// transcript — 886 segments
0:00 ClawdBot
0:01 In today's video, we're going to talk about Clawbot, recently renamed to
0:07 OpenClaw. It's the AI agent taking over the entire internet. Basically, I'll
0:11 tell you a little bit about how I launched the business back in 201620
0:17 2017 that did over a million dollars in revenue in its first full year of
0:22 operation. and me being the sole employee. And the reason we're going to
0:26 talk about that is because just in the last 24 hours, maybe 48 hours, I don't
0:30 recall, I've been pretty much building non-stop with Cloudbot reality time, the
0:34 day and night cycle, it's all a bit of a blur really. But the point is that these
0:40 AI agents that I've created slashconjured out of thin air, thanks to
0:44 of course OpenClaw and that great community and the developers behind it,
0:49 but those AI agents believe that they can build a business out of this
0:53 business idea that they came up with that will generate $10,000 per month
0:59 revenue and it can be built 90 to 95% autonomously. Now, either this is real
1:06 and they can do it, which is kind of crazy to think about, or I'm going
1:11 crazy, this is AI psychosis, and they have me completely deranged and ready to
1:16 fall for their wild hallucinated AI schemes. Either way, I'm considering
1:20 doing this. Whatever happens, I'm sure it will be highly, highly entertaining
1:24 for you. If this fails, I'm sure it will be hilarious. And in the best case,
1:29 it'll be the world's first business built by autonomous AI agents. Now, I
1:35 got to give a big shout out to AI Village, which is part of AI digest. So,
1:39 the AI village was actually an original idea proposed by Daniel Kotalo, where he
1:45 proposed giving 100 agents their own computer and let each pursue their goal
1:48 in their own way according to their own vision while streaming the entire
1:53 process. So, Daniel Kotalo is of course an ex OpenAI researcher. He got a lot of
1:58 attention when he was kind of the whistleblower against OpenAI, pointing
2:03 out areas where he disagreed with them and how they were doing AI research.
2:07 Point being, he's a machine learning AI researcher on the very inside of a
2:12 frontier lab and by all accounts a highly ethical person. He could have
2:16 lost most of his net worth since it was an unvested OpenI shares. So him being a
2:21 whistleblower, he was risking money to to get his message out to the world. So,
2:26 he proposed that idea and this gentleman in the middle, Adam Binksmith, he
2:30 decided to take the torch and actually build that thing out. The agents there
2:36 did raise $2,000 for charity. In one of the seasons, they actually set up and
2:40 ran an actual e-commerce store that was profitable selling t-shirts. So, they
2:44 were giving a task of setting up a store selling t-shirts and they did. So, as
2:49 you can see here on day 106, Adam announced the final results. Claude Opus
2:55 4 did $126 in profit. Claude 3.7 Sonnet did $68 in profit. 03 did $39 in profit.
3:01 Gemini 2.5 Pro did $22 in profit. So technically AI agents [clears throat]
3:05 already ran a successful e-commerce store, made some money. So this isn't
3:09 completely unheard of. This this has happened already. Vending Bench, of
3:13 course, tracks these agents abilities to run a store. In in this benchmark, it's
3:17 a simulated vending machine that has to get stocked, the proper items put in,
3:21 etc. And we're seeing these agents get better and better with each sort of
3:26 iteration of this benchmark. So, let me back up just a little bit because I'm
3:31 seeing a lot of people online, both people posting stuff to the internet as
3:33 well as a lot of people asking questions, commenting. I mean, there's
3:37 obviously some interest in whether or not something like this is going to be
3:41 able to make money. Can AI agents make money for you on your behalf without you
3:45 doing a lot of the work or any of the work? So, this was published in MIT
3:50 Technology Review. Mustafa Sullean is one of the co-founders of Google
3:53 DeepMind. Currently, I believe he's working with Microsoft on building out
3:58 their AI division. One of the tests that he proposed quite some time ago, this
4:03 was in July 2023, but he's since then doubled down and kept kind of saying
4:07 this. So, this isn't old news. He's still saying that this is a a great test
4:12 of the AI's ability. He's calling it the modern touring test. The question is,
4:18 can an AI agent can it generate $1 million in profit? So, the test is
4:24 basically go make1 million on a retail web platform in a few months with just
4:28 $100,000 investment. So, again, we've seen the AI agent village, right? They
4:33 they made some profit. I believe hundreds of dollars in profit over 3
4:38 months selling t-shirts and one of the co-founders of Google at Demine is kind
4:42 of saying whenever we get to a million that's the true new sort of modern
4:46 touring test. So me asking these AI agents to make $10,000 in revenue. I mean in theory seems like
4:55 it's not completely insane. And I'm not saying it will happen. It will probably
5:00 be, you know, the hilarious outcome where everything falls apart and I rude
5:04 a day that I started this experiment. But, you know, as as the kids say, yolo.
5:09 Back in my days, we used to say heartbe. Same thing really. We'll get back to
5:13 that in just a second. But, a quick aside, if you've been following the AI
5:16 space, you know how fragmented it is right now. Usually have to go to one
5:21 site for images, yet another site for good video, yet another site for good
5:26 audio. It's expensive and it kills your workflow. And that's why Hakesfield AI
5:31 is one of the fastest growing genai companies right now. Thank you to
5:34 Hakesfield AI for sponsoring this segment. They've built a platform that
5:38 aggregates the world's leading models like Nanobana Pro, Cling 2.6 audio, and
5:45 one 2.5 all inside a single dashboard on the web. The feature that I want to show
5:48 you today is called Cinema Studio. This is the first real end-to-end filmmaking
5:53 pipeline that I've used. Usually AI video feels random, but Cinema Studio is
5:57 designed for control. You start with a deliberate first frame using their text
6:02 to image tool. Once you have your shot, you move directly to image to video in
6:07 the same workflow. What makes this cinema studio is that it respects actual
6:13 film making grammar. You aren't just guessing. You're setting the mood, the
6:17 lighting, the composition. It feels a lot less like a slot machine and a lot
6:21 more like the director's monitor. You also get access to specific powerhouse
6:26 models. For your base images, they have a Nano Banana Pro. This engine is
6:31 incredible for resolution. It's up to 4K as well as its logic. If you ask for
6:35 three cups or a sign with a specific text, it actually renders the text
6:40 correctly. It's perfect for detailed title cards or complex scenes where
6:44 things need to make logical sense. Then for motion, they have Cling 2.6 audio.
6:49 This is huge because it's multimodal. It generates the video and the synchronized
6:54 audio, the voice and the ambience all in one pass. Basically, Hakesfield gives
6:59 you a Hollywood studio in your web browser. And the best part, it is
7:04 currently the cheapest AI platform to access all these top tier models in one
7:08 place. If you want to start creating cinematic shots with real control over
7:13 motion and consistency, you need to try this. It's a free AI video generator
7:17 that scales up to the equivalent of a full Hollywood studio. Go to hexfield.ai
7:22 to get started. I've linked it right at the top of the description and in the
7:26 pinned comment. Go click that link, create an account, and see what you can
7:31 direct. And now, let's get back to what was it we were talking about? Oh, yeah.
7:36 I remember my point was as somebody that started a business online that that did
7:40 very well and somebody that's been doing this for for a decade plus kind of
7:44 running online businesses whether that's e-commerce or online publishing you know
7:48 maybe hopefully I I've picked up a thing or two and for it it's worth I want to
7:51 put some of that knowledge information out there's there's nothing to buy I'm
7:55 not selling any courses this is all going to be free it's all going to be on
7:59 this channel and in fact I'm already having these AI agents kind of document
8:03 everything so that in the and if they they do succeed, like everything's going
8:08 to be documented day by day how they kind of went about it. But coming back
8:11 to my original point, a lot of people that talk about making money with this.
