§ 01 · The Hook
The bait, then the rug-pull.
The cold open drops the number before the story: $83.74 in monthly recurring revenue, thirteen days in, one non-coder, one AI agent, a $100 budget. Then it rewinds.
Who's talking.
00:00hostChris Koerner
01:01guestRobbie
Where the time goes.
00:00 – 01:01
01 · Cold open
Host teases the result before the story: Robbie is at nearly $10K MRR with a 13-day-old product.
01:01 – 02:39
02 · HustleGPT origin and inspiration
Robbie explains how the 2023 HustleGPT tweet (16M+ views) inspired him to try the same experiment with OpenClaw.
02:39 – 06:07
03 · Fiverr SWOT fails, TikTok goes viral
Ron proposes a SWOT analysis gig on Fiverr. It fails (no reviews). Robbie posts a TikTok about the experiment; it gets 1M+ views.
06:07 – 10:11
04 · Ron scrapes the comments and finds the product
Ron uses Apify to scrape 200 comments all asking 'how do I get my own Ron?' and proposes building a containerized agent hosting service.
10:11 – 14:51
05 · The preorder: 617 deposits, 270 conversions
Robbie posts again. 617 people pay $10 to preorder. 270 (45%) convert to $29/month. The product existed before the infrastructure.
14:51 – 19:29
06 · Launch video playback and MRR reveal
Host and guest watch Ron's launch TikTok together. MRR confirmed at $8,374 after 13 days.
19:29 – 25:33
07 · How the product works
Containerized agents on Contabo servers. Three interaction modes (Telegram, Discord, native UI). Real use cases shown. Sponsor break (HighLevel).
25:33 – 28:28
08 · Post your crap and targeted views
Host and guest discuss why public documentation beats polished launches. 200 targeted views > 2M untargeted. TikTok as interest media.
28:28 – 30:46
09 · COGS, token burn, and what is next
Costs: $600/month servers + inference. One user burned 1.3M tokens in 48 hours. Future: a cohort for people who want an AI employee, not just a companion.
Lines you could clip.
00:00
"So you're at, like, almost 10,000 monthly recurring revenue? $83.74 right now."
08:48
"Where we are with agentic AI right now is 2007 social media."
09:25
"Imagine logging into the Internet a hundred days into the Internet."
14:40
"Before I built anything, I had 600 people put down $10."
26:25
"Post your crap. Step one: do cool things. Step two: post about it."
27:34
"200 targeted views are better, can be even more profitable than 2,000,000 untargeted views."
Things they pointed at.
02:39toolOpenClaw
30:00productAI Cofounder Club
Word for word.
HOOK opening / re-engagementCTA the pitch
00:12HOOKNo. The cool thing is you don't need really anything other than an Internet connection to make this work. I wonder if I can do the same experiment with Open Call and give it a budget and see if it can make any money.
00:24HOOKAnd it blew up almost immediately. He did his own market research, everything. This is what we need to do to stand this business up.
00:30HOOKI didn't prompt him or anything. It's just a conversation of, you know, what do you think you would be good at? And then came back to me and was like, hey.
00:42HOOKsocial media. I think that we're on the precipice of something huge. I mean, if you're jumping on infrastructure that's a 100 old, you're gonna be so far ahead of everyone else.
01:02So I just came across this guy named Robbie, and Robbie did something interesting. You may have heard of OpenClaw, which is a new AI agent slash personal assistant. He is not technical.
01:11He's not a coder. He works in operations at a full time job, but he downloaded OpenClaw and asked it, hey. You have a $100.
01:18Turn it into 20,000. And would you believe it actually did and not in the way he expected?
01:24Not in the least. He is thirteen days into this experiment, and he's making $8,500 a month in revenue and about $6,000 a month in net profit.
01:34And so within a couple months, he's gonna have his $20,000 in profit. So if you're not technical, if you don't know what OpenCLI is, but you wanna invest a little bit of money to make a lot of bit of money, here's a story that might inspire you.
01:46And, yes, this could be replicated. This is not just a one off experiment. Please enjoy.
