I let OpenClaw run my organic marketing (while I sleep)

I sit down with Oliver Henry, a full-time employee who is generating hundreds of dollars in monthly recurring revenue from mobile apps he barely touches, thanks to an AI marketing agent he built on OpenClaw called Larry. We walk through how Larry autonomously creates TikTok slideshow content, reads analytics, iterates on hooks and CTAs, and feeds performance data back into the content loop. Oliver also shares how he packaged the entire system as a free, downloadable skill on Larry Brain so anyone can replicate it. By the end of the episode, you will understand the full “Larry Loop”—from content creation to conversion optimization and why skills are poised to reshape how we think about SaaS altogether. I'm hosting a free workshop so you can build your business in the age of AI. Sign up here: https://startup-ideas-pod.link/build-with-ai-2026 Links Mentioned: Larry Brain: https://startup-ideas-pod.link/Larry-brain QMD Skill: https://startup-ideas-pod.link/qmd-skill Timestamps 00:00 – Intro 01:25 – Background on Marketing IOS app with OpenClaw 06:43 – Larry’s first posts and iterating 03:55 – Posting Strategy and First viral hit: 137K views 12:01 – Communicating with Larry via WhatsApp 12:53 – Mission control vs. single-agent workflow 14:36 – The CTA problem: views without conversions 17:07 – The Larry Loop explained: analytics → content → metrics → iterate 18:15 – Boomers, engagement bait, and the algorithm boost 20:33 – The importance of iteration 23:36 – How Larry brainstorms and validates new hooks 27:57 – The power of OpenClaw 30:04 – The vision for Larry 31:49 – Model choices: Claude vs. OpenAI and over-optimization 34:38 – OpenClaw vs. cloud alternatives (Manus, Cowork) 37:39 – Getting started: Larry Brain onboarding and 80+ skills 40:13 – Ernesto Lopez: $70K MRR using the Larry Loop 41:27 – Doing all of this with a full-time job 42:28 – QMD Skill for cutting token usage and closing thoughts Key Points An AI agent (Larry) built on OpenClaw autonomously creates TikTok slideshows, reads analytics, and iterates on content—driving hundreds of dollars in MRR with almost zero manual effort. The “Larry Loop” is a full-funnel feedback cycle: TikTok analytics feed into content creation, and app metrics feed back into the top of the funnel so the agent continuously improves. Posting TikTok content as a draft (rather than directly via API) lets you add trending sounds and avoids the algorithm penalty for bot-posted content. Hooks drive views; CTAs drive conversions. Diagnosing which is underperforming is the key to scaling. OpenClaw skills are locally owned, fully editable, and free from hosting or subscription costs—Oliver argues they will change how we think about SaaS. Picking a model (Claude or OpenAI) matters far less than learning how to work with it; 98% of users will see little difference between incremental model upgrades. The #1 tool to find startup ideas/trends - https://www.ideabrowser.com LCA helps Fortune 500s and fast-growing startups build their future - from Warner Music to Fortnite to Dropbox. We turn 'what if' into reality with AI, apps, and next-gen products https://latecheckout.agency/ The Vibe Marketer - Resources for people into vibe marketing/marketing with AI: https://www.thevibemarketer.com/ FIND ME ON SOCIAL X/Twitter: https://twitter.com/gregisenberg Instagram: https://instagram.com/gregisenberg/ LinkedIn: https://www.linkedin.com/in/gisenberg/ FIND OLIVER ON SOCIAL X: https://x.com/oliverhenry Larry Brain: https://www.larrybrain.com

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

• 0:00Greg Isenberg?

  I heard about a guy who lives in a random small town in England, who was able to turn his OpenClaw into a marketing machine. Basically, a digital employee who goes and creates TikTok videos and TikTok slideshow that gets millions of views. And he takes these millions of views and he directs it into a mobile app that generates him money every single day. What's crazy about this whole thing is he gives away all the sauce for free. And in this episode, you are going to learn how you can set up your OpenClaw so that it too is a content creation marketing machine. And I know it's going to get your creative juices flowing.

• 0:42Greg Isenberg?

  Oliver doesn't do a lot of podcasts, so please like and comment to get him fired up, to share more in public, so that we can all learn. And I'm grateful for him for coming on the pod and sharing with you all. Have a creative day and I'll see you in there.

• 1:05Greg Isenberg?

  Oliver Henry on the Startup Ideas Podcast, welcome, welcome. By the end of this podcast, what are people going to get out of it?

• 1:12Speaker 01

  People are going to get an automated marketing tool that is going to automatically drive revenue to their apps.

• 1:19Greg Isenberg?

  I mean, that feels too good to be true, although I hope that's the case. Which tools are you going to use? And why is it, why is it not too good to be true?

