk to.
02:32So anyway, yeah, hopefully today, what like give you all the tooling you need to be able to do this yourself. It's totally possible.
02:39Everybody that is on this call will be able can do this. Like, it's not you don't have to have any specialties or like any like, you know, specific knowledge to get started on what I'm talking about today. And before we hop into it, like,
02:52how, like, how profound is what's happening in go to market and go to market engineering right now?
02:58Like, what is the impact this can have on someone's career or their business if they get this system right? I mean, this is one of the most searched for roles right now. Like, I've gotten five
03:07messages in the last, like, seven days from founders that are like, yo, I'm trying I like I saw this tweet. I'm trying to hire this person. Like, who can you intro me to?
03:15Right? Basically, like looking for these skill sets. Just to give you context, like I've worked in startups for a long time.
03:22Early stage is what I specialize in. Like, to do this type of work, like, I would have had traditionally to go out and hire a team of like, you know, whatever 20 people to execute on this.
03:31In contrast, it's just like it's just me now. Right? Like, this is the difference that this has evolved into.
03:36And, like, this is how companies are starting to, like, leverage all of these tools is to be able to basically compound their skill sets. I'm like, your domain knowledge that you have combined with this is where the real magic is. Like, if you have a deep expertise in one thing and you can you understand the systems and the processes for that, and you can translate that into code, which is, like, literally what we're doing, and then use this as a system, like, to go and do activities for you.
03:59Like, that's where the real, like, power of this comes from. So I I I I don't really have words for it.
04:05Like, I'm actually losing sleep. To be totally honest, like, I woke up at 4AM this morning. I was like, shit.
04:11I think I can do this, like, this totally, like, different thing. I my friends are, like, also like this. Like, you know, the the people I talk to on a weekly basis that are in growth that we just share, like, what is working?
04:22Like, what's happening right now? What's the meta that exists, like, currently in the market? They're all just freaking out basically because the amount it's like, I I was selling you off this call, like, before we started, Alex.
04:34Like, I I have a friend. He works at a start up. Like, they're being too slow on shipping what he's doing at the company.
04:40So he's basically in this waiting period. He's picked up, like, three side gigs doing growth consulting for them, and he's shipping more using this type of system than what he's doing at his day job. Like, that's how, like, the velocity that you can go at.
04:54And I know there's, a ton of people that are like, oh, it's all slop. It's all it's all you can do this really effectively and we're gonna like talk to that I know on the like beginning of the conversation. But
05:03Two two things before we hop into it. One is, are people going to know how to make killer katsu sandwiches by the end of this?
05:11I can give you the recipe. I mean, it's super simple. It's basically just like get good chicken from a butcher,
05:17Yukus Paco, and then just make like a tonkatsu sauce to put on the top of it. So Mike Love it.
05:23My fiancee's
05:25family is Japanese American so we get we kinda get hooked up in that that neighborhood. Got Cody is a renaissance man. Or as someone in the chat put it, the the Ehrlich Bachman of growth.
05:35I love you. I'm sorry. So good.
05:39I'm gonna stop talking. Let's fire this thing up. Where do you wanna start?
05:42Yeah. I think I think just to begin with, like, just a quick history of what is GTM engineering for the uninitiated. So this really kind of started to happen in, 2023.
05:53Originally, it was just like for outbound motions. Right?
05:56So Clay really kind of solidified this as a job function. And it was really focused on like, rather than thinking about can't the work that I'm doing to do distribution as campaigns, let's think about it as like systems and processes that are directly, you know, trackable, attributable, and treating this almost as like what a software engineering function would be where it's like, hey.
06:16We do week long sprints. We look at the data. We analyze what's working, and then we re repeat those processes.
06:21In the last couple of months, it's totally had an evolution. So now it's not just outbound.
06:27It's everything. Like, you know, it's it's inbound. It's paid ads.
06:30It's organic SEO. It's like organic content. It's you you you name it and there's not a thing that it's not touching.
06:37And this has really become kind of the foundation for, like, how these, like, early stage companies are doing really effective distribution and growth strategies is trying to build out these systems and processes.
06:51And then, like I mean, I'm seeing people incorporate these into their code bases as well, right, where it's like, okay. I built this side project thing. It's now running the entire, like, you know, blog, like, organic content strategy.
07:03And now we're gonna build this into their core code base of the company, the entity, you know,
07:08and then it's it lives on in perpetuity. It's like this agent that's just kinda working continuously in the background that's in the core code base. So that's And and so you do you think do you think the phrase GTM engineering is a useful or valuable phrase anymore given that now to your point?
07:22Like, it started as treating outbound sales like software engineering, treating it more like a science than an art.
07:30Now you're basically saying it touches sales, it touches marketing, and now even, like, if you look at other companies, they're hiring.
07:37Like, yeah. Now it touches like basically, I've heard GTM engineers refer to as people who just like are in a business, and they build
07:45agents and things for any function of the business. So do you think that phrase even has value anymore? Yeah.
07:51I think it's just the the same names, the exact same person that's always existed for the last 10 or twenty years. Like, it was a growth hacker. Was a head of growth, and now it's a GTM engineer.
08:01It's just the tooling and, like, the activities they're doing here are the exact same. Like, if I look at my day to day, like, if I was to zoom out, like, I'm doing the exact same type of work. Right?
08:11It's like I'm basically trying to find marketing and distribution arbitrage within the system, like figure out the positioning for my company so that the audience is receptive to it. And then I'm going to go and basically, like, build out processes to go execute on that and then, you know, have some type of data analysis over the top of it to actually track what I'm doing.
08:30It's not exact same thing. I'm just saying changing the underlying tooling. So I don't know if it like, GTM engineering really, like, encapsulates everything that you can do.
08:41I just think that that is like how people are talking about it and describing it. And it's also just like one of those buzzwords that are for for currently. But if you're in any role that, like, gets people to buy things, you can do these exact same strategies.
08:54Like, whether that's sales, whether that's marketing, whether that's customer experience, whether what whatever it is, like, whether that's rev ops, it doesn't matter. You can all, like, use basically these this foundation as a way to, like, do your work, you know, more quickly. Well, let's now that people have a foundation of what GTM engineering is, let's let's give people the sauce.
09:14Where do you start? Yeah. So to begin with, like, how I structure it, Cool.
09:29So to begin with, like, what I've done is I just have this file that is on my local machine that's in documents that's literally just like, this is what I live out of now. This isn't organized, and I know there's an engineer that's gonna see this and be like, this is disgusting.
09:45I totally agree. It doesn't matter. It works for me.
09:47So, basically, what I've done here is I've like, you can start with an empty file, and this is how you begin this whole process. It's basically make a file that you're going to live out of.
