Skip to content

Podcast

This Guy Cured His Dog’s Cancer With ChatGPT + 4 Other Crazy AI Stories

My First Million

Source ↗ ← All highlights
  • Pokemon Go Pivoted Into A Mapping Data Goldmine
    • Niantic repurposed Pokemon Go’s user-captured street images to become a valuable mapping dataset for AI companies building delivery and navigation models.
    • That data is unique because players effectively drove global coverage of street-level imagery across cities, useful for robotic delivery training. Transcript: Shaan Puri Like 19 or 20? Yeah, like, you know, seven years ago or something like that. It was a huge deal that people, it was this new style of game. You would take out your phone and your phone would just be on like kind of camera mode. You’d walk around a city and then you would like discover a Pokemon and it would show up like using like like the kind of, they would overlay the little Pokemon on top of like the street Where you were standing. All right, cool. So that became like a kind of like a one to $2 billion game. I think they sold the game for two or $3 billion because it had this huge hype cycle that it kind of died down. Well, there’s this really interesting story about the fact that they’re now, they now, they sold the game, but they kept all of the photos, the videos. And so what that means is like, they had, I don’t know, 100 million players or something like that walking around every city in the world with their phone out, capturing what the street Looks like. And if you think about it, that’s actually really valuable data for AI that wants to do not self-driving, but like self-delivery, right? So like delivery bots that are going to go on sidewalks in through neighborhoods or down city streets and have to deliver something. Well, they need to recognize where am I? What does this look like? What is the terrain? How does this work? Etc. And so they took that data and they’re licensing it out to all these AI companies now because they’re the only ones in the world who have this data. And like, you’ve heard this kind of like, data’s the new oil. That’s a little misleading because there’s just oil. With data, there’s like very specific, like niche types of data. And so you might have like a specific type of oil that nobody else has. And Pokemon Go, the company behind it, Niantic, has that data for real world behavior. (Time 0:29:53)
  • Karpathy’s Job Map Shows AI Exposure By Occupation
    • Andrej Karpathy’s jobs visualization maps 143 million US jobs and scores them by AI ‘digital exposure’ to show which occupations will reshape versus remain stable.
    • The tool highlights large at‑risk categories (cashiers, secretaries) and cautions scores don’t account for demand growth or regulatory barriers. Transcript: Sam Parr So have you seen this Karpathy Jobs thing? No, but I saw the Anthropik version, but what’s the Karpathy one? Do you know who Karpathy is? I don’t even know who he is, but I know that when people say, Karpathy says blank about AI, you’re supposed to take it seriously. Shaan Puri He was one of the OGs of OpenAI. He was then poached by Tesla. He ran the Tesla self-driving program for like a decade or something like that. He went back to OpenAI after that. And then now he’s just like indie hacker. He’s just Peter Levels with like six brains instead of one human brain. And a billion dollars. He’s just fucking around and he just posts on Twitter like random AI little experiments he’s doing. So he made this thing called Karpathy.ai slash jobs. And you can see this on the screen here. So basically what he did is he took all of the U.S. Jobs from the Bureau of Labor. And okay, so that’s 143 million different jobs in the U.S. Economy. And then each, the size of the square is how big of a part of the U.S. Economy that job is. So you could see like registered nurses. There’s 3.4 million registered nurses in the United States. Construction laborers, there’s 1.6 million. Secretaries, 3.5. Sam Parr Okay, so it’s percentage by quantity of workers. Popularity of the job. Because it says like cashiers a huge percentage of that, which is insane. So cashiers take 3.2 million jobs. That’s wild. That’s cool. Okay. Correct. Shaan Puri And so then what is green and red, right? Looks like a stock market chart. So basically then he told AI to estimate the likelihood that this job is going to disappear based on the capabilities of AI. So based on what the job does and then what AI can do, how disruptible is this job? How at risk is this job? So for example, construction laborers are green because they’re more in the safe category from AI because AI today can’t do what those people do. But customer service is in red, obviously. And so he went through every job in the US market. And then you could sort of almost see like how turbulent will this be for the labor market? Because you’ve got how much red are you going to see on the screen? Because if it’s going