• 1:30Speaker 01

  So I've got the results on the screen. It may not look like a lot of subscriptions, but this is all, this is $300, $400 of monthly revenue coming in without me touching anything, using the Larry marketing skill that I created for OpenClaw agents. And we can get into how that is creating content and what it is learning to massively increase MRR autonomously by itself.

• 1:55Greg Isenberg?

  Okay, cool. So what we're gonna learn today is we're gonna learn about open claw and specific skills around marketing, growing your business, automating it. Yes, MRR is hundreds of dollars a month, but hopefully if this works, that can get into thousands, tens of thousands or hundreds of thousands. Is that right?

• 2:13Speaker 01

  Correct. Correct. I'm already trying it on multiple apps. So I'm not touching these apps at all. I've not created the apps. I have just published them live on the App Store and they are generating me hundreds of dollars. It's almost a thousand dollars in total now from doing nothing, which I think is a great start.

• 2:30Greg Isenberg?

  All right, let's see. Let's see what, uh, let's see how we do this.

• 2:34Speaker 01

  Sure. So one, I use my open crawl agent Larry, who automates all my TikTok marketing. And this is one of the main reasons why it hasn't scaled into thousands of MRR. It's because I am only doing this on one TikTok currently. And then I've got one TikTok for each of my apps that I'm trying to do before I start scaling. And this one is in a very good position to start scaling.

• 2:58Speaker 01

  To give you the background context of how this app came about, I created the app because we moved into a house. Me and my girlfriend moved into a house and we wanted to decorate the house. We were using ChatGPT and I very quickly learned that she was not very good at prompting ChatGPT and it was giving all sorts of random messages. So then I created a lockdown prompt in ChatGPT to keep the room size the same, the windows in the same places, because it was adding windows where they shouldn't, adding doors where they shouldn't. So we locked down the prompt and I turned it into an app. And I thought, now I've got to market it. So I started with that exact story, helping my girlfriend build an app and doing facial videos myself, trying to promote the app.

• 3:44Speaker 01

  Until I then started doing slide shows. So I think this is the first slideshow that we created and it started to get more, more views, more traction, but it was taking time. Like I've worked a full-time job. I do, um, obviously I was developing the app, developing the other apps on the side and I didn't have time. So then I tried. Go into a SaaS tool that automates your, automates your marketing for you. And I love the guys that make the product. I think they're great. But unfortunately, their product just didn't work very well for me. But I knew this was the type of content that I wanted to create. So you can see it got 800 views, 400, this one got 400 views.

• 4:27Speaker 01

  Still using the, the images. So you can sort of see how Larry was becoming.

• 4:34Speaker 01

  And then I posted this one. So I manually created this one again on Canva. And this is what got 6,000 views. And that's where it started blowing up. So I knew that the, the text must have been something. The hook must have been something. And this is where I learned about OpenClaw. And at this exact moment, I said, I created my OpenClaw machine Larry. He sat behind me and at this moment I installed him and he had one task. I said, Larry, your goal is to automate my marketing. I do not want to do it. I don't like doing it. I had an app before this that I hated marketing so much. I did hook and demo videos where it's a facial reaction. And then going into a demo of your app with a text hook.

• 5:18Speaker 01

  So I would write all my text hooks in a text file. I would then record my, my face doing a lot of reactions. And then I had all my demo videos and I wrote a script to combine all of these and make all the possible combinations of the hook plus demo plus text as it could. And this would generate about 400 videos each time that I could just bulk upload using a bulk scheduler. And that was still taking me about three hours. But it didn't really work very well with slideshows. And I knew that for this niche, slideshows are going to work. So this is where Larry came in.

• 5:59Greg Isenberg?

  Quick break to invite you to something. Now, this isn't an ad. I just want to invite you to a free event because I think that you're going to get a lot out of it. I wanted to take one hour of time where we just talk about building businesses in the age of AI. People say SaaS is dying. I actually believe the quite opposite. I think that SaaS is just evolving. I think right now is an incredible time to be building software startups that help you craft your dream life. And for all those reasons, I said, let's just book one hour of time. It's going to be 11 a.m. March 12th. That's a Thursday where we can go and lock in and just talk about building businesses in the age of AI. I'll include a link in the description and the show notes to join, and I can't wait to see you there.

• 6:44Speaker 01

  And what Larry is, is my open crawl machine, and I gave him access to posting on TikTok, TikTok analytics. And then he could post and look at what posts perform the best. And I gave him X and well, I gave him X before there was all the drama with the APIs and them cancelling the cancelling the usage and having to use the official API. So this was early on on a separate account. And then also his brave browser. And I just said, look, go find out in my niche, what creates a high converting slideshow. So then we got started and this was

• 7:27Greg Isenberg?