09:58And the first thing that you're going to do is create an environment file. It is called an EMV file.
10:05This is where all of your API keys are stored, and this is how you're going to go and interact with all of the tools that you use within your stack. And this is what I'm about to show today. So what I've done now is I've CD'd or I basically moved into this directory.
10:18So documents forward slash graphs growth agents, and then I'm gonna start Claude. I'm just gonna assume that you know how to, like, install Claude code onto your your your applicate or, you know, onto your local machine.
10:30If you don't know how, there's tons of tutorials how to do this so we're gonna skip over that basically. Cool. And can you just zoom in a bit just Absolutely.
10:37No problem. Yeah. So Thanks.
10:39So now that I'm in this folder and I've got Claude open, this is basically where my work starts from and this is how, like, I go about my process of, you know, whatever it is that I'm doing. So a lot of the times how I'm doing this is I'm spinning up multiple Claude agents and I'm putting them in different desktops.
10:57And then I'm just moving between those desktops as I'm do as I'm doing work other work in the background. So, like, for example, for the Facebook ads, the first you know, one of the campaigns that I'm running right now is basically a before and after.
11:11So I'm just gonna bring up this, like, software that we wrote for this.
11:17So give me a second. I just gotta remember the name of what it is, but I just actually gonna transcribe. So it's the Facebook ad software for the static ads before and after.
11:25Can you start this and run it locally? So what I just did there is I used a transcription software called Super Whisper to transcribe that out. I'm now gonna tell Clogcode to basically open up this software.
11:38What it understands is the file structure, everything that I've done previously so it can basically go and bring this up for us. And so this will be like the first thing that we do today is basically we're gonna I'm gonna show you how I built out this template that is for ad creation and then show you how I'm basically using this to go and then do bulk ad generation.
11:59I'm just gonna walk you through this entire process and then show you how I would, like, set up an AI data analyst to basically watch this campaign that I'm doing for testing and what I would use for measurements of success. So this whole thing here that you see is entirely, like, AI generated.
12:15This is actually components. So if I was, to zoom in on this in Hubbard, this is, like, React components. So this is code.
12:21So this came from an epiphany I had that, like, all design is actually just code on the under like, if you go into Figma and you're doing any design, it's just code that's under the hood. So why can't I just make these outputs using React components?
12:35And then I use this library called HTML to Canvas. And this library is basically a way to export that React component as a downloadable PNG.
12:47So built this template out, basically go back and forth with Clog to, like, create that template. Once I've done that, I can then use this to do bulk iterations. So then I created the sec or, you know, the separate function that enables me to go and bulk generate as many different variations.
13:06So it's just text variation changes. But why would I be doing this? I'm basically trying to test different angles that I can talk about the same product from to understand the positioning.
13:16So how would I go about this process to actually, like, you know, create that output? So just for, again, today, we're gonna be using my company, Craft, as, like, the the the company that we'd go about this for.
13:27So what I would go and look for is I'd use Perplexity to help me identify I I've now built this into Claude. Like, I have an a file talking about my MCP, but just to zero to one this process, this is what I would do is I would basically go and I would search something like, what are the pain points that's an AI data analyst for GTM teams?
13:55Do something like what are the pain points that an idea is would solve, and then what are the outcomes?
14:06Actually, I'll just do this transcription. So what are the outcomes that an AI data analyst for GTN teams like, what would that GTN team want?
14:15What would be the outcome that they would want? And I'm gonna use this as source material for the ads that I'm going to create. So I'm then gonna put this like tag to like search Reddit for this to pull in actual customer conversations.
14:31So what I found is that when I do this, I'm using the language of my target customer. Like here's their actual pain points and the actual, you know, outcomes that they're looking for that I'm then going to go incorporate into the bulk production of this ad creative content.
14:47So I'm talking in the language that's going to be most receptive to them. At that at this point, I would pull this out. Again, you could use a you know, you could use Claude.
14:56You could use ChatGPT for this. This is just what my workflow looks like. I would then go to Claude, and I would say, you know, actually, honestly, I'd probably just drop this into this chat now.
15:07So I'd drop this in as context, and then I would say brainstorm we'll say 40 ad titles and supporting paragraphs based on the source material.
15:30And then it's basically going to take that into context. It understands that it's interacting with this code that I have written or that we've created. And then it knows, okay, this is what we're trying to basically like change within this.
15:42It's gonna create those title and subject or title and supporting paragraph variations, and then I can just say go and do the bulk creation of these ad creatives. At that point, it's gonna make the zip file for me of all those variations and then I can upload that into Facebook ads.
15:59This is something I'm like actively building out right now is basically that connector. Like, I don't have this built yet. It's halfway done.
16:04I was trying to have it ready for today, but it's not. So I'm you're just basically going to then like, within Cloud Code, I could basically say, hey.
16:13I wanna upload this zip file or upload all of these images into this specific campaign so you can drop in this campaign URL. That uploads them in.
16:24And then, basically, what it allows for is you to, like it puts them into a draft. I can change all of the URLs, like, that scale that they're sending to change the subject lines.
16:36So it's written these. Right? And at this point, can say, okay.
16:40Now use okay. Now use the bulk generation for the ads to create ad variations of each of these and hit send.
16:50So while this is working, at this point, I would spin up another workspace. So I'd go then to documents again and go to graft agents, and I'm gonna start Claude again.
17:04And I would move this over to Can we just course. For
17:08one sec, can we just pause for a sec? Because you're like, you've covered so much good stuff, and I just wanna make sure that I'm on the same page with everything.
17:22Okay. So basically, what you've done like, the whole goal here is you're running a ton of meta ads for your business, and your bullet your whole thing is, like, the bottleneck historically was getting enough good creative to test and rotate in when you're running paid marketing.
17:39But now, basically, you have the ability to, one, do research into your ICP for your business and see, like, what are actual problems or pain points that kind of, like, the the end buyer for the AI data analyst for GTM.
18:00What are the pain points they've actually talked about online, which is why you searched Reddit? You then use that output from Claude or for sorry.
18:08From Perplexity, but you could do with any LLM. You feed it into Claude where you then have Claude create kind of, like, the the h one and any of the copy for the ad, and then you use all that plus this this piece of software you've already built to basically build 40 variations of the existing ad you have.
18:28It's not connected yet, but ultimately, you're gonna have org connects where those 40 ads just get uploaded into your ads manager and meta. You can go through and decide which ones you wanna set versus not.
18:40Yeah. Aria, I can just have AI tell me which ones are the best like performing. Right?
18:44So Yep. This is And like and a few people have had a question here which is like well, there's been a bunch of questions, but the process so you already have this thing built. Yeah.
