Transcript · November 3, 2026 · 11.5K words
Ep. 89: How an Investor in 300+ Companies Picks What to Build With AI
Captions uploaded with the episode. Expect the odd misheard word. Tap any timestamp to play the episode from there, or go back to the episode for the summary, receipts and chapters.
0:00Meet Anthony Pompliano
Play from 0:11 What's up, everybody? Fast Hours Podcast, episode 89. Drew Brucker, Rory Flynn, special guest today, Anthony Pompliano. Anthony, welcome to the show, man. It's great to have you here. ~ I'm gonna go ahead and dive right in and try to do my ChatGPT background on you, and we'll just see what of this is true and if we got this right. But it's it's such a fascinating background. So you were an army infantry sergeant, you were in Iraq, you pivoted that into tech.
Play from 0:39 You worked as a growth product manager at Facebook on pages, on a team, and you worked on a team that handled Zuck's social media presence or Facebook presence, maybe even Cheryl Sandberg's according to this. then you kind of moved into the crypto VC side, you co-founded Full Tilt Full Tilt Capital, Morgan Creek Digital, and then obviously became like a real loud TV voice in Bitcoin bull. ~ and says you've invested in over 300 private companies now, you now run.
Play from 1:07 Your own podcast called the Pomp Podcast, the Pomp newsletter, which is called the Pomp Letter. And then you've got millions of followers across X, YouTube, whatever you want to call it. You've published your second book as well. now you're the chairman and CEO of Silvia which is something we're gonna get into today. But Silvia is an AI CFO for independent investors. And there's some really interesting nuggets there and too, in terms of how you've set up the people on that team or the non-people on that team as well.
Play from 1:35 So How was that? pretty good. You you want a job? You you can come over here and just keep saying that over and over. There we go, man. Hype man. it's good to have you on the show, dude. Look, we you know, for the for the audience that's listening, just to give everyone a heads up, look, you know, obviously we talk about a lot of things on the creative side. And I think, you know, Anthony has a a way different perspective and a way different experience in a lot of the d the market sectors that maybe we don't have insight into.
Play from 2:01 So we've de we wanted to get in here and really like dive in on a few things that we thought, you know, especially on something like Silvia, right? Like that's that's an interesting one We talk about like I said, a lot about the creative tools, but not on the finance side. from a from a general ~ overview of the entire market, you basically using AI as ~ a predictive analytics computer thing, something it's very good at would be something along the lines of ~ math and finance.
Play from 2:26 Right. So I just like to just to kick this one off, man. I I think getting this getting into this, ~ why was it why was it that? Like what what Why was it like, we're going, ~ I don't want to say open source, but custom trained in that sense to to build something that would be, ~ this product for this market, and and how did that come about? Yeah, so story of Silvia is pretty interesting in that ~ I wanted it for myself, w which ~ every business I've ever been involved with that ended up actually being valuable is started with somebody, whether it was me, somebody else, whatever had had a problem.
2:47The Personal Problem Behind Silvia AI
Play from 3:01 And the reason why I wanted this was all the fintech apps, I I would try them and I just never use them. And the reason was because they were all focused on budgeting and saving. And, you know, I've been very fortunate in my life where ~ five dollars saved on Starbucks every day doesn't change my life in any, you know, material way. And so it was like, okay, well, what if I can use this technology to do what I want to do, which is from an investor perspective, grow my assets? So not a contraction or savings and budgeting. I want to grow.
Play from 3:27 And I don't really think that like hiring some financial advisor is going to be the solution. Or I don't think that, you know, going and getting a bunch of these like external experts. Instead, I think that I know just enough to be dangerous, you know, in terms of allocating my own money. So I used to have this spreadsheet. Every single tab was a different month. And I basically would just go in and manually update it ~ and track my ~ wealth over time. And I said to ~ Shane, who is my co-founder the business, ~ who was the best engineer in any of our companies, I said, dude, could you just build a dynamic dashboard?