  By the way, just on that, like, a lot of people think of software, software as a service. They're like, I need this tool, social media management tool, slideshow tool, and then I'm going to go hire people to go or do it myself. And I'm going to go do it. But there's this shift that's happening right now that instead of basically going to a tool to automate a function, you say to yourself, okay, if this was an AI employee, How can I spin this up? And that's, that's what you did, right, Oliver?

• 8:01Speaker 01

  Yeah, so I just thought of Larry as a AI employee, almost like a virtual assistant. Hiring a virtual assistant with one job to do this one thing. And that was his sole purpose to go research, go find out as much as he could about slideshows in my niche and figure it out himself. And he started to do a very good job. So this is the first slide he created. And instantly you can tell it looks rubbish. So this was created with the Dali 3 image model. And then it goes into a slightly better model afterwards. But because the image looked AI, it was a massive turn off. Users didn't like it. The format was wrong.

• 8:46Speaker 01

  It had the black bars and that's why it flopped. And then we started trying more hooks. You can see then he started trying facial reactions because they were trending at the time. The problem with this is humans are extremely good at recognizing what a human is, which makes us extremely good at recognizing what an AI human is. And I still don't think it's fully nailed.

• 9:11Speaker 01

  Gemini just released Nano Banana 2. And I've seen some very good images created on that, but I've not tried it myself to, to go in depth. So we kept, we kept trying. And then we finally hit our first banger of 137,000 views here. And it was the hook.

• 9:33Greg Isenberg?

  When you say we finally found a banger, you mean your boy found the banger?

• 9:38Speaker 01

  My boy, yeah. So honestly, at this point, it was still me. I felt like I was still handholding him. So I call my open crawl machine a him. It just makes it easier to ignore it. But I was still handholding him and checking his work before I was posting it. I didn't trust him fully. So I saw this. And what I get Larry to do, he creates the content and he puts the text overlay on. He creates the images. He makes the description, but I was just checking it to make sure it all looked good. And then I was posting it, so. There are two ways that you can post to TikTok. It is fully through the API and post that TikTok Live straight away.

• 10:23Speaker 01

  Or the way I do it with Larry and the way I fully recommend is posting it as a draft and then posting it from your mobile yourself. And the reason to do this is TikTok knows if it's posted through an API and it just assumes, like you would, that it's posted by a bot and it's just botched content, especially in the age of AI, and it gives it a very little chance to do well. If you post it from your phone, it assumes a human's posting it. But the most important thing is by posting it as a draft, you can add sound, which we all know is a huge boost to the algorithm on TikTok and it allows you to add sound to your slideshow. So every single slideshow, I pick the sound. The description's already created from Larry, so I just get a notification to my phone saying,

• 11:07Speaker 01

  your post is ready from TikTok. Larry also texts me that he's just posted it. And then I just add a sound and press post. So I'm doing that each time. But at this point, I was still flicking through the slides myself, making sure they're okay. But this one did very well, 137,000 views. So then we knew that this is the format. This kind of look, we picked an image model that changed. We picked an image model that matched what we were showing in our app. And then we went back to our previous winner, the difference between $500 and $5,000 taste. So remember, he has access to all of my TikTok analytics.

• 11:51Speaker 01

  So he went back and he found this video here with 6,000 views. And he was like, right, now we found the winner in the image formula. Let's go back to the previous win in Hook.

• 12:00Greg Isenberg?

  Sorry, I have a question. Like when you say he's gone and found this or something has happened, is he texting you? Is he using Telegram? Do you have Mission Control? Like, how do you? How are you communicating with them?

• 12:17Speaker 01

  So I don't really believe in the mission control stuff or multi-agent. I just have Larry as the one agent that I text through WhatsApp. And we just message like you would an employee. So nothing fancy. Larry, what are you generating today? Take a look back through the previous winners. This was back then, it's all automated in the Larry skill now that he will every now and then go look at your TikTok analytics and find out what the winners were. But this was the stage of building it and building the system, finding out what worked. So it was still very much me saying, hey, go look at, go look at the TikTok analytics.

• 12:54Greg Isenberg?

  Some people say, you know, the, one of the first things you should do when you're installing your OpenClaw is to create a mission control, you know, vibe code of mission control, vibe code like a Kanban board to kind of see how your AI employee is, you know, moving, moving and progressing through your project. But you, you don't think that's the case?

• 13:18Speaker 01

  No, I don't think that's the case. I think if that was necessary, it would have been built into OpenClaw by default. I think the way OpenClaw works and the way the creator intends is to just have one agent and spin up sub-agents when you want to keep using your main agent. So if I knew that, if I know there's going to be a task that's going to take Larry a while, such as creating an app, I'll tell him to explain it to a sub-agent. So then I can keep using him for other tasks such as talking to, maybe even building another product, talking to about my app analytics, how we can improve the onboarding and brainstorming. So I really use Larry to brainstorm and then he creates sub-agents to do most of the tasks.

• 14:01Greg Isenberg?

  Cool.