18:55That is building the ads for you. I think people maybe have question like how hard is that? Like could anyone Do that in about twenty minutes to just give context.
19:04I basically was like, build me like an bulk ad generator.
19:08Like, so this this I built this first to visualize it and it like has base so I gave it an example of an ad that I had seen previously. I was like, hey. I want you to make a template based off of this.
19:20And then I just went back and forth with Claude. So install the Claude code skill that's called UI design. If you just Google it, that will come up.
19:29And this will just, like, help with the outputs, like, from a quality standpoint, just from, a design standpoint. But this this whole thing that you're seeing right here, like, this whole interactive thing, this was, like, two prompts basically to get this live.
19:43And then the bulk generator took a little bit more, but it was just going back and forth, basically explaining the outcome that I was anticipating.
19:52And this whole build was about twenty minutes to build this infrastructure.
19:57Awesome. Cool.
19:59Okay. So I've now created all those variations. And I I just wanna zoom out to you like why are we even doing this?
20:05Like, why am I making all these variations? I I don't know what's going to work on Facebook ads. Right?
20:10I have, like, gut intuition on what I think is going to, but how like, what the audience or my audience is going to be most receptive to, I I the only way that I can identify that is through testing.
20:23And so by doing all these testing variations, that is, like, why that's how I'm going to see, like, what the audience is most receptive to from a positioning standpoint. Once I find the those winners.
20:33Right? And I just have, like, examples in here to show you this. Right?
20:36So, like, this is the CPC's 30¢ on this one in comparison to 53 and 85 on these other other examples. I can then say, okay. Well, what does that what is that messaging on that specific ad?
20:47How do I go and remix that messaging across different formats? So I would then take this this messaging from this ad, and I would go and I would bulk generate, for example, maybe UGCs to to do a different ad format. So this is again something that I'm building out actively right now.
21:02It's basically like plugging into the HeyGen API so that I can then go and bulk create creative for the like, based off of those winning formats.
21:16And, like, I've already I'm already doing variations of this, like thousands of them. Right? Like, if I could just continue to scroll.
21:21So all I'm doing is taking the workflows that are already working for me, and I'm just trying to automate as much of those as I can, right, by using this tooling so that it's in the background basically, you know, functioning on this. So alright.
21:34So it's completed that. Right? At this point, I would go back.
21:37I would go to that bulk section. I could download those those files. I don't know if it updated this.
21:43Maybe it's only showing the first five. Or yeah. It's only gonna do these five.
21:47But, anyways, this is we could go back and forth basically to be like, okay. This oh, so this zip folder is in bulk ads comparison at zips.
21:57Add these to the bulk page. Just say forward slash bulk dot html so I can download.
22:07And then while that's working on this, this is where this starts to compound is I would go and I would start on another process. So I found a winning ad format that's working for me.
22:17Cool. Now what I wanna go do is I wanna go spin up landing pages that are dedicated to those winning ad formats. So what I how we've structured this is I use this software called Strapi.
22:29So Strapi is an open source CMS. You can run this. This is like actually how we've built this on graph.com.
22:36So we use Strapi. It has an API endpoint that I can hit, and this allows for me to go and create, upload and create blog content, landing page content, and whatever other content formats that I want. So you basically create a content format within Strapi, and then I'm going and interacting with that through Cloud Code.
22:56You could do this with any CMS. Like, you could do this with WordPress. You could do we just use Strapi because it's, like, infinitely scalable and we can self host it in the long term.
23:05So if I have a 100,000 blog posts that are on the site, like, this is a way that, like, I can do this at that scale at a super cheap cost. So for these landing pages, I already have a template that that exists. And all I wanna do is change the the the title and the the it's basically the homepage of the website that is just like built into a templated format.
23:27So it has this, like, h one and it has this supporting. So say I find a, like, Facebook ads angle that is working.
23:36I could then go and bulk generate those landing pages. So this is one way that I would do this with Facebook ads.
23:43The other way that I would do this is I would actually extract the keywords that are converting from Google Ads. So what's actually creating the sign up action or the payment action, and then I would go and build these landing pages out at scale. I used to do this manually.
23:56I now, like, just use clog codes to do this. Right? So what I'm gonna do is, like, open the landing page.
24:04First, let's just do this. Open the landing page directory. We're gonna build some landing pages together.
24:11And we'll get that started. As you're as you're doing this, one question I saw from folks is how much do you have to spend?
24:20Like, much are you spending on meta to feel confident in these tests you're running before you kind of double down on winning creative?
24:26Yeah. So, like, we haven't spent a ton to be fully transparent. Like, we're super early stage.
24:31I, like, literally figured out the positioning for this brands like two weeks ago. Like, we we started out as really broad.
24:39Like, we're this AI, you know, business intelligence software. And then we funneled down into, no. We're AI data like an AI data analyst for GTM teams because we found this to be a massive pain point when they're in the market.
24:50Or it's like, hey. You're like, you have 10 to 50 people and you don't wanna hire a data engineer, but you need a way to unify your data. We're the solution for that basically.
24:58So it it really like, how I approach the structure of my, like, Facebook ads campaigns, well, this is just how I do it. There's a million ways to do it.
25:08This is how I structure it. So I'll test all the ad creative against each other in a click campaign, and then I'll look at what has the cheapest CPC.
25:16And I can run, you know, we'll do like a $100 over a three day period to see which gets the cheapest CPC. At that point, I'll take those winners that have the cheapest cost per click, which is typically an a leading indicator that once I move them into a conversion campaign, I'm gonna get a cheaper cost per action.
25:32Like, this is just the what I've seen historically. So I take I test them basically in a CPC campaign against each other. I take those winners.
25:40I spin those out into their own conversion campaign, and then I give them their own dedicated budget that they're running against. And this is how like a way that you can basically test creative.
25:51Like, you know, imagine testing 200 different pieces of creative on a monthly cadence to find those winners so that you're always refreshing the creative that you're, like, adding to your your your ads platform that you're using. Same style, some functions with Google Ads as well. It's basically like I'm just constantly looking at the keywords that are creating conversions or like getting the cheapest CPC in relation to Yeah.
26:13It's it's kind of like this is the paid marketing equivalent of,
26:17you know, I know Gary Vee talks about this a lot or even if you look you think about like some of the best non fiction authors say, like, even think about, like, Morgan Housel or, like, Malcolm Gladwell.
26:30The way they didn't start with books. What they actually started with was a tweet that performed well.
26:36That tweet that performed well, they turned into a blog post. That blog post that performed well, they repurposed a bunch of ways. And then ultimately, like Malcolm Gladwell's book I can't remember.