to really disrupt a very small niche job, that’s not that big of a deal. But if it’s going to disrupt, you know, secretaries and admin, cashiers, I’m surprised waiters and waitresses here, high school teachers, I’m surprised that one’s there, and red. Kindergarten Elementary is like kind of in an orange state here. That’s insane. What? And then he has a caveat. So he says, caveat on digital exposure. So he basically, let me just read the whole thing. So he said, the digital AI exposure option is one example. It estimates how much of current AI, which is primarily digital, will reshape the occupation. But you could do this for any other type of thing, like offshoring risk, et cetera. And then he includes even the prompt here. It’s just a cool little hack project by him on the side. And he’s like, you know, there’s obviously a bunch of caveats, which is he’s like, software developers scored really high while some people believe, well, dude, AI is doing all the Coding now. What does that mean? And he’s basically saying, he says, because AI transforms the work, but the demand for software is going to grow as each developer becomes more productive. The score does not account for demand elasticity, latent demand, regulatory barriers, or social preferences human workers. That’d be like kindergarten teachers, for example. Many high-exposure jobs will be reshaped and not replaced. So it’s not—the red doesn’t mean replaced, but it does mean reshaped. He did another thing that was pretty crazy. (Time 0:35:50)
  • Productize AI Audits Into Scalable Services
    • Build repeatable AI products for industries by auditing one client’s workflows, automating monotonous tasks with prompts, then scale the solution across similar businesses.
    • Sam automated landing pages and a retail tenant tracker as examples that deliver quick ROI and become sellable templates. Transcript: Shaan Puri Yeah, I feel like I have superpowers, but I’m like the early, you know, in Spider-Man, when he like realizes he could shoot webs, but like really all he’s doing is just shoot, he’s like Summoning the Coke can to him. It’s like I could just get up, but or I could just web it to myself. I didn’t have to get up from my bed. Like, that’s kind of what I’m doing. I’m doing a lot of stuff that feels cool to me. It feels like a superpower. But like, does any of it matter right now? No. Like, still, all the useful stuff is just chat with, chat with ChatGPT, chat with Grok, chat with Claude. Sam Parr Dude, I’ve done like really simple stuff where like, I’ll be like, make 20 landing pages in HubSpot for when people search X, Y, and Z that are related to my company. That’s a very monotonous, boring thing that I normally would have had to do by hand. And it just made it all in five minutes. And this is for a lot of people listening. The agency business model just got really interesting now. Historically, agencies are a pain in the ass, but they’re actually pretty amazing because you don’t need to test if there’s demand for an agency, right? Everyone needs an agency. Every business hires a lot of agencies. You go through a lot because you don’t particularly like them, but running an agency has always been a huge pain in the ass, regardless of how great of an agency it is. But now it’s significantly different. I think that if you own an agency, I think you could take the path of like, my business is over, or you could take the path of like, my multiples are going to go up a significant amount. Shaan Puri My buddy Romine sent this blog post to just me and Ben, maybe a year or two years ago. And it was called Service as a Software. So instead of Software as a Service, Service as a Software. And he basically had this thesis. He’s like, service businesses have historically had lower margins and lower multiples because two things. One, because it required so many people to do, the gross margins were worse. And two, because it required skilled people, you couldn’t scale it the way you can, a piece of code that can just keep running anywhere in the world. Whether one person is using that code or 10 million usually doesn’t really matter. And he was like, AI seems to have changed this. And he called this early. And he was like, now services businesses are going to have high gross margins, like 75% gross margins instead of like 50%, 40% gross margins. And he’s like, the reason why is because obviously AI is going to make the service company more efficient. With one person, you can do what, you know, four, five, six, seven people were doing before. And then the second thing is that the more people have software, like service is going to become more rare. It’s going to become more like, it’s going to become the harder thing to get. And in addition to that, you have, so not only does it have the higher multiple, but also it’ll scale better, right? Because if you’re using the more internal software you have, the more clients you