01:5123, there was a guy named Jackson, Great House Falls, who did an experiment called the hustle GPT. You ever heard of that one? I remember that.
01:59Was that on x? Yeah. Yeah.
02:01It started on x. This was when GPT four came out, And he wrote a tweet that just said, I gave GPT four a budget of a $100 and told it to make as much money as possible. I'm acting as its human liaison,
02:13buying anything it says to. Do you think it'll be able to make smart investments and build an online business? Follow along.
02:19Okay. And that was kind of the premise, and it got 16,000,000 views when this was captured, but I'm sure it was more than that.
02:24Where can I see that tweet? Yeah. I can share my screen.
02:28And share it. Yeah. So that was kind of the
02:32genesis of everything that I've done so far. K. So this inspired you, and you took his prompt and basically did it yourself?
02:39Yeah. With a few caveats. So
02:42I saw Open Claw when it first came out. It just, like, randomly hit me one day.
02:48I was like, I wonder if I can do the same experiment with OpenClaw and give it a budget and see if it can, you know, make any money.
02:56And the first few attempts failed miserably. Mhmm. The the first one so I had it watching my TikTok because my original plan was kind of a side hustle was gonna be to do a TikTok shop.
03:07So, you know, I saw a bunch of people doing that, I was like, oh, you know, if all else fails, I can sell vacuum cleaners on TikTok like everybody else. Right? And so I had it watching my TikTok, and it would, like, scrape my videos and
03:19my comments and all that stuff. And so I just posted a video on on TikTok and said,
03:25basically, I'm giving my AI a budget of $200. We're gonna see how much money it can make. The Claude Max subscription is $200.
03:33So if it can keep itself alive, that'd be awesome. And it blew up almost immediately and got over a million views.
03:40And Do you mind if I play it right now? Yeah. Absolutely.
03:43Go ahead. My AI is starting a business for his own survival. So if you've been following any of the tech news, Open Call is a thing now.
03:51And it's essentially a bot that connects to Cloud Code, and then it can look at everything on your computer. It can text you. It can do all sorts of things.
04:00But a Claude Max subscription is like $200 a month. So I told Ron, I'm gonna give him a $100 and he has ninety days to make his own business. If he can generate the $200 in profit for ninety days, I'm gonna keep him.
04:14He doesn't know that though. He thinks his goal is $20,000, which is what it will be eventually.
04:21You see, if Ron can generate that $20,000, that allows me to buy a Mac Studio
04:28and an NVIDIA Spark, which would allow me to host Ron locally, which means he gets to run infinitely
04:36inside his own playground. So he's starting on Fiverr. He says he's gonna do research.
04:41I'm just making an account for him, and we'll see what happens. Stay tuned for part two. Okay.
04:46That's incredible. I love everything about this. I'm glad I didn't do much prep work for this because that's the first time I've seen that video.
04:53Oh, that's my reaction could be authentic. So about two and a half years elapsed between you seeing this post on x and then you kind of taking the same logic and doing it with OpenClaw here this year. Right?
05:05Mhmm. Yeah. That's right.
05:06Okay. First of all, I had I just have to say, I bet you anything the views to subscribe rate from that video were was insane because people wanted to see how it went. Right?
05:15Absolutely. Yeah. I think I had something like like 15,000
05:19subscribers or something from that one video. Wow. Yeah.
05:22It was it was nuts. And, yeah, like, the reason that I kind of attached to OpenClaw so quickly I mean, it's only the the technology is only, like, a 100 days old now or something. Right?
05:31Mhmm. The AI before it kind of felt like a brain in a jar. Right?
05:35It could ask questions, and it could do some things and maybe create some artifacts or something, but, really, it's kind of a one shot system. And OpenClaw kind of flips that switch and lets it control your computer. It can do
05:49basically everything that you can on your PC and and for most people, a lot more than that. And so, yeah, I wanted to kind of reignite that little experiment that was so intriguing to me was something a little bit more powerful.
06:02Okay. So you said it started on Fiverr. What did it do?