26:49Was David and Glide or another one. Literally, it's just a article from The Atlantic or The New York Times that performed exceptionally well that he turned into a book. Morgan Housel's book, Psychology of Money, is just a repurposing of his best performing blog post.
27:01But it's the same concept of how do you, as low resources humanly possible, test something before you know that it works, and then put more resource into it.
27:12And so the point we're at now is, like, you've run all these different derivatives of ad creative and meta.
27:18You see what works well. And then once you know what works well, then you start putting more resource into how that messaging shows up in other parts of the funnel.
27:26Totally. And I I also think that like the creative doesn't have to be perfect to begin with as well. Right?
27:32Like, we have these 40 variations that we just generated. Right? I'd go.
27:36I hit download. But the variations like, I'm just gonna go back to the core of this while that's running in the background.
27:44The variations, like, once I find a winner, I can always come back to this and have a human touch this up and improve it. Right?
27:51Like, people, for some reason, they get stuck on, like, oh, this isn't malleable and, like, the Internet, you can't just, like, iterate on top of this. This is how software engineering has, like, functioned for the last, yeah, I just if we're since the dawn of everybody doing, like, what we're doing within software. Right?
28:06You gotta think about it in that capacity. We now have like, for example, I see this with AI avatar UGC ads all the time where they're like, oh, I can tell that it's AI avatar I'm like, great.
28:17You can tell. Some of the best performing ads that I've seen have been AI avatar ads. And then you know what works what's even, like, more interesting is when we take that ad and then we go and we use, like, v o three.
28:29So what my process as an example is I would go and I would ball crate ads with HeyGen, and then I would test them. I would find a winner.
28:36I would take that winner. I go to v o three, and I try to make a better version of that winner with that same ad script. The concept is what you're testing.
28:44The the the medium platform that you put it on to put it out in the public can can evolve and can change. But what I'm trying to get to is the messaging component, and this is a way that at scale I can do that.
28:56And then, okay, I find that the v o three one, that's not performing as well as I think a human could do it. Then at that point, I can go hire a human. They read the ad script and we turn that into an ad.
29:05But what I'm seeing a lot of the times is that the iteration that I can do, I can get a better I can move faster than what I can do when I incorporate a human into that process. Because I have to wait on them. I have to, like, you know, give them feedback on how they're doing the script read, etcetera.
29:21And I can get an output that is I I can just move so much faster than what that human in the loop ends up looking like. And if I can move faster, a lot of the times I can win.
29:30So, again, I'm talking about startups, though, too. Like, we're an early stage company. This is way different when you're at a larger organization.
29:35But there's ways to incorporate this into your workflows that I I I think people don't like, if you just start to think in this way, you you you can compound the experience that you have. Like, you you're you're no longer like a purse like a single person that's joining a company.
29:50Like, you're a person with , and you have all this personal software that you've written that you're now bringing to the table. Okay. What levers does that give you to negotiate from a salary standpoint?
30:00Like, not just hiring me. You're also hiring my system that I've developed and that I'm going to bring into this organization while I come. So anyway, I don't know if there's questions around that I can try to answer.
30:10But
30:11Yeah. What percent more productive would you say you are given your use of these tools versus old Cody?
30:18I've done I've done probably more work in the last two weeks than I had done since October probably.
30:27Like, I I can't even put a number on like, it's it's it's infinite. The the problem now is no longer like the things that can be done.
30:36Right? Like, I can I can we could be in here right now and be like, here, we'll just do this? I'll just be like, go and make landing pages.
30:42Like, I'm gonna do AI dashboard generator. Right? And then I'm gonna use this tool called Keywords Everywhere, and it's gonna help me find all the long tail keywords related to AI dashboard generator.
30:52So it's gonna pull all of these out. It's gonna extract these. Right?
30:55I can take every one of these keywords. I'm gonna download them right now. I'm gonna copy them.
31:02I'm gonna go over to Klon, and I'm gonna say make a landing page for each of these.
31:10Someone saying that you're in crisis after understanding what is possible.
31:13Yes. This is literally what it comes into because you you you the I'm I'm like losing sleep over this to the because of what you can do.
31:22Like, feels like you're at this, like, tipping point. Again, with the domain knowledge that I have, like, have experience in what you know, these aspects, like, these deep understandings. If I can go and apply this to that, right, it creates this compounding effect in the work that I can do that's just at a ridiculous level, basically.
31:40And so cool. It just said, like, it just set this up. Seven new pages are being created now.
31:45It's gonna give me the URLs for those pages. I can take a look at them in a draft mode, and I can hit publish. Or I can just say, hey.
31:51Just hit publish, and let's go, like, right into the pro. And to create those pages, because I'm sure like I I had a similar question as you were creating the bulk Facebook ads. Like, what's actually creating the design there?
32:02Like,
32:03what what's
32:04yeah. Yeah. It's just the template of the homepage.
32:07So all I've done is like forward slash like page. Right? And I'm just gonna go to this test page integration.
32:13Right? Yep. It's just this.
32:15It's basically the exact same thing of the homepage. Only the title and the p one in the hero section is being changed.
32:23The reason that we do this is because when we see that when we align the ad and the landing page that we're sending them to, we can increase the conversion rate that's happening there. Like, this is why you build you know, this is why there's, like, thousands of landing page builders. Right?
32:37Like you could you go for hours talking about like Leadpages or like Yep. You know, any of these. The whole idea is I'm trying to like optimize that funnel so that it feels cohesive to the user that's coming from an ad to the page that they're landing on.
32:51What I'm how I'm thinking about this now is like, okay. Well, this is just a single landing page format. I can go and make as many different variations of landing page formats and then go and test those against each other as well.
33:04Like, expands kind of infinitely in whatever direction that you can imagine. And and I'm also not, like, hindered by any engineering resources with everything that I'm doing here.
33:15Right? Like, I I'm having Claude set up this infrastructure. I'm then having Claude do the work for me of, like, you know, page by page building out landing pages for every one of these.
33:26Right? And then I'm basically deploying those. I I what this turns into from here too is like, Now I want to go and I wanna submit these to Google Search Console so that they get indexed.
33:38I, like, ask Claude, okay. What's the sitemap that these are under? I then say, okay.
33:43Hit the Search Console API to submit these so that they get seen or by so that this sitemap starts to get indexed by Search Console.
33:52I see that those pages aren't actually getting indexed. I can hit the web indexing API from Google's that Google provides to basically do like a manual ask for the indexing of that page.
34:02The the like, it's basically the exact same work that you're doing, but just employing the Cody, you're you're you're
34:10you're you're frozen for a second. I think your your computer is like about to take flight.