might be able to serve. And so you get the scalability and you get the margins. And so he had this idea, and this is exactly what we’re seeing. We were talking to a bunch of PE folks, and they’ve all basically shifted budget away from buying SaaS to buying service companies. And they value the service companies like they used to value software companies, which is kind of crazy as a shift to make for somebody who was like ingrained in my mind that like, you Know, software is the thing and services are bad, you know? And he was 100% right. All right, today’s sponsor is a company that I use that I actually built a company on that I sold for millions of dollars. And it took me zero upfront capital. We had one employee and those are the types of businesses I love. I love lightweight businesses, things that don’t require a lot of capital, don’t require a lot of employees, and you can just get them off the ground quickly. And so Beehive has a platform that lets you launch a newsletter about anything. And the great thing about newsletters is it could just take one person. You’re just writing stuff that you already know and already in And and it can grow and grow and grow. For us, with our crypto newsletter, we built a brand called The Milk Road. And in one year, we went from zero to being the biggest crypto newsletter on earth, and we sold the company for millions of dollars in a year. And the only reason we could do that is because Beehive had all the things we needed out of the box. We needed to be able to write the newsletter, send it, had growth features, referral programs, monetization, everything was built in. And so highly recommend if you’re looking for a lightweight business, something that you can get going very quickly, highly recommend Beehive as a place to start. Go to beehive.com/MFM and you can actually use the code MFM30 and get 30% off your first three months. Let me show you this thing that I made. All right, so I think I’ve told you this before, but my brother-in who’s come on the podcast before, his name’s Sanjeev, he is one of the biggest owners of retail shopping centers in the Country. He owns, I don’t know, like some, I don’t know what the number is, but like maybe like, you know, a couple million square feet of real estate of retail shopping centers in the country. Probably one of the top five to 10 operators. But he, you know, it’s like a blue collar type of business as in like, you don’t, you know, his job is he does a lot of construction. He does a lot of leasing with brick and mortar businesses. And so he’s not hugely exposed to AI. So he, he joined like Tiger 21 or something like that. He was at a meeting and he, and they were, he’s like, dude, it was all about AI. He’s like, I feel like I got to like know more about AI. He’s like, seems like, you know, I’m, I kind of know about it, but I got to know more, you know, like, it seems like I’m probably under investing in this. So last night I just did, I did a little demo for him. So I basically went into Claude and I just go, I’m a retail shopping center owner. This is my business. Tell me 10 ways I should be using AI. So it gives me 10 ways. So it’s like, here’s 10 things you could do. A retail expansion and contraction tracker. Okay, so like, you know, your tenants are these retailers. You could scrape earnings calls, 10Ks, press releases, trade publications, and flag which stores are opening more locations, which ones are closing locations. And that would create a hit list for your team to go call those tenants and basically be like top of mind for them as they look for new locations. Lease expiration. You have all these leases, right? If you have, let’s call it 50 centers that you own and each center has 30 tenants. Now you have 150 tenants that you would have. Each one has a unique lease on it, a unique timetable. You need to know who’s expiring and when. Just upload all of it to a folder and it’ll tell you what is expiring, when it’ll give you notifications. And it’ll cross-reference that with their public announcements. So you know, hey, they had a hot quarter, you know, go and look for an expansion of that lease. They’re contracting. That’s going to be at risk. You should start thinking about possible new tenants to put in. And it gave a list of like, you know, 10 different things. And then I just said, cool, give me the prompt to build this, to build number one, that like expansion tracker. So check this out. So then I go and I basically take that prompt. I throw it into the new, I wanted to try the new Perplexity Computer. I don’t know if you’ve played with this at all, but Perplexity has their… Sam Parr Is Perplexity computer just OpenClaw, but different? Shaan Puri Yeah, it’s like O… (Time 0:40:16)
  • AI Raises Valuation Of Service Businesses
    • AI turns service companies into high-margin, scalable businesses because automation increases productivity and embeds software into delivery.