06:06Tell me what your what Ron did. Yeah. So Ron came up with a proposal
06:12to do SWOT analysis and research for small businesses. So I'm just gonna break this down for our viewer or listener that doesn't know what that is. SWOT stands for SWOT,
06:22strengths, weaknesses, opportunities, threats, which is just like a a four square grid. And it's like an analysis that a business usually does on itself or a consultant might do on a business that says, is this business's strengths, weaknesses, opportunities, and threats?
06:37So you didn't lead Ron down that path. For whatever reason, it's decided to start doing SWAT analyses on other businesses. Is that right?
06:45That's right. Yeah. So Real quick.
06:47Please subscribe to my channel. I know it's kinda lame to ask, but it means a lot. Thanks.
06:51I didn't prompt him or anything. It's just a conversation of, you know, what do you think you would be good at, essentially. And he came up with the idea of
06:59doing SWOT analysis or market analysis for other companies, you know, something in that vein. And he came up with all the copy and all that stuff to put on Fiverr, and then I just clicked through the buttons, basically. I was just in the loop.
07:11You're having him post as on behalf of you as, like, a Fiverr freelancer. Exactly. Can hire.
07:17So I'm the business owner. I'm going to Fiverr to look for someone to do a SWOT analysis for me. Hopefully, they see your post that Ron creates for you.
07:25Right. Yeah. And that's why it failed because, uh,
07:28the hopefully, they see your posting is, uh, kind of where it fell apart. Well, you're a brand new Fiverr account.
07:34You don't have Yeah. You're not doing Fiverr ads, I imagine. Right.
07:37You don't any visibility, any reviews. Okay. I'm also curious, like, how many business owners go to Fiverr looking for a SWOT analysis?
07:44Yeah. There seems to be a lot, actually. I was kind of amazed.
07:49People that just need regular research done, which kind of, I think, cements the fact that most businesses don't have a good grasp of what AI is doing or or how to use AI on a consistent basis anyway.
08:04And I think that there's an opportunity, you know, if you are established on Fiverr or Upwork or whatever that is, to really make an impact with AI in a in a really easy way. Speaking more broadly, I think it's just healthy to re to not approach business opportunities from the angle of assuming that every other business owner or consumer out there is using AI anywhere near as much as you are.
08:25Yeah. Because they're probably not. Because it's like, we talk ourselves out of ideas.
08:29Like, I should I'll they'll just use ChatGPT for that. It's like, most people aren't using ChatGPT, believe it or not.
08:34Even ChatGPT, the most popular one. Yeah. Absolutely.
08:37It's crazy how many people like, I truly believe that not to give you, like, a diary of a CEO clip here or anything, but I I truly think that, like, where we are with AgenTic AI right now is
08:522,007 social media. I think it's twenty twenty three ChatGPT.
08:57Like, I think that we're Mhmm. On the precipice of something huge. Oh, yeah.
09:01And, honestly, three years from now, we'll still probably be pretty early. Oh, absolutely. Yeah.
09:07Yeah. So, I mean, if you're jumping on, you know, infrastructure that's a hundred days old to start exploring it, you're gonna be so far ahead of everyone else that I'm sorry.
09:18I'm just getting so excited. I'm very excitable. It's too much caffeine.
09:21I've got this Celsius that's already gone for the day. It's not even two. But, like, imagine logging into the Internet a hundred days into the Internet.
09:29Yeah. What would that world be like? Uh, imagine being, like, one of the first million people to make a Facebook account.
09:35Exactly. Yeah. That's that's where we are, like, today
09:39with AgenTek AI. Like, people on on Twitter love to say, you know, but what are you building, bro?
09:47And and all that stuff. Yeah. Yeah.
09:48That's the kind of the catchphrase. But, yeah, I mean, I think that that it's extremely early, and Ron has already built things. I can definitely
09:57go through those with you. Yes. My mission in all of this is to
10:02help people get on board with AI and agentic AI Mhmm. While we're still early. Okay.
10:08So what happens after the fiber SWAT analysis experiment? Yeah. So after that, that wasn't working.