Play from 3:58 ~ it had nothing to do with AI at first. It was literally just like, I'm just tired of the spreadsheet. And ~ it was really his idea of like, well, what if we started to incorporate some of the AI stuff? And as soon as I saw it, I was like, my God, I could now. I can talk to the AI with the context of my portfolio, ~ you know, eye-opening. And so long story short, is ~ that's pretty much what we launched with. It was ~ for all intents and purposes, a ChatGPT wrapper, right? I I don't like hide from that in any form or fashion. I actually think that's where a lot of these products start, is we were outsourcing the intelligence to the Frontier Labs and we were focused on building the software and the product and the interface, et cetera.
Play from 4:32 Now we ended up switching from ChatGPT to Claude. Claude kind of had a moment there where I think everyone got really excited about, you know, it was getting ahead. and and we saw that too. But then then the bills came. And when the bills came, ~ that did not feel as great. and so I've publicly talked about ~ the peak, I think was in if I remember correctly, March of 2026. We were paying I think the bill was $550,000 for the mark.
4:46The $550,000 AI Bill That Forced a Rethink
Play from 4:57 Yeah. Now you gotta remember, this is a user generated platform, meaning that like every time I tweeted and told people about Silvia, people would go sign up and they would start hammering on it. And so ~ it became this weird thing of I don't want to tweet because it's like cost me money, but I also want more people using the platform. And so it became clear hey, we gotta fix it. Now, what I will say is start with maybe one point, which is there was a bunch of things that we could do inside the product architecturally, so that we just didn't hit the model as often, right?
Play from 5:26 We killed a feature that had been running that we took it away from people and we were like, well, let's see if anyone complains. No one complained. Well, guess what? We guess we don't need that feature, right? ~ when you refresh some pages, we were like hitting the model. Let's stop doing that. Right. So we just went into like optimization mode. But the real zero to one was going to be like, how do we get off the frontier models? And in you know, Q two of this year, talking about getting off the frontier models, like you might as well just, you know, put on a clown mask and go dance on the street because everyone thinks you're insane.
Play from 5:51 Right. You're like, what do mean you're gonna get off the frontier models? And we have a very good ~ engineering team. And they were like, well, we don't know if that's actually possible, but we do think that we could like fine-tune some models and stuff. Let's just like go see how we can do this. Long story short, as you fast forward to today, ~ we have been able to ~ take a material percentage of our compute cost and shift it to our own hardware ~ and and kind of AI stack, which I'll talk about in a second. And I think that over time.
6:17Inside Silvia's Custom AI Stack
Play from 6:18 ~ we're gonna be able to get rid of the frontier models. And we're gonna be able to have, you know, exclusively ~ an AI stack for Silvia. Now, why is that ~ interesting? If you think of the AI stack that we have today, my description of it may be a little bit different than some of the engineers, but the way that I describe it is there's basically four components to the AI stack. The first is we have a proprietary AI harness, the harness kind of air traffic control. It's doing everything from we built a custom model router, we have all the tool calling, we've got all of the agent orchestration, kind of all the stuff you would expect in a harness.
Play from 6:48 The second layer I think about is the data pipelines. So tax as an example, we went and we indexed the federal tax codes, state by state tax codes, and then we put it into a database. We created the schema and architecture that our agents uniquely understand, able to retrieve and have memory with. And so when you ask a tax question on Silvia, every AI product has what the model's been trained on plus web search. We now can also reference our data pipelines. And so you just increase accuracy, re reduce latency, et cetera.
Play from 7:14 So you've got harness, then you've got data pipelines. Third step is we fine-tune models. And so we have our own proprietary models there that are trained for our specific environment, our specific use case. So going really after specialized workflows and proprietary, intelligence. And then the last step is we built a custom coding agent. Now, the reason why a coding agent for our specific use case is interesting. Most people, when they hear coding agent, they think about competing against clawed code or codecs or something like that.