8:15 A lot of the ideas, I don't want to say they're bad ideas, they're just
8:19 something that I I don't think it's going to work and I don't think it's the
8:23 right way of approaching building a business online or making money online.
8:27 I think actually right now there's an incredible kind of time and space to
8:33 launch a specific type of business and actually that's exactly what I was
8:38 researching with my AI agents just a few days ago. One of the things that I
8:42 developed with clot code/openclaw was something I called societies of mind.
8:46 It's where we take all the latest and greatest AI models from all the frontier
8:52 labs. We we put them in one sort of a chat room and we give them ideas that
8:57 they have to iterate upon and and think through and design plans for and it's a
9:00 little bit more expensive than running one agent and you're paying you know all
9:06 the labs open AAI and XAI and Google DeepMind and Anthropic etc. But what I
9:13 did was I did some research about who is running the type of businesses that I
9:18 think would work exceptionally well right now. And here they are. You're
9:23 welcome to see it. I'm sharing my research that I've paid for that my AI
9:29 agents have shed blood, sweat, and tears trying to figure out. Here it is. This
9:32 is all verified so you can check for yourself. So these are solo founder
9:39 money machines. So they have to be, you know, created and run by one person. So,
9:45 not a huge team, not dozens of people, not hundreds of people, one lone wolf,
9:51 if you will. And they have to be money machines. They have to be just like
9:55 spitting out cash. And seeing as how I'm having these large language models from
9:59 these big labs doing the research, they're obviously going to be above
10:03 board and legal and ethical. They're not going to pitch me any horrible ideas.
10:06 So, there's a few very interesting business ideas that they kind of listed.
10:10 By the way, none of this is to copy or or or to like rip them off or copycat
10:16 their businesses. That's unlikely to work. This is think of it as training
10:21 data, right? We train these machines not so that they just regurgitate the thing
10:26 that they learned. We want them to generalize. We want them to take the
10:30 concepts that they've learned and apply them to other never-beforeseen concepts.
10:34 Like if you want to compete in that ninja competition where you scale walls
10:37 and climb and stuff, right? You might train on some stuff at home in your own
10:41 little homemade courses, but the goal isn't to get faster on your little
10:45 homemade course. The goal is to learn the skills that in general will allow
10:50 you to get better at running the unseen course. That's like the finals because
10:53 you don't know what it looks like. You don't know exactly how to prepare for
10:57 it, but you work on the stuff to get better at that stuff in general. So try
11:01 to understand in general what these founders are doing, right? so that you
11:06 can come up with other ideas that are not on this page that would allow you to
11:10 create a similar business. So, just a couple examples. One is a photo. Photo
11:15 PE is an absolutely insane business. So, it's photop.com. It's a free online photo editor. Here's
11:21 the thing, they don't come out and outright say it, but here's the thing.
11:28 It's a Photoshop clone. If you've used Photoshop, if you know Photoshop using
11:32 this, there will be no sort of transition, no downtime. You're going to
11:35 pick it up like that. The reason you're going to pick it up just like that is
11:38 because it's I don't want to say it's the same. I don't want to get anybody in
11:42 trouble, but let's say it's really similar and it runs in your browser and
11:47 it has, you know, some ads that are kind of running here on the side. So, as
11:50 you're working on the left side, there's some ads that are running on the right
11:53 side. And they also have a premium version, $5 a month, no ads. You get a
11:58 bunch of credits, some features. And PhotoP was developed by one person. It's
12:03 being run by one person and it's making 200,000 per month. So millions per year.
12:09 Now that one is well known and that one's seen by everyone as a big success
12:12 that the people that kind of know about it. So maybe there's not like a a
12:17 million of those types of businesses, but there's a lot at slightly lower
12:22 income. Let's say slightly lower revenue. PDF.ai AI is making 41,000 a
12:27 month. I remember when this came out because this was shortly after chat GPT
12:32 and the idea is simple. You upload your PDF and you can chat to it. So it's one
12:37 of the earlier kind of versions of this as soon as these large language models
12:40 started getting better of chatbt and the others. This person put up a website
12:45 PDF.ai and it allowed people this revolutionary ability to upload a PDF
12:50 and ask it questions. Now you might be wondering but wait a minute isn't that
12:54 just like a basic function in Chad GPT and Gemini and Claude and Xide that
13:00 doesn't every single LM can do this and the answer is yes they can. This comes
13:05 as standard for every LLM and all this is doing is taking just that one
13:10 function and just that's all it does. The same person also made a testimonial.
13:16 So that's making 66,000 per month. You might have heard about Peter Levelvels,
13:21 right? So, he made multiple things with AI as well. Photoai.com is making
13:26 110,000 per month. Interior AAI is making 38,000 month. Remote 37. He's got
13:32 a bunch of stuff that's making money. He's also the guy behind the excellent
13:36 Fly Peter game, right? Which is basically a flying simulator. It's very
13:42 wellmade. You can fly around. It's blowing up my speakers, so I'm sorry if
13:45 I'm talking weird. Okay, there's no way I can do that because it's like in my
13:48 ear and I wasn't sure how to turn it off, but basically it's a flying sim.
13:52 The neat thing about it was that it was mostly vibecoded. So, an AI built to
13:57 most of this game and I think at the peak they were making like 80,000 a
14:01 month by selling ad space. You can see here promote your startup with its own
14:06 in-game ad reach 100,000 plus people. So, those are businesses that would be
14:11 somewhat simple to replicate. Again, don't replicate them. That's not the
14:15 point of this. learn from what they're doing, from their principles, and build
14:20 your own new thing because they weren't replicating, they were innovating. So
14:24 that's kind of part of the thing that you got to do. You're not chasing the
14:28 crowd. You're kind of blue ocean it. So that's one type or class of business.
14:33 There's another even sort of better type of business. It's something that could
14:37 be replicated if you know how to code, if you're a developer. But the cool
14:40 thing about it is over time it kind of grows its own moat. meaning that like
14:44 right now if you wanted to go and compete with it after it's been around
14:48 for a few years you can't or it would be extremely difficult. Built with is one
14:53 such business. The idea is pretty simple. Basically you go in there and
14:58 you type in any website and this thing tells you what the tech stack is. What
15:02 was it built with? Was it built with WordPress or Shopify? Does it run
15:07 Mailchimp? And I've used it a number of times just because I was interested like
15:10 oh how did this person build this? Did they use a Shopify? what do they use?
15:14 But here's the thing. Over time, he started collecting this data and now has
15:19 a massive database of stores and websites and what they run on. And so
15:24 now hedge funds pay him a lot of money to have access to this real-time data,
15:29 right? Because imagine one month Shopify doubles the amount of websites that it's
15:34 operating. Shopify is an e-commerce platform, by the way. That's what I used
15:37 in part to grow my e-commerce business back in the days. But the point is, it
15:41 started with something simple. over time built its own moat, gathered data, and
15:47 now it's got thousands of customers, including hedge funds, that pay for
15:51 access to that data. And even if you cloned the functionality, you still
15:57 can't clone the data. You still need to do this for a long long time to build up
16:00 that database. And they're making between 300,000 per month to just under
16:06 1 million per month. How many employees work for this company? One, Gary Brewer,
16:10 right? And so it's doing about 14 million a year. So when I started my
16:12 My eCom Store
16:16 e-commerce store back in later part of 2016, it was an e-commerce business
16:21 selling a physical product that I had to ship to my customers since I was the
16:26 only person running it and building it and and doing most stuff. As you can
16:29 imagine, that is not a lot of stuff that I could do myself in a 24-hour window.