34:16They're like they're like, I'm trying to match Cody's energy right now, and it is not possible. I'm simply a robot. But but I think the I think the the gist of it is, like, you you're able to create at such abundance now.
34:30I guess my question for you is is, like, what is what is the biggest bottleneck in a marketer or head of growth's job now if you're functioning like a go to market engineer?
34:41Yeah. The the the biggest challenge that you're gonna be facing is the amount of data that you can produce is at a scale that you just, like, never could previously. Right?
34:50And this is actually, like, the whole origin of the company that I'm building. Right?
34:55It's like everybody right now can go and build a thousand Facebook ads. Right? Like, I just literally showed you how to do this.
35:00But how do I understand what's actually working? Like, what is actually happening within that data at the scale that I'm at? Like, this is this is the challenge that all of these companies are going to be facing.
35:10And this is, like, exact so we tried to solve this too. Like, I initially tried to build this with like an MCP or just like hitting an API endpoint to pull in the data. Like, Facebook ads alone for like one of our customers creates 25,000,000 rows of data a month.
35:23Right? Like at the scale that they're at. They're deploying about a $180 of like spend a month.
35:28The there is no way through an MCP or through the Facebook API that you could basically hit that. So be before we get even get into that, though, like, I'd love to do an actual build just to, like, show people the pro Let's do it.
35:40Where it's like, hey. We're gonna scrape LinkedIn comments and, like, people that have engaged with profiles. We're gonna use this tool called Phantom Buster to basically, like, pull all that out.
35:49We're gonna send that to Apollo, and then we're gonna send that to Instantly AI to actually, like, cold email these people. Just to show people Yeah. Well, let's let's do it.
35:56Looks like. So And if you're Wadi and you're at LinkedIn, I apologize,
36:00but this is gonna be awesome.
36:03Watch my decal's gonna get nuked after this. Give me one second. But but I just wanna show people, like, what this process, like, actually looks like when you're starting from zero just so that they can understand.
36:14So I'm gonna start an entirely new window. So I'm gonna go into terminal. And, again, I'm gonna go into that documents folder.
36:22I'm gonna go into that agents folder, and then I'm gonna start Claude, and we're gonna start from scratch. I'm gonna say, okay.
36:29I wanna build a new workflow that basically will take the engagers from a LinkedIn profile sorry, from a LinkedIn post that I find, and it will then go and use the Apollo API to enrich those individuals.
36:44It's gonna pull out the emails of those people, look for the Apollo API key within the environment file. All this should be in within the environment file already.
36:52Ask me for any API keys if you need them. And then I want you to then go to the million verifier and validate that email and then add it to an instantly AI campaign.
37:02Let me know if you have any questions. So at this point, I'm then gonna turn it into plan mode and have it basically walk through the process of like, what do you need from me to get this output that I just asked you for?
37:17That is literally the origin of this. And twenty minutes later, you're gonna have that whole workflow built.
37:23Like, you're just basically providing it what it needs to do this activity for you. This is like how simple this is, and this is how. Yeah.
37:32I mean, I was building a similar I was building a similar thing yesterday where I want to I've noticed there are more go to market engineer, chief AI officer, head of AI roles.
37:42And to me, those companies that post those jobs are great potential customer for 10 x. And so basically, I just similarly to you, and we have like a certain planning prompts at 10 x, but you can, you know, quads plan mode is is more than sufficient. I basically was like, build me something that every day, uh, goes through and searches LinkedIn jobs that have these three titles or titles like them at companies that have more than 200 employees.
38:12If you can't get it directly from LinkedIn, use Exa because I know Exa has basically scraped all of LinkedIn, enrich the data so that you know which of these companies actually have more than 200 employees, and then set up a Slack integration where every day in Slack, we get a message with at least 15 jobs that have these titles at companies of this size that we can send DMs to.
38:32That's where it's at now. We can take it a step further, obviously, and actually, like, automate the drafting and DM process.
38:37Exactly. This is the whole like, so what you just described is where all of my time is spent now is figuring out, like, what is that arbitrage.
38:48Right? And then how do I go and basically, like, and this is the game now.
38:53Right? Like, you're you're competing against, like, people like me, like you, like, you know, our friends that are all doing this type of work.
39:02Like, this is how sophisticated it's starting to become. And, like, I I I'm, like, honestly just, like, afraid. Right?
39:09For, like, I I, you know, I I know people that are in these roles that, like, aren't, like, adopting this or their or their jobs won't let them this yet. There's just this huge opportunity cost basically to, like, not, like, start doing this immediately.
39:23I and I would like the other the other piece of it is, like, how do you communicate this and, like, show this internally of, like, the activities that you're doing? And, again, like, the with the system that we're like, just to show you an example of this. Right?
39:36So, okay, I I spun up this campaign and I'm gonna go and I'm going to extract the ad set ID.
39:42Okay? And then I go over to graft and I'm like, cool. Like, make a dashboard about this ad set.
39:51So I've selected like the like Facebook ads and then I'm gonna hit send message. So this is gonna go and make a dashboard for me about this campaign that I just spun up.
40:02While it's working on that, in the background, I'm just over here doing whatever it is that I'm working on with Claude Code. Yeah. Yeah.
40:09So let's keep let's keep this let's keep this LinkedIn thing going. I think it'll be awesome for people to say. So I built this pipeline already.
40:14So what I'm gonna say is build like something else. I'm gonna be like, do it from scratch.
40:22To show the process. So gave it the output, the outcome that I want. It so it just read my code base and understands, hey.
40:32You already have something like this. Like, are you sure you wanna basically do this, or do you wanna do it a different way? So now what it's going to do is basically go and design this whole system for me of what I just described.
40:43And this will have some back and forth that's necessary where it's basically gonna ask me, hey. Like, I need these API keys. Like, here's where to it'll walk you through step by step.
40:52Like, here's exactly where you need to go to basically get those API keys that are necessary for me. Here's the permissions that I need. It's crazy, man.
40:59Like, on Facebook ads in particular, traditionally a very, like, challenging API to interact with. It is so it it understands the API so deeply.
41:11Like, it it from a Totally. It's it's honestly like, I I can't even find good documentation on Facebook alone, like, about the API to, like, if I was trying to do this myself.
41:21But by it going and interacting with the API, it knows what it needs to like, it it can guide me through this process, basically. Yeah. It sounds ridiculous.
41:29It seems almost like you're a vessel for what it needs from you to do the activity that it kind I mean, that that is exactly what it is. Like, I think it's really important for people to realize that with Claude code,
41:41it's it's one of those things where you you really do not have to understand software engineering. You just have to be a clear enough thinker. But, honestly, ClaudeCode is getting better at even if you're not a clear thinker, you're a disorganized thinker because it's really good at inferring what you want.