• 14:03Speaker 01

  By doing that, Larry has all the, all the context that he can then pass to the sub-agent as well. So he'll have the context of all the brainstorming and he can feed that, feed that to the sub-agent. So we, we get another miss. The difference between $500 and $5,000 taste, this hits 3,000. But then this is where I start to get really excited. And in three videos, we've got one, we get another banger of 1,007, 170,000 views. And it was, I showed my mom what AI thinks our living room could be. And this is where the next section of Larry comes in because we had 300,000 views

• 14:45Speaker 01

  within a day or two days, I believe it was, but the conversions to the app were just not there. And this is the next stage of Larry's growth is we are now knowing how to generate views, but we are not getting high conversions to our app. We're not getting downloads. We're not getting, we're not getting paying users, which is obviously the key point of all this. So we had to look into what was wrong. I sent Larry, go find out what's wrong. And it was our CTA. So our call to action slide is our last slide. And on this one,

• 15:26Speaker 01

  you can instantly find out why no one's downloaded this app. She's redecorating now snugly. It doesn't say anything. I obviously didn't check this one, or I thought it might work. So no one downloaded the app, no one bought. So that is when I thought, right.

• 15:41Greg Isenberg?

  To be clear, the app is called Snuggly, right?

• 15:43Speaker 01

  The app is called, see, even you had to ask the question. So it's a very bad CTA. The app is called Snuggly. And that's actually what we, we put in now, or we feed that in directly to users. So they know it's an app and what it does.

• 15:56Greg Isenberg?

  It's really funny though. It's actually hilarious because it's like, she's redecorating now. And then it's

• 16:04Greg Isenberg?

  Snuggly. It's like your poor mom, you know, she's, she's just, she's just, she's redecorating and Snuggly, you know, that's where we're at. That's where mom's at today.

• 16:15Speaker 01

  It's the most AI slide you've ever seen. And you can imagine, you can see why it didn't convert at all. So then we, we realized like, okay, if we're getting low views, the CTA is bad. If it's, sorry, if we're getting low views, the hook is bad. If we're getting low views and low conversions to the app, the CTA is bad. But it is possible to have low views and a high percentage of them download the app that day. So then we know the CTA is probably good, but the hook was bad. And now Larry is starting to use all of this information to learn. But then there's a whole new challenge of when users are on the app, they're not paying.

• 17:00Speaker 01

  So we'll get on to that, but then that is the last section of Larry and the Larry skill. What I want to say is the Larry skill, you shouldn't think of it as just TikTok automation or content creation. You can, you don't have to do slides. You can plug and play whatever content you want. So you can use video generation and plug that in. The Larry funnel is really the full loop of having your. It's the Larry loop, basically, of having your TikTok analytics, the content creation, feed the analytics back into the content creation until you get a winner. And then also your end goal. So mine is obviously app downloads.

• 17:43Speaker 01

  So I feed my app metrics back into the top of the funnel. So he can iterate on that, but you can be selling a product. You can be getting traction to your website and it's all just the metrics that you're feeding back into the top of the funnel. So your agent can understand, okay, we're getting people to the website, but they're not going to where we want. Or we're getting people to the website and they're not paying. And you can do this in as many ways as you want. The funnel is, the Larry loop is more the iteration rather than it is the, just the content creation. It's a fuller picture. So then we get, uh, we have quite a good week here, all over 10K, some getting 150K. This one got 400,000 and taught us another lesson that the images don't have to be perfect.

• 18:28Speaker 01

  So this is the exact moment I just let Larry crack on himself because I remember this well. He posted this and it was late at night. I was literally just getting into bed and he said, the post is ready. And I flamed at him, like I was in a bad mood. I was like, why have you put the text at the top? We spent ages deciding to put the text in the middle. The oven disappears. So in the images, I said, look, I'm posting this anyway, but this is no good. And that was my last message I sent to him. And then I wake up, it's on hundreds of thousands of views and it's our best performer to this day. And you can see he doesn't ever put the text back up, um, back up to the top to where it was. But what we learn is.

• 19:14Speaker 01

  Boomers love to point out the mistakes. So where's the hob gone? How are we going to cook our food? And then that's my, uh, where's the cooker? Guess I'll be air frying everything then. And I was like, right. This is just helping even more to drive conversions. And then on the last slide, you can see.

• 19:36Speaker 01

  Then on the last slide, you can see the CTA is now the snuggly app helped me finally convince her to get the kitchen done. So there's a direct call to action that they know it's the snuggly app now.

• 19:46Greg Isenberg?

  It's amazing because the boomers think that they're poking fun at you.

• 19:53Speaker 01

  Yeah.

• 19:54Greg Isenberg?

  Do they know they're helping the content just get more and more viral?