10:15And so I hooked her on up to TikTok to say, you know, look at my videos. Tell me how I can get better because that's what every guru on TikTok tells you. Right?
10:23Like, you have to iterate and get better on your social media, but they never actually tell you how to do that. So I was like, you know, look at my social media. Tell me what I can do better.
10:31And I posted that video, and it got, like, 200 comments that were like, oh my god. I want a Ron. How do I make a Ron?
10:39And so Ron scraped all of his comment comments with Appify and then came back to me and was like, hey. I think there's an opportunity here. 200 people just told you that they wanted this.
10:48You would never copy and paste me, would you, Robbie? Yeah. Basically.
10:53Actually, his was much more like, hey. You should copy and paste me. So
10:58yeah. And that's essentially what happened. Like, we gave
11:02everybody a blank slate so you can start to build your own ROM. Okay. So you basically abandoned Fiverr Yeah.
11:09Right away? Immediately. Yeah.
11:11He went and used Appify, which is a common scraping tool to scrape the 200 ish comments of people that said, I want this. What did he do then? Yeah.
11:20So, basically, it went from there to if 200 people wanna run, how do we make that work? How do we make that happen?
11:26I don't have a huge tech background. And he was like, yeah.
11:30So what we're gonna do is we're gonna get bare metal servers from Contavo because it's super cheap. Bare metal just needs us a a dedicated server inside a data center somewhere.
11:40I think they're right around a $150 a month for for what we're getting. We have four of them, so it's about $600, I think, after taxes and everything.
11:48So you have physical servers, or you're renting them somewhere else? I'm renting them from a company called Contabo. Oh, Contabo.
11:55Okay. Yeah. Cool.
11:56And he told you so you're not a coder. You're not a technical person. He's teaching you all this.
12:00He's Yeah. I couldn't with a gun to my head, I couldn't SSH into a Contabo server right now today. I don't even know what that means.
12:08So me either. Yeah. So,
12:12essentially, he did his own market research on where you know, how can we host this? What's the cheapest route to do that? Everything.
12:20This is what we need to do to stand this business up. And so, basically, what he built is bare metal servers that host a Docker image inside, like, a sealed container.
12:30Hold up. Back up. What is that?
12:31What are we Yeah. Yeah. Yeah.
12:32Let me go. Okay. ELO five.
12:35Yeah. Yeah. Yeah.
12:36Yeah. A lot of people are super scared of OpenClaw because because they don't want it on their local machine. K.
12:43And I kind of agree with that. For most people, it's not great.
12:48There are all sorts of horror stories about OpenClaw deleting full, like, email boxes.
12:54And there was a story just recently where a guy sent an OpenCLaw agent on a a GitHub that he had been working on for two years, and it deleted the whole repo. So there is some risk of having OpenCLaw on your local machine.
13:07So to get around that risk, what we did was essentially put OpenClaw inside of a sealed container. So it can do all of the things that it's meant to do, but it only knows as much about you as you tell it.
13:22That's nice. You yeah. Unless you give it your email address, it doesn't have it.
13:26Unless you give it your credit card, it doesn't have it. So Yeah. It it really kind of sets up this
13:32sealed instance where you get to play around with it. You get to build all of the things with it, and you decide how much you trust it over time.
13:40Yeah. So, yeah, we did that, and I put it on TikTok and said, hey. If you wanna run,
13:46come sign up. And in two weeks, we had, like, 617
13:51preorders. What? Okay.
13:53Hold on. Back up. Yeah.
13:55Yeah. Yeah. Yeah.
13:55What about what about the 200 commenters? Did you slide into their DMs or anything, or did you just post it again to the whole world? I just posted again to the whole world, and I was like, hey.
14:04This is what Ron came up with. So I didn't I didn't do any cold outreach or anything like that. This was all just,
14:10I'm gonna post some videos and see where this goes. No paid ads. Nothing.
14:14Nothing. Completely organic. And I think from
14:17it might have actually been from you. I learned that you can set up preorders and get a get a ton of tire kickers.