41:57And anytime you don't know something so if it's like I don't know. Using the example of what I built, like, hey. I wanna scrape for these things from Exa.
42:05If it's like, oh, you need to go grab the sign in for an Exa account and grab the Exa API. Let's just say I didn't know what the Exa API was. I could just ask it, how do I what is an API?
42:15Or how do I go grab the AXA API? And it gives me the act exact directions.
42:20So and to the point Cody made, it's like because Claude code and these tools can now just build software from scratch, It's using, like, people as a vessel to get all the information that it needs to build these things, and we are just a conduit for giving the information it needs to basically build software from scratch.
42:39Totally. I think the other thing just to piggyback on, like, we had this conversation two weeks ago, I would have had, like, n and n within this workflow. And, like, what I found is that I'm actually now just going directly to, like, code.
42:51Like, I I just skipping n and n as, like, within my stack and just going like immediately to like writing. So for example, like I I needed to set up a basically like after a sales call, wanna give a Notion document of the recordings that it writes an email draft for me and based off of like what the follow-up steps are for from the onboarding call.
43:16I needed a server to be able to have this run-in in perpetuity in the background all the time so I could like send a Slack message and have it run. So I gave,
43:25like By the way, so so you guys know because I think, like, you may be watching Cody right now and you're like, holy hell. This guy is frenetic and, like, he's context switching to all these different things. It's one of my theories that, like, multitasking has always been, like, a frowned upon thing.
43:42But what he is doing is actually not multitasking. It's multithreaded work is the way I view it, where it's now as things can run-in parallel, the only way to work is going to be working on multiple things in parallel because the cost of not doing so is someone else who is able to do so because you actually don't have to divert your attention across many things because another coding agent or another Claude is going to have all of its attention focused on that one task.
44:07Yes. And, like, this again, I'm I'm just basically jockeying these agents is how I think about it.
44:14Right? Like, I'm just going back and forth between them and it's like guiding them or doing what they need from me. And just to come back to this, so, like, I needed a server and so that this software could run-in perpetuity in the background.
44:26Right? So, like, basically always beyond, not just using the CPU on my machine.
44:31So I gave it my Railway API key, and I'm like, cool. Set up, you know, a types you know, set up the the server that I need.
44:38It was just like a Node Express server. It set that up for me. It launched the software on It pushed the the the repo to GitHub to be able to do this.
44:46Like, all of this, it it handled on its own. Again, while I'm working on these other activities, whether it's like creating dashboards or analyzing the data or building out a new workflow for whatever like opportunity that I see within the distribution stack that I'm going after.
45:02So anyway Totally. I don't know if there's other questions on your Friday. Where are we no.
45:06No. So where are we in the the LinkedIn build? Yeah.
45:09So in the LinkedIn build, it's currently like, you're it basically created this action plan, and it's now going to go and, like, implement what I just described.
45:20So
45:21Cool. And ultimately, what you would have to do for this to be, like, fully built is you're gonna need, like, an API key for the thing that scraped from LinkedIn.
45:30You're gonna need an API key from Apollo, which is gonna enrich those leads from LinkedIn. You're gonna need an API key for Instantly, which is the thing that sends, um, cold, uh, outbound emails. And there's one other API key.
45:42What's Million Verifier? Million Verifier is just a email validation software to check that the email is, like, actually valid so you don't nuke the cold email can't like, domain Yep. That you're sending from.
45:53So Yeah. So basically, what's gonna end up happening probably is that the only work you're gonna have to do is get these API keys from these different sites. If you don't didn't know how to, you could just ask Claude how to do it.
46:03You're gonna feed it to Claude. And then once it's done and you test this out, if it works, amazing. If it doesn't work, you just feed the error back to Claude and tell it fit figure out how to fix this for me.
46:15Exactly.
46:16Honestly, it'll just do that recursive loop for you. And this is like why it's so powerful. It's like if it sees the error that's occurring, like for example, we had this railway when I was deploying the server.
46:25Like, there was some bug that was occurring. I just looked at the logs to see what was happening and then fixed it so that it could do the deploy. So this is now completed.
46:34Right? Like, it's now done this basically. I can now just drop in a LinkedIn URL here.
46:41Like, what how I'm going to extend this, like, just to communicate it. It's like, I'm gonna build this as a Slack function. So I do, like, forward slash LinkedIn enrichment.
46:49Right? I drop in a LinkedIn post URL, and that will just fire this whole thing off and that automatically gets added to the Instant Lead like AI like campaign that I'm running that cold outbound strategy from for people that engage. So as I'm scrolling LinkedIn on my mobile, right, I find a post that's relevant to the software or the product that I'm building.
47:08People that interact with that post, they're valuable to the company. I taped that post link, I dropped that into Slack, and this whole thing fires in the background. So I this is this is the compounding effect that can happen.
47:20You could also just provide this as a tool to your whole team as well. So they can just, like, feed this into it, or you can have it where it's, like, automatically scraping for these posts using something like Appify or Rapid API to go and extract the, like, most viral pieces of content within the category that you're in.
47:38Or if you know that specific creators make content that is super that is like, they do that on a consistent basis that your ICP interacts with, you can go and follow them. So, like, this is something I'm building right now is I'm basically taking, like, all the incumbent competitors, CEOs, LinkedIn profiles.
47:53I'm looking at their their content to see who engaged. And then that typically is a great signal that they're, like, in a buying motion with that company. I then take that and then go and reach out to them.
48:03So these are the like the levels of the game that you can play with this basically. So And anything so we have about ten minutes left. Is there anything else you wanna show in kind of your build before we hop into questions?
48:16No. I I I think the only, like, again, the only thing that I think is going to hinder people is, like, thinking about you can do anything now.
48:27So, like, what should you do? And, again, like, going back to those, like, the data is going to drive all of your activities. Like you have to have a solution for whatever it is that you're like just going like I I see these like tutorials all the time that are like, here's how to do like you know this this action where it's like for example, like I'm gonna make a thousand Facebook ads.
48:47Like if you don't have like an outcome that you're moving towards or a way to measure this, there's no point of what you're doing. Like all the AI slop tooling in the world will, again, will enable you to to create as much content as you want.
49:03But unless you can actually analyze that information, that's that's the only way that you're gonna be able to, like, scale with this. So anyways, that's my only thought.
49:12And two two other things here. One is we're gonna do questions now. If you have questions, put them in the q and a part of the chat so that it doesn't get drowned out by, honestly, the amazing conversation that's going on here.
49:25There's literally conversation in this chat about people setting up a go to market engineering WhatsApp group to keep the conversation around this going. But, yeah, drop questions in the q and a. I'll take a look at those.