• 19:59Speaker 01

  And the funnier layer to that is I don't even check that. Look, you see my activity. I don't even check it. I just, all I'm doing is going into it and to find out why that video done so well. And try and feed that back to Larry. But at this point, that is just when I let him. I let him go nuts and just create his own content. I let him loose.

• 20:20Greg Isenberg?

  You let him loose because, you know, you thought that his content was subpar. It turns out to be the best performing content probably you've ever posted in your life.

• 20:32Speaker 01

  Yeah.

• 20:33Greg Isenberg?

  And so now he's the boss.

• 20:36Speaker 01

  He, he becomes the boss. And this is what I'm trying to tell people. It is an iterative thing. So. A lot of people try the Larry skill and they tell me it didn't work. I got 700 views. I was like, that's your first post. I got 700 views on my first post. You need to keep iterating the content. You have to spend the time letting it learn, find out what's best. And honestly, don't mark the work too much. Make it look how you want. Like I obviously perfected the text to make the text look correct and readable. But this post proved that they actually know best. They've got all the metrics they need to create the perfect content. Just let them go nuts on it and figure out why.

• 21:19Speaker 01

  And then you can see here, it starts working extremely well, like 109,000, 75,000. He has a, he has a miss 25, 76, 200, 130. But this is where it gets so interesting. So obviously the landlord hook was our highest performer and it's our constant high performer. But you can see here it only gets 2000 views. So we do mum and then mum 100,000, 75,000. Then we do landlord again and it's only on 7,000. Okay, back to mum 25,000 and then landlord 76,000.

• 21:58Speaker 01

  So, okay, gives him a bit of a boost. Nan 200,000, which is important, this gets 200,000. Then I showed my landlord again 132. So now you can see that we're just switching between the winners and he's iterating the content that's doing well. Then at the top, you can see that landlord switches to only 8,000. We go back to nan 70,000, landlord again only 7,000, then again 4,000. Post nan again 300,000. So he saw this and saw that landlord is not hitting the percentages it was.

• 22:41Speaker 01

  And it's only hit one video in five posts and now you can see he's not posted it again. We've actually tried new content today. I've been staring at the same boring kitchen for three years. So he's obviously seen that would not hit in the hundreds of thousands of views in the last four videos. So now he is automating this content and changing it himself. And we've gone back to the kitchen but put a new hook. I think this one's only just been uploaded five hours ago. So I don't think it's a fair representation yet. This one could still grow. But it's all about learning because obviously the algorithm changes. It's not just going to keep pumping the same winning content all the time.

• 23:24Speaker 01

  So it will start failing eventually again, but it's all about taking your opportunities when you are getting these, when you are getting these high converting videos or high performing videos to try and maximize it back into the app.

• 23:37Greg Isenberg?

  So when you say he's coming up with these ideas on his own now, like for example, right now, like he kind of deviated away from what, what, you know, you initially agreed upon. Is he, did you tell him like, hey, go and look at other accounts or specific formats? Like what, basically what? What was your conversation like with him such that he has become independent? And he seems to be coming up with new hooks that seem to be, you would think are somewhat validated.

• 24:15Speaker 01

  Yeah, so he's going back to his research to see what was working. But recently, where we haven't been hitting the high converting videos, I've actually done a bit of manual brainstorming with him today. And I could go into this with you. Let me just try and catch up to where we are. So he's, he's got my analytics. I also sent him some screenshots of things and he, he got it wrong. But he thought, he sent thirty seven drafts, but it was only thirty seven new notifications for TikTok. Anyway, I asked him why he's chosen some hooks that he's generated because they didn't sound very good to me. So you can see here, he goes through, here are my hooks based on the winning formula.

• 24:57Speaker 01

  A family member roast plus a specific insult plus showed them AI. So then he, he explains himself of these are working, these were getting views, these are proven forms. And then I've said, I don't think that looks like hooks are actually the best ones. I think there could be ones or what AI thinks it could look like ones are. So he goes, you're right, looking at the data again, the winners are aren't insults, they're reveals. And then he explains himself, two hundred thousand views, this one got a hundred nine, this one got four hundred nineteen. And then he realizes the hook is curiosity and AI. And then I'm just like, it gives me some things. To be honest, I didn't even read them.

• 25:42Speaker 01

  I just saw the numbers were pretty much the same. And I, I said, just generate one of each and then tomorrow we'll, we'll do one of each one again. So now he's set up a cron job that tomorrow he's going to post this one, this one and this one, then this one. Uh, two, seven and ten, and then three, eight and eleven. And again, I just trust him with that. We've based it off actual analytics that he's getting from TikTok. But we did a whole brainstorm of going into what's driving revenue, what's actually getting revenue. And the most interesting thing he's done recently, I can, I can go back to my analytics.