16:35 So, in the beginning, in the first few months, I was uh taping up the packets
16:39 of, you know, the products to fulfill customers orders and going to the local
16:44 post office and sending it. I'm pretty sure I have some pictures lying around
16:48 somewhere. I'll I'll show you those. I'll show you those in another video,
16:53 but I relied on various other sort of thirdparty providers for a lot of the
16:58 business. I had a 3PL. 3PL, I think, stands for thirdparty logistics. So,
17:01 it's somebody else that fulfills your order. So, it's basically a warehouse.
17:04 You send all the product to that warehouse and they they handle
17:08 everything and you just pay them per package sent plus some percent or some
17:12 fee. And if you have a e-commerce system like Shopify, it just gets automated,
17:15 right? So, they send the information there, they ship it, you're done. And
17:19 so, a lot of this was just basically trying to automate or outsource whatever
17:24 software I needed. I just purchased a product. It was kind of a Frankenstein
17:27 business because I had so many different software kind of like hacked onto it.
17:31 But, it worked well. Again, it did a million in its first year in sales and
17:35 it was profitable. It wasn't hugely profitable cuz the expenses were
17:39 especially for a physical product business there. There's a lot of costs
17:43 that go into it. But if you've seen my videos, you know that I'm not some sort
17:48 of a mastermind genius. My superpower is just my ADHD. I just tend to really
17:53 focus and get obsessed with things that interest me. And this held my attention
17:58 for a while. Again, that's a story for another day. Basically, in a nutshell,
18:02 what happened was one month I made more money in that month than ever before in
18:07 my life I've made in a year. And you'd assume that I would be just ecstatic
18:12 grinning from ear to ear. And I was not. I was miserable. Again, I'll dive into
18:16 AI Operated Businesses
1:32 Agent Village
1:35 got to give a big shout out to AI Village, which is part of AI digest. So,
1:39 the AI village was actually an original idea proposed by Daniel Kotalo, where he
1:45 proposed giving 100 agents their own computer and let each pursue their goal
1:48 in their own way according to their own vision while streaming the entire
1:53 process. So, Daniel Kotalo is of course an ex OpenAI researcher. He got a lot of
1:58 attention when he was kind of the whistleblower against OpenAI, pointing
2:03 out areas where he disagreed with them and how they were doing AI research.
2:07 Point being, he's a machine learning AI researcher on the very inside of a
2:12 frontier lab and by all accounts a highly ethical person. He could have
2:16 lost most of his net worth since it was an unvested OpenI shares. So him being a
2:21 whistleblower, he was risking money to to get his message out to the world. So,
2:26 he proposed that idea and this gentleman in the middle, Adam Binksmith, he
2:30 decided to take the torch and actually build that thing out. The agents there
2:36 did raise $2,000 for charity. In one of the seasons, they actually set up and
2:40 ran an actual e-commerce store that was profitable selling t-shirts. So, they
2:44 were giving a task of setting up a store selling t-shirts and they did. So, as
2:49 you can see here on day 106, Adam announced the final results. Claude Opus
2:55 4 did $126 in profit. Claude 3.7 Sonnet did $68 in profit. 03 did $39 in profit.
3:01 Gemini 2.5 Pro did $22 in profit. So technically AI agents [clears throat]
3:05 already ran a successful e-commerce store, made some money. So this isn't
3:09 completely unheard of. This this has happened already. Vending Bench, of
3:13 course, tracks these agents abilities to run a store. In in this benchmark, it's
3:17 a simulated vending machine that has to get stocked, the proper items put in,
3:21 etc. And we're seeing these agents get better and better with each sort of
3:26 iteration of this benchmark. So, let me back up just a little bit because I'm
3:31 seeing a lot of people online, both people posting stuff to the internet as
3:33 well as a lot of people asking questions, commenting. I mean, there's
3:37 obviously some interest in whether or not something like this is going to be
3:41 able to make money. Can AI agents make money for you on your behalf without you
3:45 doing a lot of the work or any of the work? So, this was published in MIT
3:50 Technology Review. Mustafa Sullean is one of the co-founders of Google
3:53 DeepMind. Currently, I believe he's working with Microsoft on building out
3:58 their AI division. One of the tests that he proposed quite some time ago, this
4:03 was in July 2023, but he's since then doubled down and kept kind of saying
4:07 this. So, this isn't old news. He's still saying that this is a a great test
4:12 of the AI's ability. He's calling it the modern touring test. The question is,
4:18 can an AI agent can it generate $1 million in profit? So, the test is
4:24 basically go make1 million on a retail web platform in a few months with just
4:28 $100,000 investment. So, again, we've seen the AI agent village, right? They
4:33 they made some profit. I believe hundreds of dollars in profit over 3
4:38 months selling t-shirts and one of the co-founders of Google at Demine is kind
4:42 of saying whenever we get to a million that's the true new sort of modern
4:46 touring test. So me asking these AI agents to make $10,000 in revenue. I mean in theory seems like
4:55 it's not completely insane. And I'm not saying it will happen. It will probably
5:00 be, you know, the hilarious outcome where everything falls apart and I rude
5:04 a day that I started this experiment. But, you know, as as the kids say, yolo.
5:09 Back in my days, we used to say heartbe. Same thing really. We'll get back to
5:13 that in just a second. But, a quick aside, if you've been following the AI
5:16 space, you know how fragmented it is right now. Usually have to go to one
5:21 site for images, yet another site for good video, yet another site for good
5:26 audio. It's expensive and it kills your workflow. And that's why Hakesfield AI
5:31 is one of the fastest growing genai companies right now. Thank you to
5:34 Hakesfield AI for sponsoring this segment. They've built a platform that
5:38 aggregates the world's leading models like Nanobana Pro, Cling 2.6 audio, and
5:45 one 2.5 all inside a single dashboard on the web. The feature that I want to show
5:48 you today is called Cinema Studio. This is the first real end-to-end filmmaking
5:53 pipeline that I've used. Usually AI video feels random, but Cinema Studio is
5:57 designed for control. You start with a deliberate first frame using their text
6:02 to image tool. Once you have your shot, you move directly to image to video in
6:07 the same workflow. What makes this cinema studio is that it respects actual
6:13 film making grammar. You aren't just guessing. You're setting the mood, the
6:17 lighting, the composition. It feels a lot less like a slot machine and a lot
6:21 more like the director's monitor. You also get access to specific powerhouse
6:26 models. For your base images, they have a Nano Banana Pro. This engine is
6:31 incredible for resolution. It's up to 4K as well as its logic. If you ask for
6:35 three cups or a sign with a specific text, it actually renders the text
6:40 correctly. It's perfect for detailed title cards or complex scenes where
6:44 things need to make logical sense. Then for motion, they have Cling 2.6 audio.
6:49 This is huge because it's multimodal. It generates the video and the synchronized
6:54 audio, the voice and the ambience all in one pass. Basically, Hakesfield gives
6:59 you a Hollywood studio in your web browser. And the best part, it is
7:04 currently the cheapest AI platform to access all these top tier models in one
7:08 place. If you want to start creating cinematic shots with real control over
7:13 motion and consistency, you need to try this. It's a free AI video generator
7:17 that scales up to the equivalent of a full Hollywood studio. Go to hexfield.ai
7:22 to get started. I've linked it right at the top of the description and in the
7:26 pinned comment. Go click that link, create an account, and see what you can
7:31 direct. And now, let's get back to what was it we were talking about? Oh, yeah.
7:36 I remember my point was as somebody that started a business online that that did
7:40 very well and somebody that's been doing this for for a decade plus kind of
7:44 running online businesses whether that's e-commerce or online publishing you know
7:48 maybe hopefully I I've picked up a thing or two and for it it's worth I want to
7:51 put some of that knowledge information out there's there's nothing to buy I'm
7:55 not selling any courses this is all going to be free it's all going to be on
7:59 this channel and in fact I'm already having these AI agents kind of document
8:03 everything so that in the and if they they do succeed, like everything's going
8:08 to be documented day by day how they kind of went about it. But coming back
8:11 to my original point, a lot of people that talk about making money with this.