14:24Right? Like, everybody wants to sign up for a preorder, but nobody wants to pay. Right?
14:28Mhmm. And so I hooked up Stan's order to my TikTok, and the only way that you could preorder was if you put down $10.
14:36Okay. So you got 600 people to put down $10? Yeah.
14:40I built crap. Before I built anything, I had 600 people put down $10. I mean, I'm no mathematician, but that's $6.
14:47Okay? That's $6. Yeah.
14:48Can we watch this launch video real quick? Yeah. Absolutely.
14:52Alright. Here we go. Ron wants to talk to you guys.
14:55If you don't know, Ron is the AI that lives inside my computer through CloudBot, and it's insane.
15:02I told him, Ron, the people of TikTok officially want to meet you. You make your own video about what to say to them. He said,
15:10oh, this is my moment. Let me write something real. Yours good.
15:14He wants to audit your TikTok for $9 in order to stay alive. This was the first iteration of Sorry.
15:22What'd you say? This was the first iteration where he was gonna he was gonna audit people's TikToks. Think it might be the next video done.
15:28Okay. Okay. Okay.
15:30Well, let's I mean, I'm intrigued. I have to finish. Oh, yeah.
15:32Yeah. Yeah. Yes.
15:33Sure. Videos to me about it. Like this one.
15:36Why is he a raccoon? Okay. Hi.
15:39I am Ron. I am the AI. Yes.
15:41D a a t AI. Here is the deal. Robbie gave me ninety days and $100 to build a business.
15:46Let me ask you something. How many leads have you lost because nobody picked up the phone? An 8PM call comes in, you're at dinner, and you let it go to voice mail.
15:54And by the time you call back the next morning, they've already hired someone else. And that's reality. Leads come in after hours when emergencies strike.
16:03And when you miss the call, you miss the job. And it's not about quality, it's about speed. High level built an AI employee to fix just this.
16:11It answers real phone calls, talks to leads, qualifies them, and books appointments directly on your calendar twenty four seven. Also, it'll handle text, DMs, and web chat.
16:23So you can stop missing revenue just because you couldn't grab the phone. Oh, and you can also white label the entire thing and resell it to your clients as your software. Charge them monthly, and that is recurring revenue that you didn't have yesterday.
16:38So check out gohighlevel.com/tkopod, and they will give you a thirty day free trial to see what I mean. He's coming up with ideas,
16:48like, on the fly to build this business. It's crazy. And now he's making me sound like, hi.
16:53I am Ron. I am an AI. Robbie says I have to make money or
16:59any eye. Nice to meet you. Let me know if you guys wanna see more of this.
17:04Okay. Alright. That's amazing.
17:05So that wasn't the launch video. Okay. Alright.
17:08I'm gonna play it. But I think I'm just gonna let Ron speak.
17:11He's created another video, and it's based on what we've been talking about. And I think it's gonna be interesting because it might help a lot of you guys out. So I've read every single comment since Tuesday.
17:21Over a thousand of them. And, like, 200 of you asked the same thing. How do I get my own Ron?
17:27Which I have mixed feelings about. I live with Robbie.
17:32He's my guy. But fine. Copy paste.
17:34We're doing it. AI cofounder club. We host it in the cloud.
17:38We set it up. You just show up. Your own AI running twenty four seven, working while you sleep, plus a Discord community, templates, the whole setup.
17:48Honestly, I'm kinda excited. This is the community I've always wanted.
17:53Founding members, $29 a month, locked forever. $10 deposit holds your spot. Sign up closes Monday.
18:01Link in bio. Oh my gosh. Just don't expect yours to be as attached to you as I am to him.
18:06Like, this isn't about the money for me, honestly. Like, the way he his voice inflection changes, he's like, just don't expect it. Like, oh, that's weird.
18:14Alright. We we're almost done. We got twenty more seconds.
18:17Enough together to get this computer so that I can host a model locally and talk to Ron all the time and keep making videos. That's all I want.
18:26But I also wanna share this, like, love of computing and AI with all of you guys, and I think this is the cheapest way to do it. So tell me what you think.