    • PE shifted to valuing service firms more like software due to improved gross margins and scale from internal tooling. Transcript: Shaan Puri Buddy Romine sent this blog post to just me and Ben, maybe a year or two years ago. And it was called Service as a Software. So instead of Software as a Service, Service as a Software. And he basically had this thesis. He’s like, service businesses have historically had lower margins and lower multiples because two things. One, because it required so many people to do, the gross margins were worse. And two, because it required skilled people, you couldn’t scale it the way you can, a piece of code that can just keep running anywhere in the world. Whether one person is using that code or 10 million usually doesn’t really matter. And he was like, AI seems to have changed this. And he called this early. And he was like, now services businesses are going to have high gross margins, like 75% gross margins instead of like 50%, 40% gross margins. And he’s like, the reason why is because obviously AI is going to make the service company more efficient. With one person, you can do what, you know, four, five, six, seven people were doing before. And then the second thing is that the more people have software, like service is going to become more rare. It’s going to become more like, it’s going to become the harder thing to get. And in addition to that, you have, so not only does it have the higher multiple, but also it’ll scale better, right? Because if you’re using the more internal software you have, the more clients you might be able to serve. And so you get the scalability and you get the margins. And so he had this idea, and this is exactly what we’re seeing. (Time 0:41:40)
  • Sell AI Transformation To Local Businesses
    • Sell AI transformation services by learning AI nights and weekends, auditing businesses, and offering no‑risk implementation tied to measurable value.
    • Sam built a live retail expansion tracker prototype for a landlord using prompts, Perplexity Computer, and retail announcements as signals. Transcript: Shaan Puri All right, so I think I’ve told you this before, but my brother-in who’s come on the podcast before, his name’s Sanjeev, he is one of the biggest owners of retail shopping centers in the Country. He owns, I don’t know, like some, I don’t know what the number is, but like maybe like, you know, a couple million square feet of real estate of retail shopping centers in the country. Probably one of the top five to 10 operators. But he, you know, it’s like a blue collar type of business as in like, you don’t, you know, his job is he does a lot of construction. He does a lot of leasing with brick and mortar businesses. And so he’s not hugely exposed to AI. So he, he joined like Tiger 21 or something like that. He was at a meeting and he, and they were, he’s like, dude, it was all about AI. He’s like, I feel like I got to like know more about AI. He’s like, seems like, you know, I’m, I kind of know about it, but I got to know more, you know, like, it seems like I’m probably under investing in this. So last night I just did, I did a little demo for him. So I basically went into Claude and I just go, I’m a retail shopping center owner. This is my business. Tell me 10 ways I should be using AI. So it gives me 10 ways. So it’s like, here’s 10 things you could do. A retail expansion and contraction tracker. Okay, so like, you know, your tenants are these retailers. You could scrape earnings calls, 10Ks, press releases, trade publications, and flag which stores are opening more locations, which ones are closing locations. And that would create a hit list for your team to go call those tenants and basically be like top of mind for them as they look for new locations. Lease expiration. You have all these leases, right? If you have, let’s call it 50 centers that you own and each center has 30 tenants. Now you have 150 tenants that you would have. Each one has a unique lease on it, a unique timetable. You need to know who’s expiring and when. Just upload all of it to a folder and it’ll tell you what is expiring, when it’ll give you notifications. And it’ll cross-reference that with their public announcements. So you know, hey, they had a hot quarter, you know, go and look for an expansion of that lease. They’re contracting. That’s going to be at risk. You should start thinking about possible new tenants to put in. And it gave a list of like, you know, 10 different things. And then I just said, cool, give me the prompt to build this, to build number one, that like expansion tracker. So check this out. So then I go and I basically take that prompt. I throw it into the new, I wanted to try the new Perplexity Computer. I don’t know if you’ve played with this at all, but Perplexity has their… Sam Parr Is Perplexity computer just OpenClaw, but different? Shaan Puri Yeah, it’s like OpenClaw, but in their sandbox. Less technical work to start, but it’s in their environment. So it built it. So check this out. It’s basically like, check this out. Is this just a mock-up? No, this is like an active live website now that he can use. Sam Parr But this is not his data, or this is his data. Shaan Puri This is not his data. I just said, hey, here’s seven states that I operate a lot in, Texas, Nevada, New Mexico, Utah, whatever. So just, and then I didn’t tell it anything else. So we went in and said, okay, Dollar General is a retailer that’s expanding. You know, they’re going to do 450 new stores. And then you click in and it says they’ve planned this many new stores for 2026. They’re going to do 4,000 remodels, 20 relocations. They currently have 20,000 stores. Here’s where they’re likely to be expanding. (Time 0:44:34)
  • Sell AI Transformation To Local Businesses
    • Learn AI on nights and weekends so you can offer transformation services to business owners.