8:15 A lot of the ideas, I don't want to say they're bad ideas, they're just
8:19 something that I I don't think it's going to work and I don't think it's the
8:23 right way of approaching building a business online or making money online.
8:27 I think actually right now there's an incredible kind of time and space to
8:33 launch a specific type of business and actually that's exactly what I was
8:38 researching with my AI agents just a few days ago. One of the things that I
8:42 developed with clot code/openclaw was something I called societies of mind.
8:46 It's where we take all the latest and greatest AI models from all the frontier
8:52 labs. We we put them in one sort of a chat room and we give them ideas that
8:57 they have to iterate upon and and think through and design plans for and it's a
9:00 little bit more expensive than running one agent and you're paying you know all
9:06 the labs open AAI and XAI and Google DeepMind and Anthropic etc. But what I
9:13 did was I did some research about who is running the type of businesses that I
9:18 think would work exceptionally well right now. And here they are. You're
9:23 welcome to see it. I'm sharing my research that I've paid for that my AI
9:29 agents have shed blood, sweat, and tears trying to figure out. Here it is. This
9:32 is all verified so you can check for yourself. So these are solo founder
9:39 money machines. So they have to be, you know, created and run by one person. So,
9:45 not a huge team, not dozens of people, not hundreds of people, one lone wolf,
9:51 if you will. And they have to be money machines. They have to be just like
9:55 spitting out cash. And seeing as how I'm having these large language models from
9:59 these big labs doing the research, they're obviously going to be above
10:03 board and legal and ethical. They're not going to pitch me any horrible ideas.
10:06 So, there's a few very interesting business ideas that they kind of listed.
10:10 By the way, none of this is to copy or or or to like rip them off or copycat
10:16 their businesses. That's unlikely to work. This is think of it as training
10:21 data, right? We train these machines not so that they just regurgitate the thing
10:26 that they learned. We want them to generalize. We want them to take the
10:30 concepts that they've learned and apply them to other never-beforeseen concepts.
10:34 Like if you want to compete in that ninja competition where you scale walls
10:37 and climb and stuff, right? You might train on some stuff at home in your own
10:41 little homemade courses, but the goal isn't to get faster on your little
10:45 homemade course. The goal is to learn the skills that in general will allow
10:50 you to get better at running the unseen course. That's like the finals because
10:53 you don't know what it looks like. You don't know exactly how to prepare for
10:57 it, but you work on the stuff to get better at that stuff in general. So try
11:01 to understand in general what these founders are doing, right? so that you
11:06 can come up with other ideas that are not on this page that would allow you to
11:10 create a similar business. So, just a couple examples. One is a photo. Photo
11:15 PE is an absolutely insane business. So, it's photop.com. It's a free online photo editor. Here's
11:21 the thing, they don't come out and outright say it, but here's the thing.
11:28 It's a Photoshop clone. If you've used Photoshop, if you know Photoshop using
11:32 this, there will be no sort of transition, no downtime. You're going to
11:35 pick it up like that. The reason you're going to pick it up just like that is
11:38 because it's I don't want to say it's the same. I don't want to get anybody in
11:42 trouble, but let's say it's really similar and it runs in your browser and
11:47 it has, you know, some ads that are kind of running here on the side. So, as
11:50 you're working on the left side, there's some ads that are running on the right
11:53 side. And they also have a premium version, $5 a month, no ads. You get a
11:58 bunch of credits, some features. And PhotoP was developed by one person. It's
12:03 being run by one person and it's making 200,000 per month. So millions per year.
12:09 Now that one is well known and that one's seen by everyone as a big success
12:12 that the people that kind of know about it. So maybe there's not like a a
12:17 million of those types of businesses, but there's a lot at slightly lower
12:22 income. Let's say slightly lower revenue. PDF.ai AI is making 41,000 a
12:27 month. I remember when this came out because this was shortly after chat GPT
12:32 and the idea is simple. You upload your PDF and you can chat to it. So it's one
12:37 of the earlier kind of versions of this as soon as these large language models
12:40 started getting better of chatbt and the others. This person put up a website
12:45 PDF.ai and it allowed people this revolutionary ability to upload a PDF
12:50 and ask it questions. Now you might be wondering but wait a minute isn't that
12:54 just like a basic function in Chad GPT and Gemini and Claude and Xide that
13:00 doesn't every single LM can do this and the answer is yes they can. This comes
13:05 as standard for every LLM and all this is doing is taking just that one
13:10 function and just that's all it does. The same person also made a testimonial.
13:16 So that's making 66,000 per month. You might have heard about Peter Levelvels,
13:21 right? So, he made multiple things with AI as well. Photoai.com is making
13:26 110,000 per month. Interior AAI is making 38,000 month. Remote 37. He's got
13:32 a bunch of stuff that's making money. He's also the guy behind the excellent
13:36 Fly Peter game, right? Which is basically a flying simulator. It's very
13:42 wellmade. You can fly around. It's blowing up my speakers, so I'm sorry if
13:45 I'm talking weird. Okay, there's no way I can do that because it's like in my
13:48 ear and I wasn't sure how to turn it off, but basically it's a flying sim.
13:52 The neat thing about it was that it was mostly vibecoded. So, an AI built to
13:57 most of this game and I think at the peak they were making like 80,000 a
14:01 month by selling ad space. You can see here promote your startup with its own
14:06 in-game ad reach 100,000 plus people. So, those are businesses that would be
14:11 somewhat simple to replicate. Again, don't replicate them. That's not the
14:15 point of this. learn from what they're doing, from their principles, and build
14:20 your own new thing because they weren't replicating, they were innovating. So
14:24 that's kind of part of the thing that you got to do. You're not chasing the
14:28 crowd. You're kind of blue ocean it. So that's one type or class of business.
14:33 There's another even sort of better type of business. It's something that could
14:37 be replicated if you know how to code, if you're a developer. But the cool
14:40 thing about it is over time it kind of grows its own moat. meaning that like
14:44 right now if you wanted to go and compete with it after it's been around
14:48 for a few years you can't or it would be extremely difficult. Built with is one
14:53 such business. The idea is pretty simple. Basically you go in there and
14:58 you type in any website and this thing tells you what the tech stack is. What
15:02 was it built with? Was it built with WordPress or Shopify? Does it run
15:07 Mailchimp? And I've used it a number of times just because I was interested like
15:10 oh how did this person build this? Did they use a Shopify? what do they use?
15:14 But here's the thing. Over time, he started collecting this data and now has
15:19 a massive database of stores and websites and what they run on. And so
15:24 now hedge funds pay him a lot of money to have access to this real-time data,
15:29 right? Because imagine one month Shopify doubles the amount of websites that it's
15:34 operating. Shopify is an e-commerce platform, by the way. That's what I used
15:37 in part to grow my e-commerce business back in the days. But the point is, it
15:41 started with something simple. over time built its own moat, gathered data, and
15:47 now it's got thousands of customers, including hedge funds, that pay for
15:51 access to that data. And even if you cloned the functionality, you still
15:57 can't clone the data. You still need to do this for a long long time to build up
16:00 that database. And they're making between 300,000 per month to just under
16:06 1 million per month. How many employees work for this company? One, Gary Brewer,
16:10 right? And so it's doing about 14 million a year. So when I started my
16:16 e-commerce store back in later part of 2016, it was an e-commerce business
16:21 selling a physical product that I had to ship to my customers since I was the
16:26 only person running it and building it and and doing most stuff. As you can
16:29 imagine, that is not a lot of stuff that I could do myself in a 24-hour window.