Play from 7:40 That's not what we're talking about. Our engineering team still uses the off-the-shelf coding to write code. What we are specifically talking about is in the personal finance use case, there's a lot of edge cases. So to answer questions, many of our questions get routed to a custom sandbox where we are writing custom Python code, et cetera, to answer the question. So the example I will use is somebody wants to run a Monte Carlo simulation. We're writing code in the pursuit of answering that question to run the Monte Carlo simulation.
Play from 8:10 You can imagine if you're using a Frontier model to write that code, it's incredibly expensive. And then of course it is less accurate, it's higher latency, et cetera. So harness data, model, and then the coding agent, you put all that stuff together. When we process tokens on, let's say, Claude, versus if we do it on our hardware and software, it's a 97% cost savings. It's like a no-brainer. Now, that is good for me for as a company.
Play from 8:38 Everyone wants to spend less money and get the intelligence, but the user does not care. The user didn't give a shit what we're paying, right? What the user cares about is, hey, are you improving the experience for me? So the more important port, the more important part for the user is that by shifting over to our own intelligence stack, it is more accurate answers and lower latency. So they're getting faster, better answers. At the same time, we're getting lower costs. So it's a win-win for both. And so that really shifted us into that world.
Play from 9:05 And so if you zoom out, it's like, okay, if we can get to the point, right? We think we can, but let's see. Where we rip out the frontier models and we basically have a proprietary AI stack that is built for personal finance, and it is more accurate than the frontier models at answering questions like tax, mortgage, credit cards, et cetera. And it is significantly cheaper. Then, from a strategy standpoint, well, what are the use cases of intelligence that were previously cost prohibitive that now are on the table because you've been able to drop that cost so much?
Play from 9:35 We got a bunch of ideas. A lot of work ahead of us, but that's what I'm interested in, right? Is like by the shifting over to our own intelligence, now you start to open up the things that you can do that you couldn't otherwise do if somebody else was controlling the intelligence. Yeah, I like that. I'm I'm curious, how many How many people do you have working on this versus agents at this point? I feel like you probably have more agents than people. if I remember ~ correctly, ~ we have a total full-time employees at the public company of 10 employees right now, including myself.
9:5610 People, Dozens of AI Agents
Play from 10:08 So, like you think about that, you're like, that's not that many people, right? Now, we have I don't know, tens of 50, 60, 70, whatever agents. So we have more agents than people. I think that's kind of normal for a lot of companies at this point. But what I think is maybe most interesting is How we have chosen to build some of this stuff. So we definitely have some agents that are like, I use Grokbot as an example to do a whole bunch of stuff professionally across all of our different businesses.
Play from 10:36 Each one of those agents is like a dedicated agent for that specific task, right? But what I think becomes more interesting inside of the company is I'll give you a concrete example around ~ ticket support. So obviously we got a, ~ what is a fast growing number of users. And so they constantly have, feedback, issues, problems, bugs, whatever. When one of those tickets gets submitted, it goes into a Slack channel that is Silvia support.
10:48Toby: AI Support With Human Approval
Play from 11:03 And in that Slack channel, we have an agent. That agent's name is Toby. And Toby will go and say, okay, what is this issue? It will then go and do a bunch of checks to identify it, ~ investigate it, recreate it. It will then also identify what are the potential solutions. And then it might go as far as to actually write the code to solve the problem. And then it will sh put in a PR or go and send it to a human engineer to go and approve it.
Play from 11:31 Now, the process of ticket being submitted to Toby being able to actually identify issues and potentially write the code, less than two minutes. Whereas previously a human was having to go and ~ go figure it out and all the stuff, right? So that is like one example. And then you take that and you start to apply to every single aspect of the business and you're like, okay, humans are not going away. But man, we would not be able to do what we're doing with 10 employees if we had no AI.