    • Sell a no-risk audit: promise to find ways AI can improve their business, then demonstrate the value.
    • Implement the solution for one client (e.g., a dentist), then scale the blueprint to dozens of similar businesses.
    • This model can turn into a repeatable, high-margin service that can generate millions by replicating the same audit→implementation flow.
    • Shaan frames this as the modern equivalent of the SMMA playbook: a business-in-a-box anyone can start now. Transcript: Shaan Puri Well, yeah, I think you have to, you have to figure out some proportional investment of education and time to work on this, whether it’s, you know, once a week or you do three weeks or you Hire a consultant or you do whatever you’re gonna do. It’s one of the reasons I think the biggest opportunity right now, like the obvious no brainer thing that anybody who hasn’t made it, that wants to make it in life needs to do today is to Basically go sell AI transformation. So what do you do? You go find business owners and operators and you go, you become an AI expert in your nights and weekends, basically. And then you go to them and you sell a no risk offer to them, which is that I will come and I’ll find ways. I will come and do an audit and basically figure out how I can use AI to improve your business. And then you go and you do an audit and you try to figure out how to improve their business. You sell them the solution, the implementation of it. Of, hey, this is what I think AI could do for you. I’ll do it for you if you pay me this amount. And here’s the value it’s going to drive for you. And then you do it once with, let’s say you do it with one dentist, then you go do it with 50 more dentists. And you can literally make millions of dollars doing just that blueprint. (Time 0:49:30)
  • Sell AI Transformation To Local Businesses
    • Learn AI skills on nights and weekends until you can audit businesses for AI opportunities.
    • Offer a no‑risk audit: find ways AI can improve operations, quantify value, then propose implementation.
    • Start with one client (e.g., a dentist), prove results, then replicate the blueprint across dozens to scale revenue.
    • Position this like the early SMMA playbook: a repeatable, teachable service businesses need but don’t know how to adopt.
    • This approach can become a simple, high-margin business that can generate millions by rolling out the same solution to similar customers. Transcript: Shaan Puri Well, yeah, I think you have to, you have to figure out some proportional investment of education and time to work on this, whether it’s, you know, once a week or you do three weeks or you Hire a consultant or you do whatever you’re gonna do. It’s one of the reasons I think the biggest opportunity right now, like the obvious no brainer thing that anybody who hasn’t made it, that wants to make it in life needs to do today is to Basically go sell AI transformation. So what do you do? You go find business owners and operators and you go, you become an AI expert in your nights and weekends, basically. And then you go to them and you sell a no risk offer to them, which is that I will come and I’ll find ways. I will come and do an audit and basically figure out how I can use AI to improve your business. And then you go and you do an audit and you try to figure out how to improve their business. You sell them the solution, the implementation of it. Of, hey, this is what I think AI could do for you. I’ll do it for you if you pay me this amount. And here’s the value it’s going to drive for you. And then you do it once with, let’s say you do it with one dentist, then you go do it with 50 more dentists. And you can literally make millions of dollars doing just that blueprint. That is the blueprint. (Time 0:49:30)