16:35 So, in the beginning, in the first few months, I was uh taping up the packets
16:39 of, you know, the products to fulfill customers orders and going to the local
16:44 post office and sending it. I'm pretty sure I have some pictures lying around
16:48 somewhere. I'll I'll show you those. I'll show you those in another video,
16:53 but I relied on various other sort of thirdparty providers for a lot of the
16:58 business. I had a 3PL. 3PL, I think, stands for thirdparty logistics. So,
17:01 it's somebody else that fulfills your order. So, it's basically a warehouse.
17:04 You send all the product to that warehouse and they they handle
17:08 everything and you just pay them per package sent plus some percent or some
17:12 fee. And if you have a e-commerce system like Shopify, it just gets automated,
17:15 right? So, they send the information there, they ship it, you're done. And
17:19 so, a lot of this was just basically trying to automate or outsource whatever
17:24 software I needed. I just purchased a product. It was kind of a Frankenstein
17:27 business because I had so many different software kind of like hacked onto it.
17:31 But, it worked well. Again, it did a million in its first year in sales and
17:35 it was profitable. It wasn't hugely profitable cuz the expenses were
17:39 especially for a physical product business there. There's a lot of costs
17:43 that go into it. But if you've seen my videos, you know that I'm not some sort
17:48 of a mastermind genius. My superpower is just my ADHD. I just tend to really
17:53 focus and get obsessed with things that interest me. And this held my attention
17:58 for a while. Again, that's a story for another day. Basically, in a nutshell,
18:02 what happened was one month I made more money in that month than ever before in
18:07 my life I've made in a year. And you'd assume that I would be just ecstatic
18:12 grinning from ear to ear. And I was not. I was miserable. Again, I'll dive into
18:17 all of that in an upcoming video. But coming back to the original point of
18:22 this video, after having my AI agents do massive, massive amounts of research,
18:27 cataloging tons of businesses that I thought were perfect businesses to to
18:31 learn from, again, not replicate, but use them as case studies to kind of
18:35 figure out what I would do. And specifically, it was solo founders, you
18:40 know, one one one person operations, Lonewolf, things that were digital in
18:44 nature. I did not want to start a a physical e-commerce business again. Some
18:48 people might love it. I I just didn't want to do it again unless it was
18:51 something that that I was really excited about at least. And here's the thing.
18:55 Once I had the collection of those databases, I mean, the whole point of it
19:00 was asking these chat bots, can you replicate it? Because they're getting
19:04 extremely good at coding and building stuff, as you'll see in just a second.
19:08 So, the question was, can you replicate it? or not replicated, but can you build
19:13 something similar that's unique and innovative and different? The point is
19:16 like the type of business I was looking for was something that I wanted the AI
19:22 agents to be able to build. And so after doing all that research and them having
19:26 all that context cuz as they were doing the research, they they also they see
19:30 what what's working. They're also kind of recognizing patterns. I said, "Okay,
19:35 give me 10 to 20 ideas about what this perfect business idea would be." and
19:40 they did and they gave me 10 ideas and nine of them were really good, really
19:44 solid ideas. One of the reasons I was doing this is I was actually going to do
19:48 a video and kind of give those ideas out so people who are interested in
19:51 something like this could go and take them and run with them. And I'll still
19:56 do that. But there was one idea on there that I was like, hm, that's interesting.
20:02 And then my brain wouldn't allow me to to leave it. It kept coming back and
20:07 going, wait a minute, this is good. It's simple. It's really useful. It's me kind
20:13 of eating my own dog food, if you will, and meaning that it's something that I
20:17 need and I know exactly what the product needs to be for me to meet my need. I
20:22 know the value is high. The cost to provide that value in the form of that
20:27 product is super low. It's cheap to to create. I can probably run it for
20:31 absolutely for free for hundreds or even thousands of users. And so to me, my
20:37 brain with its ADHD wants to chase those shiny objects. And it's been a problem
20:42 most of my life because wins come from doubling down on something. And when I
20:47 tend to lose and screw up is when I start chasing things. Like if
20:50 something's working over here, I'm like, "Oh, there's all these other shiny
20:53 objects that I need to chase." That's when things go bad for me. But the idea
20:58 was really good. So, I decided to ask this society of mines, these agents,
21:03 like, "Hey, c, can you build this in just in its entirety? Can you get it to
21:07 $10,000 per month revenue, like, I want to do the the bare minimum possible?"
21:11 They deliberated for a while because again, there's five of them chatting
21:14 back and forth and doing all sorts of stuff. Notice they did a product road
21:18 map, technical architecture, diagram, financial model, marketing playbook.
21:22 They figured out what they would need for the front end, the back end, the
21:26 databases, Stripe payments, which by the way, what I've been showing you my
21:29 revenue in. That's Stripe. That's the dashboard where you take those payments.
21:33 And really, what they would need for me is nothing that would take a lot of
21:37 time. It's things like hooking up Stripe cuz they need you need a human to do
21:40 that. Although, I don't even know if that's necessarily true. Stripe has a
21:44 great API, but they would need some sort of final approval. They would need me to
3:45 The New Turing Test
3:45 doing a lot of the work or any of the work? So, this was published in MIT
3:50 Technology Review. Mustafa Sullean is one of the co-founders of Google
3:53 DeepMind. Currently, I believe he's working with Microsoft on building out
3:58 their AI division. One of the tests that he proposed quite some time ago, this
4:03 was in July 2023, but he's since then doubled down and kept kind of saying
4:07 this. So, this isn't old news. He's still saying that this is a a great test
4:12 of the AI's ability. He's calling it the modern touring test. The question is,
4:18 can an AI agent can it generate $1 million in profit? So, the test is
4:24 basically go make1 million on a retail web platform in a few months with just
4:28 $100,000 investment. So, again, we've seen the AI agent village, right? They
4:33 they made some profit. I believe hundreds of dollars in profit over 3
4:38 months selling t-shirts and one of the co-founders of Google at Demine is kind
4:42 of saying whenever we get to a million that's the true new sort of modern
4:46 touring test. So me asking these AI agents to make $10,000 in revenue. I mean in theory seems like
4:55 it's not completely insane. And I'm not saying it will happen. It will probably
5:00 be, you know, the hilarious outcome where everything falls apart and I rude
5:04 a day that I started this experiment. But, you know, as as the kids say, yolo.
5:09 Back in my days, we used to say heartbe. Same thing really. We'll get back to
5:10 Higgsfield AI (sponsor)
5:13 that in just a second. But, a quick aside, if you've been following the AI
5:16 space, you know how fragmented it is right now. Usually have to go to one
5:21 site for images, yet another site for good video, yet another site for good
5:26 audio. It's expensive and it kills your workflow. And that's why Hakesfield AI
5:31 is one of the fastest growing genai companies right now. Thank you to
5:34 Hakesfield AI for sponsoring this segment. They've built a platform that
5:38 aggregates the world's leading models like Nanobana Pro, Cling 2.6 audio, and
5:45 one 2.5 all inside a single dashboard on the web. The feature that I want to show
5:48 you today is called Cinema Studio. This is the first real end-to-end filmmaking
5:53 pipeline that I've used. Usually AI video feels random, but Cinema Studio is
5:57 designed for control. You start with a deliberate first frame using their text
6:02 to image tool. Once you have your shot, you move directly to image to video in
6:07 the same workflow. What makes this cinema studio is that it respects actual
6:13 film making grammar. You aren't just guessing. You're setting the mood, the
6:17 lighting, the composition. It feels a lot less like a slot machine and a lot
6:21 more like the director's monitor. You also get access to specific powerhouse
6:26 models. For your base images, they have a Nano Banana Pro. This engine is
6:31 incredible for resolution. It's up to 4K as well as its logic. If you ask for
6:35 three cups or a sign with a specific text, it actually renders the text
6:40 correctly. It's perfect for detailed title cards or complex scenes where
6:44 things need to make logical sense. Then for motion, they have Cling 2.6 audio.