Play from 11:59 Right. And so like that, I think is the the interesting thing is like, how do you mash humans plus agents together so that they it's like one plus one equals three? And that's what hopefully, you know, we're we're kind of on the cutting edge of of figuring out. It's crazy. we just think about before AI, there were so many obstacles or roadblocks you'd run into for ideas that you would have or things that you wanted to do. Right. ~ from a career standpoint, it was hard to break into another field if you didn't have a degree from that or you had direct experience with an employer.
Play from 12:28 Right. But now you're seeing even with people that are doing this full time, even people like us, right? Where it's like, I wasn't necessarily a creative director, but now I'm doing creative director type roles or I'm wasn't doing consulting before. But it's all about taking advantage of the the democratization, if you will, that's taken place there. It's really fascinating. I do want to ask you, going back to that experience standpoint. this is an interesting question, but ~ if if somebody else, like a major lab were to copy one of your best features over there at Silvia, or or do like why what would what about that experience?
Play from 13:00 What about your platform would keep somebody there? Right. Because it's also like, you know, now it's it's it's always like this thing too, where man, the moats are drying up everywhere. There maybe are a few left, but they're always the, you know, they're very shallow. And what about maybe something like that would keep customers paying you versus maybe somebody else and not worrying about somebody else eating your lunch in that that capacity? We're the R D department for these ~ companies, these trillion dollar companies.
13:23Competing When Big AI Labs Copy Your Features
Play from 13:27 We launched stuff they fucking copy 100%. good good luck keeping up though. Good luck. What do I mean by that? Well, we're recording this on a Thursday. Four days of the week. We've launched four separate products or features so far this week. Hopefully we'll go five for five tomorrow. And these aren't like small things, right? Today we announced something called Silvia Legacy. The whole idea is Well, what happens if you pass away or something happens, you're incapacitated?
Play from 13:55 Your spouse, your loved one, your child, your somebody that it could be an attorney, whatever. They want to be able to understand where are your assets. And so we have an entire flow in the product now where you can actually delegate to that person, hey, this is the person I want to get the information. But then you have to think about control. And so who gets it? When do they get it? What information do they get? How does it get delivered? All this stuff. And so we built an entire product around basically for me.
Play from 14:22 I picked my wife, right? And I said, okay, if I'm not active on the platform for 60 days, it will then send me a notification, both email and text, every single week for three weeks. If I still do not engage with the product, it will then wait, I think, seven or 14 more days. And then it will send her all of my information that I've put into Silvia. Well, then you have a question of does she just get the assets and the accounts? It's like where is the money? right check why should we chat history.
Play from 14:52 One, I think it'll entertain my wife because she'll think it's funny to read through all my chats with the with the AI, right? But probably more importantly is then she can hopefully try to deduce why did I own certain things? Why did I make certain decisions? So all of that is putting the user in control, but it is trying to make what is, you know, usually a horrific situation, chaos, uncertainty, et cetera, a little bit easier on loved ones, et cetera. That's one thing we launched this week.
Play from 15:18 Remember, we have 10 people at the company. And so we've just done this over and over and over again. Now, the reason why I say that is the single biggest moat I personally think, speed. It is impossible to keep up with a company that is iterating quickly and continuing to improve their product. Now, the second thing that I think is a big moat is focus. And the reason why I say focus is if you think about why our AI tech stack is outperforming the frontier models.
Play from 15:43 They're pursuing general intelligence. We're pursuing expertise in a specific vertical. And so just naturally those two things are at odds with each other. And so the way that you create the mode or you extend away from the frontier ~ models is you basically have to figure out what are the data pipelines, what are the training, you know, kind of strategies, how are you going to build more accurate, more expertise than them?
Play from 16:09 Focus is incredibly helpful in that pursuit. That's a that's a great point on the focus side. Cause we see it all the time. We see ~ when these frontier models launch stuff, you're like, for what? it's just it feels like a wasted a wasted release or just something that you're just like, this is novel. Did you just do this for a press release? 'cause it doesn't feel like it has anything to do with the rest of the company. But no, it's it's interesting. Even like that set like that second brain sort of modeling there.