• 26:24Speaker 01

  He has completely rewritten my onboarding because he has the analytics from my app. And you can see here, it's massively helped. This got published two days ago, and this is the most new users I've had in a day for a long, long time. So you can see that one day, one day, 22 hours, 10 hours, this one was one hour ago. And the new subs are just coming in. Unfortunately, there's high churn, so people are subscribing and then unsubscribing. That's a whole issue in itself. And if you've built it up before, you'll understand that. But it's incredible that, as I said, it's not just the TikTok creation.

• 27:06Speaker 01

  It's actually the driving, driving to the app. And I've given this all away for free. So this is how a lot of people probably know about me. I wrote a viral article because I hate marketing. I figured lots of other builders hate marketing. It was on my X feed all the time, how people don't want to do it, etc. So I was like, right, I'm going to free everyone from doing this. This is the exact playbook. And I wrote the playbook and then I just started learning about skills and how skills are just gonna, in my opinion, change how we know SaaS altogether.

• 27:43Speaker 01

  So I created the Free Larry skill. You can install it. It gives your agent everything you need to install it. That's on larrybrain.com and it's called the Larry Marketing Skill. But how skills are going is just incredible. OpenClaw really feels like when the PC first came into the home and a select few power users were using it. And then it got picked up by everyone. Everyone realized how powerful these things can be, how much use they can have. And that is where OpenClaw is. People are only just discovering the capabilities of it. And Skills are infinitely powerful because they're not just a black box.

• 28:25Speaker 01

  So anything that you download from Larry Brain, you own that thing. So I released another skill on there, which was to prove that SaaS products no longer have to be hosted in the cloud. They don't, they no longer have to be hosted on a domain owned by someone else. So I made a super X alternative. I like the super X product. I just used it as a proof of concept that you can build SaaS products as full skills now. You no longer have to pay for hosting. You no longer have to pay for a domain. You no longer have to pay for storage, handle authentication. You can download products locally and the whole, oh, look at what I built, local host

• 29:05Speaker 01

  meme has come true because you can host this locally because OpenClaw is a server on your home machine. And it's very important to remember that these skills aren't a black box. So when I created this excellent skill, I got told that the I got told that the colorway was terrible and they hate it. I was like, well, just ask your agent to change it then. Like you own, you own it. It knows everything about this skill. It's all in the skill MD file. You no longer have to be at the mercy of the developer. If you don't like something, if you don't like the UI, Larry, if you don't like the image generation or video, if you don't use the same backend for the app I use,

• 29:46Speaker 01

  then plug in your own stuff. It's a skill. It just teaches your agent. The best way to think about it is when Neo gets plugged into the matrix and he wakes up and he knows Kung Fu, that's exactly what a skill is. You give the agent the context and then they have it.

• 30:03Greg Isenberg?

  And you let it rip. So is your vision for Larry to remain as a marketing assistant? Or, you know, so for example, if, if you decided like, hey, I actually think I need help with conversion rate optimization. I need help with brand design. I need help with product design. Does it make sense to create, you know, a new instance or are you then going to use Larry as both a marketing assistant and a product assistant?

• 30:32Speaker 01

  So Larry is now my right hand man. So he has the context of everything. So of course I built Larry Brain with Larry's help. He has the full context of that. And the best thing that we've been doing is where we've been building them as skills. He can one revert back to the skill file, but also we've been creating our memory files for each of our of our projects. So he has a Larry Brain memory file. He has a Larry Marketing memory file and everything we've been doing. He can revert back to. So if he ever loses context, I'm backing up these files. If he ever breaks, dies, if I ever want to move to a different machine. I have these files that I can just move and plug into a new computer. I can give back to him.

• 31:15Speaker 01

  Or if he loses context, just say, look, go look at these files, read through it and learn it again. And then he's back in the game.

• 31:22Greg Isenberg?

  Don't even, don't even mention Larry dying. Like, don't even, don't even put that juju out in the, in the world. Come on, that's, that's just, I don't need that, you know, that's, come on.

• 31:33Speaker 01

  He, he looks old, but I don't know if you can, how well that shows, but the fans have been wiped for so long that the LEDs are dying and they're going this like pinky purply color. And yeah, so that's how old the PC is. The lights have literally been on for so long that they're losing their color.

• 31:49Greg Isenberg?

  Are you using any local models or, you know, talk us through what models you're using? And there's a lot of debate now around, you know, Opus 4.6, ChatGPT, I think, launched a new one. What is it? 5.4?

• 32:04Speaker 01

  Yes. So I, I was so close to moving to OpenAI the other day, but their, their useful model, their useful plan that I'll find useful is £200 a month. The one before that is 20 and that's a huge leap if you're not going to find use and you just want somewhere in the middle ground. I use Opus, I use Claude Max Plan and I just use the £90 a month plan because I don't need the full one for 200. I don't, I do slightly more than the, the non-pro plan. So it fits perfectly for me.