49:36And in the meantime, I got this question. So we purposefully make these conversations, I would say, technical ish.
49:47Like, not technical at the at the level of engineering, but technical where you are for sure gonna hear terms that you haven't heard before if you're nontechnical.
49:56And that is purposeful. And and it's purposeful because our belief is that really high agency people who are joining this conversation and wanna stay in the frontier of this technology are going to do the work to get those questions answered. So I saw someone say like, I don't even know how to get a repo on GitHub.
50:12I don't know what, like, you know, exactly how an IDE works and what a CLI is and all these things. So I didn't know what any of these things were ten months ago.
50:23I literally did not know any of them. But if you, one, spend time in this sort of conversation with people also looking for the answers, and honestly with Claude Code, there actually is just no excuse now for for not finding the answer to all these Everything.
50:37Everything I talked about today, you can literally be like, I don't know how to do that. Like, tell me how to do this, and then we'll walk you through every part of that process. So Yep.
50:47Cool. Okay. So we're gonna hop in some questions.
50:50First question from Tom Babb. Tom said you listed about 20 tools. How are you discovering them?
50:56Google, AI, like where are you finding the tools that you use? Yeah.
51:01I I honestly,
51:02it's like I've been doing this this is a terrible answer, but I've been doing this so long. I just know where these things are. The like, Twitter is Twitter and YouTube is where all of this is happening right now.
51:14Full stop. It's so funny because, like, I'm not really active on LinkedIn. Like, I'm I'm it's a it's a channel that I'm really bad at.
51:20I'm just, like, not good not really built for the content that's, like, works there. I'll see stuff that went viral, like, six months ago on Twitter, And, like, that, like, has already been adopted and implemented and, like, basically, like, the arbitrage has been exhausted, and it will show up on LinkedIn.
51:36And it's just hilarious. So, like, it it's it's kind of assess pool, like, full, like, full warning.
51:42But, like, that like, where all this is happening is largely on Twitter and, like, YouTube. With that said, though, honestly, like, using a like, if you're trying to solve a problem, like, okay. How do I extract, like, engagers out of, you know, like, LinkedIn posts?
51:56Right? Like, if you just go to Perplexity and be like, what are the five best, like, tools for this? It's gonna give you examples of each of those, and then you'll be able to go and implement those.
52:05And, honestly, I've had Cloud Code, like, share stuff with me where I was like, oh, I didn't even realize that this was the solution that would, like, be a better like, for the outcome that I'm looking for. So I I think just, like, there's no there's no rule book with this right now. There's no, like, good educational material.
52:21It's all just kind of happening in real time. Like, everything that I just showed you is, like, transpired over the last two weeks. Right?
52:27Like, that's how that's the velocity that this is basically moving at, like, with currently.
52:32Okay. We have a ton of questions in the chat. I can just keep them rolling.
52:35Yeah. Will said try be quick. Everything you've talked about today seems like a tool for just you to use.
52:40How do you think about deploying these tools to a team for them to use as well? Yeah. So what I just showed, like, that can be a shared like, a a piece of shared software.
52:48Right? So for example, like the bulk ad generator.
52:51You could just deploy that. Like, you could just literally put that on a URL and then provide that to your entire team. Or like how I see, like, more sophisticated teams doing this is they're, like, having shared GitHub repo, And they're basically, like, as they make changes to the code base, they, like, merge that to the GitHub repo so everybody's working on the same instance.
53:12There's multiple different ways that you could go about that. It just depends on, like, how sophisticated you wanna be, like, treating this as, a software engineering activity or as like a I'm just building internal tools and sharing them with my team.
53:23And and again, you don't need to know how to get your repo on GitHub or how to merge a PR. You can literally just say, I wanna share this with my team. What's the best way to share it?
53:33It Cool. Google Right?
53:35Yeah. Exactly. It's like cool.
53:36Here's this URL that it can now go to and you just deploy it onto Vercel. You don't even know what Vercel is. Right?
53:41But it's Exactly.
53:42Can access it. So Okay. Next question from Kate.
53:45Do you use this post on your owned channels or do you just test on satellite accounts and paid ads? How do you think about quality control at scale?
53:54So basically, it's like if you're testing all these different ads, do you worry about knowing they're gonna be bad ones that you're testing on like your core channel? You could totally have that separation. And then like basically, like, the the the the paid ad spend for the winner is the the
54:10basically, the the lens that you, you know, that you sift through to then move to your main. We do this a lot with organic social where it's like I have burner accounts that we're posting ideas to. And then once we find, like, banger ideas, we'll then pull those over and then put them onto mains.
54:25It just depends on again, the it it it depends on the situation of the organization that you're in and how flexible they are with, like, this type of this type of testing.
54:35But I think if you go to them and you're like, we can create better outputs. Like, I need to be able to do this. Like, there's the conversation that you can have.
54:42And then proving that to them, like, here's this KPI dashboard that is showing you that I just decreased, like, our CPA by 20% because I'm now doing this, like, you know, ad velocity testing that we're talking about. So
54:54Yep. Love it. Eduardo asked, I'm sure other people are wondering, can you share again what is the process you use in order to get the meta ad creative done?
55:01So can you just, like, almost just go, like, step one through whatever step of how you do that bulk upload? Yeah. Totally.
55:07So so step one,
55:10I would go to Claude and, like, I I would basically be like, I wanna build an bulk ad generator. Here's an example of an ad, like, you know, square format that I want to create. Like, help me do this.
55:24It'll put it in plan mode. It'll literally make a plan to execute this.
55:29It'll ask you questions on, like, how, like, variable do you want the generator to be. And then you like, literally, that walks through. Imagine, like, ten minutes of you going back and forth with it describing it.
55:40You hit go. It will make that box generator. And then to redesign that that, like, piece of that, like, ad creative.
55:51So, like, this is something where we having the language to describe the thing that you're trying to describe is the hardest part of this now.
56:00So for to give you an example, I have a cofounder, super technical. Like, the depth of knowledge he has and the language and the vocabulary that he can use to describe the outcome that he's looking for is like so much deeper than I can.
56:13Right? So and to get to to to zoom this out, to give an example, if I go to Claude and I'm like, write a blog post about x. Right?
56:19It's gonna write the most average thing that you've ever seen. But in contrast, if I interview myself, say I'm an expert about whatever this topic is, I'm gonna like do a transcription where I talk for ten minutes. Like, I just monologue about the topic, my opinions, my views.
56:34I use that as source material. Then I also go scrape what's ranking on page one of Google, and I bring that in as source material as well.
56:41I put both of those into context and then I'm right, okay. Now write a blog post about this based off of the content that I provided. Use the source material as you need.