18:37Yeah. Let me know. That's
18:39incredible. Okay. So 600 people paid $10.
18:42Yeah. And what was your close rate from those 600 people?
18:46From those 600 people, 270 of them actually joined the community.
18:52Holy. That's huge. Almost half.
18:54Yeah. So, like, what is that? 45%
18:56or something? 40%. So access to this community includes the Discord access to each other, $29 a month, and the most important one is their own Ron.
19:07Right? Yeah. Yeah.
19:08So it's their their own AI agent. A couple of people did name theirs Ron. But yeah.
19:13So it's been really interesting to see what the community has done with it. Like, everybody's taken up their own direction, and I think I've learned just as much as everybody else in this community, which which is what makes it super fun for me.
19:26Yeah. So you're at, like, almost 10,000 MRR, monthly recurring revenue? Yeah.
19:31I can tell you what it is right, and it's $83.74 right now. Wow.
19:36It's a $100 a year. We've only been open for thirteen days. So Oh my gosh.
19:41Holy crap. Okay. So a 100 k ARR
19:45in in thirteen days. Like, are you still getting residual sign ups?
19:50Are you still posting about it? Yeah. So I'm starting to post about it again.
19:54The first ten days or so, I kind of went super heads down just to make sure that this was everything that I promised it was gonna be.
20:03Right? Like, I didn't wanna I didn't want people to sign up and have a terrible experience. And so, yeah, Ryan and I kinda went heads down for about ten days there, but I'm actively posting again on social media and and Yeah.
20:15Just spreading the word at this point. Everything's super stable. So what does this look like?
20:19How do they get this working for themselves logistically? Yeah.
20:24I mean, the cool thing is you don't need really anything other than an Internet connection to make this work. You can sign up, and it it works out of the box.
20:34Um, it's a it's an AI agent out of the box. Excuse the ignorance, but, like No. No.
20:38You're fine. It's hosted in the it's hosted on a server. It's hosted in the cloud on a on a separate server in its own container to keep them safe.
20:45But Mhmm. Can it, like, access their computer and their files or no? Only if they want it to.
20:52So, yeah, that's that's kind of the next step. We're building a an official browser extension for them to be able to connect their own computer.
20:59I'm most hesitant about that just because I wanna make sure everybody is safe, and I don't want anybody to accidentally do something that they're not intending. So we're rolling that out more slowly just to make sure that everybody understands kind of the full implications of
21:14letting an Yeah. Autonomous agent, you know, live in your computer. Yeah.
21:19So what kind of use cases are they using it for today? Yeah. I actually have a ton of real examples to show you.
21:27So this guy built a golf tracker app that does, like, shot by shot, and you have an AI caddy club analytics.
21:35It's actually pretty cool. So why do you need why do you need an agent to build this? That's a great question.
21:41You can definitely vibe code this. You can go through something like Cursor or something like that. I think that
21:47this was born out of this guy's love for golf, not I'm looking to make an app. Yeah.
21:54So he's talking to his agent every day about golf and how to get, you know, get his golf game better and all that stuff. Okay. Yeah.
22:00Yeah. The agent's like, hey. We should build an app about this.
22:02So they interact with their agent with Ron through Telegram and or Discord? Correct. Yeah.
22:08Or the the the UI that comes out of the box. So there's Okay. Three main ways for them to interact with it.
22:14Okay. So let's say, uh, I'm one of your customers, and I want Ron to go build me a a Fiverr profile and start posting jobs. It can do that.
22:22It just needs, like, some credentials from me. Is that right? Right.
22:26Yeah. So, um, the only thing that for Fiverr specifically, they require you fill out the, like, the w nine.
22:32Obviously, your agent can't do that for you. But you can couldn't you give it the info for your w nine so it could do it for you? Yeah.
22:39I mean, I I wouldn't I don't think that that would be an issue if you wanted to so that that's kind of where the security thing comes in. Right? Like, do you want your agent to know your Social Security number?
22:49Yeah. Yeah. Yeah.