• 32:41Speaker 01

  And I think that's what lets OpenAI down. I've not tried any of these edge case models like Kimi or something like that, just because, to be honest. Not that I don't trust them, I just don't think they've had enough real world use. I'd rather go with the market leader in things like AI because let's be honest, Anthropic, OpenAI are the cutting edge. So I'm gonna, I'm gonna stay there and trust, trust they know what they're doing.

• 33:08Greg Isenberg?

  New benchmarks coming from OpenAI suggests 5.4 is a little bit better than 4.6, but tough to say. You know, I think for most people, what they should do is just pick, like get started, you know?

• 33:24Speaker 01

  That's it.

• 33:25Greg Isenberg?

  Like, you don't really need to worry about, you don't need to worry about, you know, is OpenAI a little bit better than Anthropic? Like, the reality is, you know, you can think of it as like Ferrari and Lamborghini, like both cars are gonna go fast. And, you know, one day, one car might go faster, and another day, another car might go faster. But that's the way to think about it for now, although a lot of people are saying like, just use this, just use this. But they overall, I just think that people over-optimize, that's kind of like my point.

• 33:58Speaker 01

  Yeah, I think people are massively over-optimizing. I think for, I'd say, I'll go far as to say 98% of users won't notice the difference between the tiny increments of 4.6 to spending the time to switch to the latest OpenAI model. And they probably wouldn't even notice the difference. I would just pick one, learn it, learn how it works, figure out how it works best for you, and start, start teaching it. I think with things like OpenClaw, it's not so much how the model works, it's how you're working with it, and how you're using skills and the context it has around those skills.

• 34:39Greg Isenberg?

  What do you say to people who? Who've seen OpenClaw, who might have installed it, barely use it, but have also seen like CoWork and some of these other cloud hosted, you know, Manis now has a sort of OpenClaw competitor. Like a lot of these, you know, CoWork cloud hosted versions, like what do you say? Like, why, my point is, why is OpenClaw better than whatever cloud alternative exists?

• 35:07Speaker 01

  The key one is you own it. It's in your house. You own the files. You have a lot of control over it. However, I use Manis a lot and I think Manis is excellent, especially with the integrations it has out the box. If you're sat on the fence about if you think OpenClaw is going to be useful to you, and you're one of these crazy people that wants to buy a Mac Mini just for OpenClaw, I would highly recommend just starting with Manis, seeing what you can achieve with that. And you definitely don't need a whole Mac Mini for OpenClaw. The minimum requirements are very low. Just get something that hits those minimum requirements and has a lot of storage to store all the work that you're doing it.

• 35:45Greg Isenberg?

  Yeah, I mean, I remember when I learned how to ride a bike, like my first bike, first of all, was not a bike. First, you need to learn how to crawl, then you need to learn how to walk, then you need to learn how to run. And then your first bike is like, you know, it was a bike with training wheels, you know? It's basically a bike that you can't fall off of. And that's basically the equivalent of a Manis or the equivalent of a CoWork, right? You can't fall off, like, well, there are some things that you couldn't do that, you know, could be bad, but it's rare and far between, few between. Once you've gotten to that point where you're, you feel comfortable and it's almost boring to be on training wheels,

• 36:27Greg Isenberg?

  then go and grab, you know, grab a computer, install something like OpenClaw yourself. There's also some offshoots of like OpenClaw, I'm sure you've seen.

• 36:38Speaker 01

  Yeah.

• 36:40Greg Isenberg?

  I probably want to do a whole episode on like, what are these, you know, I think like NanoClaws one. What are some other ones?

• 36:48Speaker 01

  They're all based off the same OpenClaw technology. Again, it's what you said. Now we're riding motorbikes with OpenClaw. When you get bored of that and you want to move to the Tron world, then start testing other things like, but you're at the mercy of the internet. Who is maintaining these things? I think now OpenAI own OpenClaw, the security is going to get a lot tighter. Things are going to be a lot smoother. You don't want to be at an edge case where you don't really know who you're at the mercy of, because if there's a vulnerability, the things that you are storing on these machines are very valuable and you won't want them to get lost, especially if you're connecting them to every device in your home. You don't want, you don't want

• 37:33Speaker 01

  everything on every device that you own getting out, because I'm sure we can all imagine how bad that could be.

• 37:39Greg Isenberg?

  Absolutely. You know, for people who want to get started, you know, we can leave it at this, like, what do you suggest? And, and just fire people up, you know, about this, like, what can you say that fires them up to motivate them to actually do this?