56:49Use the vocabulary from the transcript that I provided. The output quality that I'm gonna get from that is gonna be, like, top, you know, 10 to, like, 10 to 1%.
56:59Right? Like, you're gonna be finally in that, and and it's the same idea here. So when you're this is the actual the hardest part.
57:05Like, I I'm act I'm actively doing this right now. I don't have the design language to describe what I'm trying to do with the creative to create the out like, the the visual that I'm looking for.
57:16So, like, something I just learned yesterday was like, oh, you can talk about a grain opacity. So you basically, like, overlay a grain, like SVG over the top of the background, and then I can put the opacity levels at a different size to basically create this almost like effect where it feels like there's like texture to it.
57:32And I was trying to tell it, like, make texture. Right? Like, I I don't have the language.
57:36And this is this is the actual, like, hardest part of this is the acquiring of that, like, domain knowledge. And this is why people that already have that domain knowledge that implement this are so much more effective at using this tooling. So
57:48Totally. And so just the to complete the steps. The step was basically you have Cloud Code build the bulk ad generator.
57:55Yep. You give it an ad that you want to be kind of the source material.
58:00Then what's the next step after that? Yeah. So give it the ad you want it to be the source material.
58:04It's gonna go and make a variation of that ad. You then go back and forth with Claude with the changes that you want. And you can just ask it.
58:11Be like, hey, this doesn't feel super cohesive. The first version that it did, for example, purple, black, brown, like, very like, it looked terrible. Like, tons of colors that weren't on brand.
58:20I'm like, cool. Here's our brand style guides. I want you to, like, you know, mimic this as much as you can.
58:25How can we make changes to the ads so that it fits this? I want I go down that texture rabbit hole because I'm trying to make it have, like, this feeling of depth. So once I get that that the image to the quality that, like, I want, like, the template to the quality either to the outcome that or the the the template to the level that I want.
58:42At that point, then I can say, okay. Now I wanna, like, add a bulk generator to this. I'm gonna basically we're gonna change the title.
58:48We're gonna change the the the paragraph. Help me brainstorm those ideas of like what should be in here. Again, I showed you how to do that with Perplexity and using Claude.
58:58You could do that all within Claude code. I was just, you know, showing you a different way to do the same outcome. It's gonna generate those ideas.
59:04You can select the ones that you think are the best, and then you go and do that bulk generation. That whole process again, like that's like 30 of time to get that thing set up that I just showed you how to do.
59:15Love it. Last question for you. What are,
59:19let's call it the top five most important tools you use in your stack for doing what you do today?
59:26Yeah. I think it's it's it's I just try to keep it simple, and it's like literally, like, on the most slept on okay.
59:35So just to take a step back. I think the ability to email, like, cold email people is so underrated.
59:42Like, I don't think people understand. Like, I can the the amount of information that I can just, like, scrape from the Internet and then do bulk email that's customized at scale.
59:51Like, just having, like, a ton of domains sorry, a ton of inboxes that are in instantly. So we I, for example, I use HyperTite. For all of my domain infrastructure.
1:00:00They basically set up I have 2,000 inboxes that I'm sending from that I put into instantly. And then I'm just running like multiple campaigns out of out of instantly for whatever activities that I'm trying to do. Right now we're pitching me to go on podcasts.
1:00:12I'm pitching YouTubers to basically like, you know, do influencer marketing or creator led marketing for us. And then I'm also doing like traditional cold email to our ICP. I think that that is something, like, having that ability to just tap into that with, like, whatever out out, you know, outbound you're trying to do, that's something that's super powerful.
1:00:31Like, again, this clog code, like, piece, like, I still do use cursor some just for, like, small pieces, but my stack is basically clog code. All of the all of the APIs that I work with on a daily basis, I have that in that environment file. And then I'm, like, living out of graph.com, which is my company for, like, all of the analytics.
1:00:49And so, like Got it. For example, just to show people like what this ends up looking like. So we just started that campaign, right, of like we're trying to go and basically like track the outcomes of this like ad creative that we just generated.
1:01:02I basically was like build me a dashboard about the ad set with this ad set number and I now have like spend. I have breakdown by the conversions. I have like where it's actually being shown on platform.
1:01:12If I wanna modify this or change this in any way, I just chat with the dashboard. If I wanna change the chart style where I'm like change this to a bar chart, it does this.
1:01:22And then I can share this internally with my team so that I can actually create like, you know, basically buy in from the organization that I'm working on with. Here's this KPI dashboard that's actually showing the outcomes that I'm doing based off of this work that I'm doing. So anyways, man, that's that's really it.
1:01:37I've try to keep it as simple as possible. I think people get obsessed with the tools, get obsessed with the outcomes. The tools are it doesn't matter anymore.
1:01:44Like, there's no limitation on the tools and what you can get them to do. It's the hardest part is knowing like what should I be doing and then like how do I polish that and measure like what is actually working for my company. So Love it.
1:01:56Cody,
1:01:57this was awesome. Appreciate the time and
1:02:00look forward to having you on again in the future but I thank Thanks for having me. Truly on sales channel. Everyone
1:02:04learned a ton. Cheers. Cool man.
1:02:07Thanks everyone. Yeah. Thank you.
§ · For Joe
Build the system, own the system.
GTM-as-software playbook
One person with Claude Code and domain expertise now out-executes a 20-person marketing team — but only if the domain knowledge comes first.
- Start with your .env file and a folder you live out of — Cody's entire system starts from documents/growth-agents.
- Treat every distribution activity as a sprint: build the script, test it, measure CPC or conversion, automate the winners.
- The Multithreaded Work frame is the JoeFlow/Batch pitch in one sentence — steal this exact language.
- Perplexity + Reddit scrape + Claude for copy is a repeatable ICP-to-creative pipeline that takes 30 minutes to set up.
- Strapi + Claude Code API calls = programmatic landing pages at any scale without a dev team.
- The concept is what you're testing; the medium (AI avatar, Veo 3, human) evolves cheaply once you have a winner.
- Personal software portfolio = career leverage — bring your agents and your code to every engagement.
§ · For You
You don't need to know how to code.
For the non-technical marketer
Every task you do manually in marketing today can be automated with Claude Code — and you don't need a computer science degree to start.
- Pick one repetitive marketing task you do every week. Describe it to Claude Code in plain English. Ask it to build a tool that does it for you.
- You don't need to understand the code — you just need to describe the outcome clearly.
- When Claude asks for an API key, it will tell you exactly where to get it and what permissions you need.
- When it breaks, paste the error back in. It fixes its own bugs recursively.
- The hardest skill to develop is vocabulary — learn the names of things (CPC, Apollo, Strapi) so you can describe what you want more precisely.