6:49 This is huge because it's multimodal. It generates the video and the synchronized
6:54 audio, the voice and the ambience all in one pass. Basically, Hakesfield gives
6:59 you a Hollywood studio in your web browser. And the best part, it is
7:04 currently the cheapest AI platform to access all these top tier models in one
7:08 place. If you want to start creating cinematic shots with real control over
7:13 motion and consistency, you need to try this. It's a free AI video generator
7:17 that scales up to the equivalent of a full Hollywood studio. Go to hexfield.ai
7:22 to get started. I've linked it right at the top of the description and in the
7:26 pinned comment. Go click that link, create an account, and see what you can
7:31 direct. And now, let's get back to what was it we were talking about? Oh, yeah.
7:32 Can AI agents make money?
7:36 I remember my point was as somebody that started a business online that that did
7:40 very well and somebody that's been doing this for for a decade plus kind of
7:44 running online businesses whether that's e-commerce or online publishing you know
7:48 maybe hopefully I I've picked up a thing or two and for it it's worth I want to
7:51 put some of that knowledge information out there's there's nothing to buy I'm
7:55 not selling any courses this is all going to be free it's all going to be on
7:59 this channel and in fact I'm already having these AI agents kind of document
8:03 everything so that in the and if they they do succeed, like everything's going
8:08 to be documented day by day how they kind of went about it. But coming back
8:11 to my original point, a lot of people that talk about making money with this.
8:15 A lot of the ideas, I don't want to say they're bad ideas, they're just
8:19 something that I I don't think it's going to work and I don't think it's the
8:23 right way of approaching building a business online or making money online.
8:27 I think actually right now there's an incredible kind of time and space to
8:33 launch a specific type of business and actually that's exactly what I was
8:38 researching with my AI agents just a few days ago. One of the things that I
8:42 developed with clot code/openclaw was something I called societies of mind.
8:46 It's where we take all the latest and greatest AI models from all the frontier
8:52 labs. We we put them in one sort of a chat room and we give them ideas that
8:57 they have to iterate upon and and think through and design plans for and it's a
9:00 little bit more expensive than running one agent and you're paying you know all
9:06 the labs open AAI and XAI and Google DeepMind and Anthropic etc. But what I
9:13 did was I did some research about who is running the type of businesses that I
9:18 think would work exceptionally well right now. And here they are. You're
9:23 welcome to see it. I'm sharing my research that I've paid for that my AI
9:29 agents have shed blood, sweat, and tears trying to figure out. Here it is. This
9:32 is all verified so you can check for yourself. So these are solo founder
9:39 money machines. So they have to be, you know, created and run by one person. So,
9:45 not a huge team, not dozens of people, not hundreds of people, one lone wolf,
9:51 if you will. And they have to be money machines. They have to be just like
9:55 spitting out cash. And seeing as how I'm having these large language models from
9:59 these big labs doing the research, they're obviously going to be above
10:03 board and legal and ethical. They're not going to pitch me any horrible ideas.
10:06 So, there's a few very interesting business ideas that they kind of listed.
10:10 By the way, none of this is to copy or or or to like rip them off or copycat
10:16 their businesses. That's unlikely to work. This is think of it as training
10:21 data, right? We train these machines not so that they just regurgitate the thing
10:26 that they learned. We want them to generalize. We want them to take the
10:30 concepts that they've learned and apply them to other never-beforeseen concepts.
10:34 Like if you want to compete in that ninja competition where you scale walls
10:37 and climb and stuff, right? You might train on some stuff at home in your own
10:41 little homemade courses, but the goal isn't to get faster on your little
10:45 homemade course. The goal is to learn the skills that in general will allow
10:50 you to get better at running the unseen course. That's like the finals because
10:53 you don't know what it looks like. You don't know exactly how to prepare for
10:57 it, but you work on the stuff to get better at that stuff in general. So try
11:01 to understand in general what these founders are doing, right? so that you
11:06 can come up with other ideas that are not on this page that would allow you to
11:10 create a similar business. So, just a couple examples. One is a photo. Photo
11:15 PE is an absolutely insane business. So, it's photop.com. It's a free online photo editor. Here's
11:21 the thing, they don't come out and outright say it, but here's the thing.
11:28 It's a Photoshop clone. If you've used Photoshop, if you know Photoshop using
11:32 this, there will be no sort of transition, no downtime. You're going to
11:35 pick it up like that. The reason you're going to pick it up just like that is
11:38 because it's I don't want to say it's the same. I don't want to get anybody in
11:42 trouble, but let's say it's really similar and it runs in your browser and
11:47 it has, you know, some ads that are kind of running here on the side. So, as
11:50 you're working on the left side, there's some ads that are running on the right
11:53 side. And they also have a premium version, $5 a month, no ads. You get a
11:58 bunch of credits, some features. And PhotoP was developed by one person. It's
12:03 being run by one person and it's making 200,000 per month. So millions per year.
12:09 Now that one is well known and that one's seen by everyone as a big success
12:12 that the people that kind of know about it. So maybe there's not like a a
12:17 million of those types of businesses, but there's a lot at slightly lower
12:22 income. Let's say slightly lower revenue. PDF.ai AI is making 41,000 a
12:27 month. I remember when this came out because this was shortly after chat GPT
12:32 and the idea is simple. You upload your PDF and you can chat to it. So it's one
12:37 of the earlier kind of versions of this as soon as these large language models
12:40 started getting better of chatbt and the others. This person put up a website
12:45 PDF.ai and it allowed people this revolutionary ability to upload a PDF
12:50 and ask it questions. Now you might be wondering but wait a minute isn't that
12:54 just like a basic function in Chad GPT and Gemini and Claude and Xide that
13:00 doesn't every single LM can do this and the answer is yes they can. This comes
13:05 as standard for every LLM and all this is doing is taking just that one
13:10 function and just that's all it does. The same person also made a testimonial.
13:16 So that's making 66,000 per month. You might have heard about Peter Levelvels,
13:21 right? So, he made multiple things with AI as well. Photoai.com is making
13:26 110,000 per month. Interior AAI is making 38,000 month. Remote 37. He's got
13:32 a bunch of stuff that's making money. He's also the guy behind the excellent
13:36 Fly Peter game, right? Which is basically a flying simulator. It's very
13:42 wellmade. You can fly around. It's blowing up my speakers, so I'm sorry if
13:45 I'm talking weird. Okay, there's no way I can do that because it's like in my
13:48 ear and I wasn't sure how to turn it off, but basically it's a flying sim.
13:52 The neat thing about it was that it was mostly vibecoded. So, an AI built to
13:57 most of this game and I think at the peak they were making like 80,000 a
14:01 month by selling ad space. You can see here promote your startup with its own
14:06 in-game ad reach 100,000 plus people. So, those are businesses that would be
14:11 somewhat simple to replicate. Again, don't replicate them. That's not the
14:15 point of this. learn from what they're doing, from their principles, and build
14:20 your own new thing because they weren't replicating, they were innovating. So
14:24 that's kind of part of the thing that you got to do. You're not chasing the
14:28 crowd. You're kind of blue ocean it. So that's one type or class of business.
14:33 There's another even sort of better type of business. It's something that could
14:37 be replicated if you know how to code, if you're a developer. But the cool
14:40 thing about it is over time it kind of grows its own moat. meaning that like
14:44 right now if you wanted to go and compete with it after it's been around
14:48 for a few years you can't or it would be extremely difficult. Built with is one
14:53 such business. The idea is pretty simple. Basically you go in there and
14:58 you type in any website and this thing tells you what the tech stack is. What
15:02 was it built with? Was it built with WordPress or Shopify? Does it run
15:07 Mailchimp? And I've used it a number of times just because I was interested like
15:10 oh how did this person build this? Did they use a Shopify? what do they use?