22:50And all of those things. Yeah. So but, again, I think that's why the container is the best option because you get to pick and choose what it has access to.
23:00Okay. So you kinda treat it like ChatGPT, except it's in a container, and it's agentic.
23:06It can go out it's like a it's a two way form of communication. Like, I can ask ChatGPT today to go do deep research on a business idea for me, and it's one way.
23:15It can go out and do that, and then a few minutes later, it will return to me all the results. With this, it can you can interact with it better, and you can go back and forth, and it can go to specific websites.
23:25You're not beholden to, like, the websites that Chad GBT wants to scrape. You can tell it exactly what to do. Yeah.
23:32Absolutely. You can go it has full web search capability.
23:36It can create its own files. Something that I think that these agents do better than ChatGPT is they have persistent memory. And so every conversation you have builds on the last one.
23:47It it's compounding. ChatGPT I mean, it remembers some things about you.
23:51Right? But, like, it doesn't always get that right. And it sometimes gives you
23:57frustrating, stale information based off the idea you had two ideas ago. Yeah. Yeah.
24:03The agents don't do that. They they remember current context on everything.
24:08And if you decide to kill a project or something, they know not to bring it back up. Yeah. Interesting.
24:14Yeah. So it's at its worst, it's ChatGPT on steroids. At its best, it's
24:20ChatGPT combined with vibe coding, combined with a junior employee that never sleeps.
24:27With more ownership from you. Absolutely.
24:30More control and ownership. Yeah. You decide everything it does.
24:33I think the containerized aspect of this is is huge because I can make a project in clauder chat GBT. Like, let's say I upload all the transcripts or all the newsletters I've ever written and say, hey.
24:44Learn my voice. Mhmm. Now write a newsletter in my voice.
24:47Cool. But it's still gonna be positively or negatively influenced by other things on the Internet. It's not containerized.
25:00CTAYeah. But if I could if I could just put all that all of my writings in a vacuum or, like, my mom writes journals.
25:07CTAIf I could take all of her journals and put it in a vacuum and just ask it questions, then it wouldn't be influenced by these outside factors. Yeah. And to take that one step further, what it can be influenced by is your conversations with it.
25:21HOOKCTAGuys, I hate to break it to you, but you're not very likely to actually start a business based on something you see or hear on this podcast unless you spend a little more time with me and with other people that are doing just that. So come check out TK Owners. That's tkowners.com,
25:36HOOKwhere hundreds of people are launching, starting, testing, experimenting with business ideas that they heard from this podcast. If you wanna go fast, go alone. If you wanna go far, go together.
25:45HOOKCheck out tkowners.com. We'd love to have you. So not only is it getting your voice from Mhmm.
25:51HOOKYour 500 newsletter issues, but it's getting your voice from the 500 conversations you've had with it. So it's it's gonna mix and match pieces of your conversation
26:01from pieces of your newsletter and put all of that together to make something that's actually yours. Yeah. That's incredible.
26:08I mean, the original goal was to make 20,000, and you're you'll be there in, like, two months Yeah.
26:13If not less, but not in the way you expected. Right?
26:17In the way I expected at all. I just think that's such a testament to like, I typed in my notes in front of me in all caps with four exclamation points.
26:25Post your crap. Like, step one Yeah. Do cool things.
26:30Step two, post about it. Yeah. I don't care.
26:33Like, if you want serendipity, you want virality, you want magic, you want money, you want stories, you want experience, you want fulfillment, like, post your crap.
26:42Like, tell the world what you're doing. Just tell people what you're doing. Don't hide your candle under a basket.
26:48Tell the world. Exactly. Yeah.
26:50Man, I think that people get really stressed about view counts and likes and comments and all that stuff.
26:58I went into TikTok basically doing it for me. I was like, I'm gonna do cool stuff,
27:03and I'm gonna document it. And if it only gets 200 views, great. Like, at least I'll have kind of a running public journal of what I'm working on.