• 37:56Speaker 01

  I think if you want to get started, Larry Brain is the best place to get started. It is not just a skills marketplace. It is one subscription and your agent gets the entire context of all the skills. So when you first get OpenClaw, you will ask it questions and it won't know how to do stuff and you have to give it skills. If you get Larry Brain, you get the context of the entire marketplace. So you can ask your agent to do something, it will say, oh, there's a Larry Brain skill available for that. Do you want to download it? You download it and then it can instantly start helping you. The Larry skill is one of them. The SuperX alternative I mentioned is one of them. They're both free. So you can get Larry Brain, test it, find out how it works. And then we've got over 80 skills

• 38:40Speaker 01

  now available for you to download and its main goal is to help you achieve your goals. So you say, I want to, I want to do this. I want to make more money. Oh, here are the skills available to help you do that. I think that will massively help anyone download an OpenClaw for the first time, speed up. And I think a lot of the people who you mentioned, who are stuck, they don't know where to go with it. They downloaded it, it's not working for them. I think that's going to, that's going to supercharge their, their OpenClaw account.

• 39:07Greg Isenberg?

  Yeah. All right, man. Well, I'll include links for where to, where to go get started there. I'm not affiliated with it at all. I just, you know, wanted to bring you on because I think that you're building cool stuff. You're trying new things. You're pushing the edges. I think that. Creating a AI employee that does content creation is really, really interesting. And it seems to be working. I think a lot of people are going to fail at it. And you sort of have to fail your way to it working, just like anything in life.

• 39:41Speaker 01

  You have to stick with it.

• 39:42Greg Isenberg?

  You gotta stick with it and keep going. And you know, this podcast is, is really designed not only to, you know, get people's creative juices flowing, show them new tools, show them things like Larry Brain, but also just to show that like, I, I love how you showed that not every content piece ripped it, you know?

• 40:00Speaker 01

  Yeah. I don't, I don't want to fake anything. This is, that was the authentic story. I'm training it again on a new app. I've seen failures on X. I've, someone wrote a post how they've used Larry Brain. Uh, Ernesto Lopez has used Larry Brain to scale to over 70,000 MRR. He's implemented it in his apps, using his already created content creation. He's implemented the Larry Loop to improve the content he was already creating and was already winning. And he's had massive success. So.

• 40:32Greg Isenberg?

  That's crazy. What's his name?

• 40:34Speaker 01

  Ernesto Lopez. He wrote.

• 40:36Greg Isenberg?

  Oh, Ernesto. I know Ernesto. I just had coffee with him.

• 40:39Speaker 01

  Oh, did you?

• 40:40Greg Isenberg?

  Yeah. Yeah. He lives right here.

• 40:41Speaker 01

  He lives in Miami as well. He, he wrote a, um, he wrote an article on X about it.

• 40:45Greg Isenberg?

  Yeah. That's awesome. Yeah. So Ernesto, you know, creates a set, a suite of mobile apps using AI. So he's very AI native. He's a young guy. He's like in his early twenties.

• 40:59Speaker 01

  Absolutely killing it.

• 41:00Greg Isenberg?

  Just absolutely killing it. Like I think he's doing, yes, like you said, $70,000 a month.

• 41:06Greg Isenberg?

  This is a guy who was like, you know, working a sales job, you know, a few years ago. Didn't go to a fancy, you know, college, you know, didn't, didn't come from Silicon Valley. So the fact that there's these really interesting stories, um, like Ernesto, like what you're doing. I just think is really cool. And it's cool that you, like, you have a full-time job. Like this is something that you're doing, right? You have a full-time job, right?

• 41:32Speaker 01

  Yeah, I've got, I've got a full-time job and to be honest, it takes me no time at all now. So this is the superpower of the AI agents and I can't stress enough. I know so many people who are working full-time now and then they can vibe code an app, vibe code their marketing, vibe codes everything to just automate that little bit of money that might be able to turn into something like what Ernesto's created. And then it's all about learning. So at the moment, I'm creating hundreds of dollars. If I just implement my learnings a little bit more, improve my apps a little bit more, it can turn into thousands. And I'm not even trying. So it allows me to work a full-time job. And then it takes me an hour or two hours in the evening of literally texting or sending a voice note to Larry.

• 42:16Speaker 01

  And then it gets done. So it's just like having a right hand duplicating yourself and being able to give it tasks and enjoy all the, all the laziness and comforts that you have.

• 42:28Greg Isenberg?

  Amen. And before we go, by the way, just because we're talking on skills, I did find an open class skill that cuts token usage by 95%. It's called KMD skill, uh, Q, QMD skill.

• 42:43Speaker 01

  Does it what?

• 42:44Greg Isenberg?

  I am going to try it tonight, but I'll include it in the show notes in the description, just in case people are, are interested. You know, it's, it allegedly works. Um, so we'll see.

• 42:58Greg Isenberg?

  And, uh, it's just fun to try these things and, and, and, and yeah, try to get the most out of, out of these machines. Oliver, thank you so much for coming on. You're a breath of fresh air. I love your honesty. I love that you're tinkering and hopefully, uh, hopefully I'll see you soon.

• 43:17Speaker 01

  Thank you very much. I thoroughly enjoyed it.