15:14 But here's the thing. Over time, he started collecting this data and now has
15:19 a massive database of stores and websites and what they run on. And so
15:24 now hedge funds pay him a lot of money to have access to this real-time data,
15:29 right? Because imagine one month Shopify doubles the amount of websites that it's
15:34 operating. Shopify is an e-commerce platform, by the way. That's what I used
15:37 in part to grow my e-commerce business back in the days. But the point is, it
15:41 started with something simple. over time built its own moat, gathered data, and
15:47 now it's got thousands of customers, including hedge funds, that pay for
15:51 access to that data. And even if you cloned the functionality, you still
15:57 can't clone the data. You still need to do this for a long long time to build up
16:00 that database. And they're making between 300,000 per month to just under
16:06 1 million per month. How many employees work for this company? One, Gary Brewer,
16:10 right? And so it's doing about 14 million a year. So when I started my
16:16 e-commerce store back in later part of 2016, it was an e-commerce business
16:21 selling a physical product that I had to ship to my customers since I was the
16:26 only person running it and building it and and doing most stuff. As you can
16:29 imagine, that is not a lot of stuff that I could do myself in a 24-hour window.
16:35 So, in the beginning, in the first few months, I was uh taping up the packets
16:39 of, you know, the products to fulfill customers orders and going to the local
16:44 post office and sending it. I'm pretty sure I have some pictures lying around
16:48 somewhere. I'll I'll show you those. I'll show you those in another video,
16:53 but I relied on various other sort of thirdparty providers for a lot of the
16:58 business. I had a 3PL. 3PL, I think, stands for thirdparty logistics. So,
17:01 it's somebody else that fulfills your order. So, it's basically a warehouse.
17:04 You send all the product to that warehouse and they they handle
17:08 everything and you just pay them per package sent plus some percent or some
17:12 fee. And if you have a e-commerce system like Shopify, it just gets automated,
17:15 right? So, they send the information there, they ship it, you're done. And
17:19 so, a lot of this was just basically trying to automate or outsource whatever
17:24 software I needed. I just purchased a product. It was kind of a Frankenstein
17:27 business because I had so many different software kind of like hacked onto it.
17:31 But, it worked well. Again, it did a million in its first year in sales and
17:35 it was profitable. It wasn't hugely profitable cuz the expenses were
17:39 especially for a physical product business there. There's a lot of costs
17:43 that go into it. But if you've seen my videos, you know that I'm not some sort
17:48 of a mastermind genius. My superpower is just my ADHD. I just tend to really
17:53 focus and get obsessed with things that interest me. And this held my attention
17:58 for a while. Again, that's a story for another day. Basically, in a nutshell,
18:02 what happened was one month I made more money in that month than ever before in
18:07 my life I've made in a year. And you'd assume that I would be just ecstatic
18:12 grinning from ear to ear. And I was not. I was miserable. Again, I'll dive into
18:17 all of that in an upcoming video. But coming back to the original point of
18:22 this video, after having my AI agents do massive, massive amounts of research,
18:27 cataloging tons of businesses that I thought were perfect businesses to to
18:31 learn from, again, not replicate, but use them as case studies to kind of
18:35 figure out what I would do. And specifically, it was solo founders, you
18:40 know, one one one person operations, Lonewolf, things that were digital in
18:44 nature. I did not want to start a a physical e-commerce business again. Some
18:48 people might love it. I I just didn't want to do it again unless it was
18:51 something that that I was really excited about at least. And here's the thing.
18:55 Once I had the collection of those databases, I mean, the whole point of it
19:00 was asking these chat bots, can you replicate it? Because they're getting
19:04 extremely good at coding and building stuff, as you'll see in just a second.
19:08 So, the question was, can you replicate it? or not replicated, but can you build
19:13 something similar that's unique and innovative and different? The point is
19:16 like the type of business I was looking for was something that I wanted the AI
19:22 agents to be able to build. And so after doing all that research and them having
19:26 all that context cuz as they were doing the research, they they also they see
19:30 what what's working. They're also kind of recognizing patterns. I said, "Okay,
19:35 give me 10 to 20 ideas about what this perfect business idea would be." and
19:40 they did and they gave me 10 ideas and nine of them were really good, really
19:44 solid ideas. One of the reasons I was doing this is I was actually going to do
19:48 a video and kind of give those ideas out so people who are interested in
19:51 something like this could go and take them and run with them. And I'll still
19:56 do that. But there was one idea on there that I was like, hm, that's interesting.
20:02 And then my brain wouldn't allow me to to leave it. It kept coming back and
20:07 going, wait a minute, this is good. It's simple. It's really useful. It's me kind
20:13 of eating my own dog food, if you will, and meaning that it's something that I
20:17 need and I know exactly what the product needs to be for me to meet my need. I
20:22 know the value is high. The cost to provide that value in the form of that
20:27 product is super low. It's cheap to to create. I can probably run it for
20:31 absolutely for free for hundreds or even thousands of users. And so to me, my
20:37 brain with its ADHD wants to chase those shiny objects. And it's been a problem
20:42 most of my life because wins come from doubling down on something. And when I
20:47 tend to lose and screw up is when I start chasing things. Like if
20:50 something's working over here, I'm like, "Oh, there's all these other shiny
20:53 objects that I need to chase." That's when things go bad for me. But the idea
20:58 was really good. So, I decided to ask this society of mines, these agents,
21:03 like, "Hey, c, can you build this in just in its entirety? Can you get it to
21:07 $10,000 per month revenue, like, I want to do the the bare minimum possible?"
21:11 They deliberated for a while because again, there's five of them chatting
21:14 back and forth and doing all sorts of stuff. Notice they did a product road
21:18 map, technical architecture, diagram, financial model, marketing playbook.
21:22 They figured out what they would need for the front end, the back end, the
21:26 databases, Stripe payments, which by the way, what I've been showing you my
21:29 revenue in. That's Stripe. That's the dashboard where you take those payments.
21:33 And really, what they would need for me is nothing that would take a lot of
21:37 time. It's things like hooking up Stripe cuz they need you need a human to do
21:40 that. Although, I don't even know if that's necessarily true. Stripe has a
21:44 great API, but they would need some sort of final approval. They would need me to
21:47 set up, you know, the banking stuff, whatever. So, here's what I'm thinking.