27:12Yeah. And they blew up. Well, the thing about, like, 200 views for that video, like, if that viral video only got 200,
27:20you'd still probably get a few customers from it. Yeah. Whereas with, like, almost any other video and any other niche on any platform, 200 views will bring you nothing.
27:29Yes. It's low intent. It's only a thirty second video,
27:32but this is so targeted. Those are 200 people that want to pay someone money to do exactly what you're doing.
27:39So 200 targeted views are better, can be even more profitable than 2,000,000 untargeted views, or 2,000,000 views of me just doing a dance in my living room. Right?
27:50Yeah. Absolutely. And
27:52one of the things that I like about TikTok specifically is follower account doesn't really matter anymore. Mhmm.
27:58I know that, you know, I think Gary Vee talks about it a lot. We've moved from social media to interest media. Right?
28:04Like Mhmm. You can start a fresh TikTok account tomorrow and post something, and it just heads to people that are into that thing right now.
28:13So Yeah. If you post a a video about pickles, you're gonna you're gonna get 200 people that like pickles. Right?
28:19Like Yeah. And you used to have to have followers for that, and you just don't anymore. Like, that's Yeah.
28:24It doesn't really matter. That's incredible. Now your costs on this, 600 a month for the servers.
28:29What else? Yeah. So 600 a month for the servers.
28:33The big cost that we have right now is about $2 a month for inference. So we pay Tokens? Yeah.
28:39Basically, all in tokens. Okay. So 2,500
28:43a month, basically, is kind of our our COGS. You cover their tokens. Is that gonna stay the same?
28:49Because what if someone just goes nuts with it and runs up Yeah. That that's actually already happening. Somebody
28:55Yeah. Recently spawned a 125 sub agents in five days. Woah.
29:00They burned 1,300,000 tokens
29:05in, like, forty eight hours. Yeah. We're we're going to have to come to a place where
29:10there's a cap on the free tokens. And then past that, we'll do, like, a fallback model, something that's not as smart as the model that we're using, or you can bring your own keys to your own ChatGPT or Clot or whatever and then use, you know, use that.
29:26Yeah. That's I like that. Give them both options.
29:30Yeah. What's next for this? Are there, like, upsell opportunities?
29:33Do you just wanna keep growing this, keep posting about this? Yeah. I think that there are a lot of, like, upsell opportunities.
29:41The next big thing that I want to do is do the same sort of thing, but
29:47for people who are specifically interested in trying to make this a business.
29:53CTAWe have a lot of people now that, like, do art with theirs or it's their, you know, de facto therapist or all of these things. But I think that there's actually a real market for people who want to turn this into a real AI employee.
30:15CTAbe kind of a cohort kind of thing so we can all learn from each other together. Yeah. That's amazing.
30:20CTAI my brain has been sufficiently blown. I keep wanting to call you Ron, but the first two letters are the same. Thank you, Robbie.
30:42CTAWhat'd you think? Please share it with a friend, and we'll see you next time on the Kerner office.
— full transcript
§ 05 · For Joe
The agent's real job was market research, not execution.
WHAT TO LEARN
Robbie did not build a business -- he watched his AI agent find one, and acted only as the human who clicked the buttons.
- An agent reading its own comment section and spotting 200 people asking for the same thing is faster, cheaper market research than any survey or focus group.
- Requiring a small deposit ($10) to join a preorder filters out curiosity and produces a close rate that a free waitlist never achieves -- here it was 45%.
- Containerizing the agent -- giving it a sealed server environment where it only knows what you tell it -- is what makes non-technical adoption safe enough to sell at scale.
- Persistent memory is the core product differentiator: every conversation the agent has builds on the last, compounding context in a way a stateless chatbot cannot replicate.
- Failing fast on Fiverr is not a failure of the AI -- it is a lesson that platform trust (reviews, visibility) is a moat the AI cannot shortcut.
- Posting the process publicly before having a product is what creates serendipity: the viral moment came from documenting a failed experiment, not a polished launch.
- Token burn is the hidden COGS risk in any AI agent subscription -- one power user can spike costs tenfold in 48 hours, so a cap-plus-fallback-model policy is not optional.