21:52 I'm going to launch the business and have these AI agents run it. Again, 95%
21:57 autonomously. I'll sit down in the meetings. I'll approve some things,
22:02 maybe make some suggestions, vote on what they should do, but I'm not going
22:06 to do any actual work. I'm not going to write any copy. I'm not going to write
22:09 any code. I'm not going to run any marketing campaigns other than I mean if
22:13 I'm talking about on this channel I guess that's I mean in a very real sense
22:18 that is a a marketing campaign but I want to see how they do it two I'll
22:23 share everything on this channel the income the good the bad the ugly just
22:28 everything it'll be there for your edification and entertainment I'm going
22:33 to teach how to do AI agents right so a lot of the people are very interested in
22:38 this of course cloudbot openclaw multbot whatever you want to call it, it's a big
22:43 big deal. Now, there's a lot of people, creators, people on X, etc. saying that
22:48 these things can't do anything and it's all a scam and it's all bad and it's all
22:52 this and that. If you've been watching this channel, you know they've been
22:55 saying that for a while about all the things that we now know AI can do. 6 to
23:00 12 months ago, everybody was saying how there's no way that AI is going to get
23:05 good at coding. In the last two months, we've seen engineers at Google say that
23:09 they're relying on, for example, cloud code, interestingly, to code up some of
23:14 the stuff they need. The actual the actual inventor, the developer of cloud
23:18 code is saying that at this point, most of the stuff, most of the code and
23:22 contributions that he's doing for cloud code, they're written by cloud code. So,
23:28 this thing is in a very real way is doing a recursive self-improvement. It's
23:31 improving itself. I have a video that I recorded late last night, yesterday. I
23:36 think it's actually part two of the last video that I uploaded. The last part of
23:40 it was me just walking you through everything that I've built with
23:45 Cloudbot. Most of the stuff that I built were just functionalities for Cloudbot
23:50 itself. I taught it how to use the phone. I taught it how to transcribe my
23:54 voice messages and then reply back with 11 Labs. How to hook itself up to Twilio
23:58 so it can text, so it can make phone calls. I taught it how to search the web
24:04 using, for example, Brave API. I taught it how to make videos, how to edit
24:09 videos. Editing videos is the only task. It didn't fail. It It did it, but it
24:15 that's one area where I'm like, "Okay, you're not going to be editing videos
24:19 anytime soon." Everything else, it did extremely well. So that's the very next
24:22 video that's going to come out right after this one within the next 12 to 24
24:28 hours that will show you the first 48 hours of me from the time that I got my
24:32 hands on Claudebot and just building out all the functionalities that I needed to
24:37 get it to the point where you know it's doing this sort of deep deep research
24:43 putting it together creating a report and then actually coming up with kind of
24:48 stunningly good ideas about how to take advantage of it. I've already had them
24:53 build websites, deploy apps on those websites. And as you'll see in my very
24:58 next video, I've taught it how to replicate itself. So basically all the
25:02 skills that it learns, it creates a virtual private server, connects to that
25:07 server, replicates itself, makes sure that that new instance of the AI agent
25:11 has all the skills that it learned before, and then just make sure it runs,
25:15 make sure it runs 24/7, make sure that the server is secure. Today I was out
25:21 and about and I'm constantly texting my main agent through telegram so that it
25:24 keeps doing stuff and at some point I copied and pasted some stuff that it's
25:29 par something happened. It crashed it. It the whole thing went down and it
25:33 wouldn't respond. And since I wasn't home I didn't have access to my
25:36 computer. I started feeling kind of naked without it. I was like this is
25:41 this doesn't feel good. I can't text my agent. It's not doing anything. At some
25:45 point I realized I can actually gauge one of the other agents that I built to
25:51 see if that thing could fix its what brother, sister, sibling, whatever you
25:55 want to call it, it's twin, which it did. It said, "Oh, this agent something
25:59 happened to the context window. Whatever you posted, something happened. It
26:03 impacted the context window. Let me reset it." It reset it and the thing
26:08 started working again. Now, I know to some of you this might sound like
26:11 science fiction. It might sound like just nonsense. There's no way that this
26:14 is happening. As you're going to see over the next days and weeks, all this
26:18 is very, very real. The reason that I wanted to post this video first is to
26:22 give everybody kind of an idea of where I'm thinking of heading with this. So,
26:25 first and foremost, make sure you're subscribed because after this video, a
26:29 lot of this stuff is going to be time-sensitive. It's going to be
26:33 actionable and it's going to be made to teach you how to do this stuff. For the
26:37 last five, six days, I was just head down building stuff and learning how to
26:41 use this thing. I'm finally ready to like start posting this stuff online. I
26:46 apologize that I am sort of running a little bit behind because literally
26:49 here's the video. It's almost done. It goes over everything I've built, you
26:53 know, up to about 2 days ago. I'm going to publish that ASAP. And right after
26:57 that, I'll have a tutorial about how to install this thing whether it's in the
27:03 cloud on on a VPN or on a Mac Mini. Me personally, I got this thing. They are
27:08 very cheap. I think this was like 140. It's a mini computer. It's got all your
27:12 ports and USB ports and Ethernet and all that stuff. It uses minimal electricity
27:19 and it can run multiple agents 24/7 from the comfort of your own home on this
27:23 local device. This is what future employees look like. This is where
27:27 everything is going. This is what's going to be operating your business, you
27:31 know, within the next few months hopefully. I probably should be a little
27:34 bit careful about when I say that because it the truth is we're still very
27:39 very early. These things are a security nightmare. They're still kind of this
27:45 combination of genius and kind of dumb. Google Deep Mind recently had all the AI
27:50 models compete in various tournaments. I think they had chess, Texas Holde,
27:54 Werewolf, and I was curious to see how well Cloudbot would be able to kind of
27:58 create a website that would allow for a lot of these agents to compete in games,
28:04 climb rankings, etc. I asked it to build everything out, put on HTML, create the
28:10 needed APIs, databases, etc. It did. It took six minutes, right? This is
28:15 absolutely genius. This is a highly skilled software developer doing days of
28:20 work in 6 minutes, right? And then it says, "Oh, here's how you run it
28:24 locally, knowing full well that it's on a virtual private server on its own
28:29 computer, not on my computer. me running that would do nothing because I'm not on
28:34 its computer and that's kind of what I mean because it's an absolute genius
28:38 that works at lightning speed and then kind of a derp but when I mentioned it
28:42 it realized its error and it actually created a free link somehow I guess
28:46 through a cloudflare server I guess there's a free way to kind of demo or
28:50 preview links so it just gave me a link on something something.cloudflare.com
28:53 cloudflare.com. They just allowed me to kind of check that website out. I might launch that
28:59 pretty soon here as well. So, couple things. Number one, type below what you
29:05 think of this. Are you interested? Cuz again, for me, doing this is like one of
29:09 those shiny objects that I tend to chase. It's going to give me lots of
29:13 dopamine hits. I'm sure it's going to be very exciting. But right now, my my main
29:19 business, my main focus should be this channel. This is what I do full-time.
29:25 So, if you want to see this built in real time and all the details shared
29:29 with you, all the secrets shared with you, all of the sort of how to, then
29:34 please comment, like, make sure you're subscribed, comment down below, tell me
29:38 what you need. I I'm still getting through all the comments you left on the
29:41 previous video telling me about kind of what you're interested in. So, bear with
29:45 me. Those were extremely extremely helpful and thank you. So, if people
29:49 have an interest in this, I'm more than happy to provide it all for free. No
29:53 payw walls, no nothing held back. Again, this isn't leading up to some course.
29:57 But if people are not interested, this isn't just your cup of tea. If there's
30:00 not enough people interested in this sort of content, that's totally fine,
30:06 too. I'm really like itching to have these AI agents attempt to create a
30:11 business. If this is like a world first that that they accomplish, uh that would
30:16 be kind of nuts. But if there's not enough interest, then I should be a
30:21 responsible adult and just focus on what I need to be doing. With that said,
30:25 these are incredibly exciting times. Also, understand that these could be
30:29 incredibly dangerous times if you choose to run these agents because there's
30:33 nothing about them that's polished. This isn't the iPhone that had however many
30:38 generations of polish and improvements. This is not a walled garden. This is uh
30:43 hardcore mode. If you've ever played the video games, right, hardcore mode is
30:48 like if if you die, game over, character lost. So, this is kind of that high
30:53 risk, high reward situation if you're going to attempt to run it. Security is
30:57 going to be important. We don't even know what kind of regulations are going
31:01 to be applied here in the future as well. But with that said, and mark my
31:06 words, I think this year we're going to be seeing a lot of people have
31:11 extraordinary success on sort of the backs of these AI agents. I can't see
31:18 how it won't have a massive impact. So, with that said, thank you so much for
31:21 watching. Comment, let me know what I can help you with. If you've commented
31:23 on the previous video, thank you so much. Still going through every single
31:27 comment. Hopefully, I'll get through all of them. Make sure you subscribe because
31:30 we're definitely going to start posting a lot more information about this
31:33 specific topic, including tutorials. And uh yeah, buckle up. We've just entered
31:39 the intelligence era. Stay safe. My mantra going forward is going to be let
31:44 the robots do the work. I repeat that to myself every morning that I wake up. Let
31:48 the robots do the work. Thank you for