Play from 16:34 If you're sending that to someone, it also feels like kind of like a licensable product. Like, you know, here's Warren Brand, you Warren, Warren Buffett's Chat history, that seems like an entire ~ digital medium in itself that could be utilized just like randomly, hypothetically. Just it's we it's weird where we are with that stuff. You know, like someone's entire personal finance decision making process could be synthesized into code and then transferred to someone who could essentially like your wife could essentially take the same patterns that you've been making and like continue it if you weren't there.
Play from 17:06 Right. Is that is that like the is that sort of like the idea too is to like keep it going if something like that happens? Yeah, you did you definitely could. look, it it's this ~ it's this very interesting question of how much of a product just to use Silvia, you keep keep at using that as the example. How much of the ~ solution is product oriented versus ~ like hardcore AI tech oriented? And what I mean by that is like obviously raw intelligence or expertise is a part of the solution.
17:12Building AI Products People Want to Keep Using
Play from 17:35 But there's a lot of things like Silvia Legacy that has nothing to do with AI. That is a software product and you're thinking about design, you're thinking about hu user psychology, you're thinking about life events, you're you're trying to better understand who is my user and what is the problems or issues that they face. And so you extend that out across the product. Another example is we have a feature ~ that is a one click generation of a personal financial statement. Anyone who's ever worked with a bank, whether you borrowed money, you work a private bank, whatever, when they ask you for that personal financial statement, it's like, ~ boy.
Play from 18:06 Let me go get a cup of coffee. Let me sit down, get out my little pen. Let me let me go figure this out, right? Yes. So if Silvia has all your assets, you should just be able to generate the personal financial statement in one click. AI really doesn't touch that. Like that's not an AI problem, right? It is a software problem. It's an understanding of the user problem. And so when you start to think through this, again, it goes back to just because a team can go and pursue AI technology or like general intelligence.
Play from 18:36 It doesn't mean that you're going to build the right product for the user base. A third thing that I'll use, and probably you guys of all people would appreciate this, who you are trying to solve a problem for can really set an incredible amount of downstream issues. What do I mean by that? Most fintech products are built for budgeting and saving. It is the largest adjustable market. It has the lowest barrier to entry, anyone, regardless of how much money you have, budgeting and saving.
Play from 19:05 And then most engineers don't have a lot of money when they're younger. So they're solving a problem that they intimately understand, right? So it makes sense why a lot of that coalesces around that vertical. The problem we're trying to solve is not budgeting and saving. We are trying to help people who already have assets. Our average user is a multimillionaire. Grow their assets. It is fundamentally a different product, a different person, a different solution. And so when you start to make decisions, it's not just like budgeting and saving versus growing your investments.
Play from 19:35 It's also like, what is the psychology? So when you're budgeting and saving, it's all about ~ put more money in your pocket. When you're talking to an investor, maybe build generational wealth. Similar, but different. And so you start thinking about design. Right. Yeah. I know you guys are like ~ have talked a lot about the creative stuff. It's like, are you building a luxury positioned product, even though it's software at a low price point? Because that is with the user base.
Play from 20:01 I don't I I I talk a lot about budgeting and saving apps. You could go to any mall and pretty much that's where your users are. Right. You could go to any sporting event, there's all your users are sitting there. For us, ~ maybe golf and tennis. Rolexes, sports cars, that's where our audience is. And so it's like a little bit higher up on the the luxury positioning, but you don't want to abandon the other group.
Play from 20:27 So how do start thinking about creative and and all these things that I don't have great answers for? we're we're actively thinking through it, but it starts to really shape around where does value get created in a business? And how much of it is just like, or AI is better than your AI, versus no, you're built, you're crafting a product. Right. And the product experience is just as important, if maybe not more important for the user. As somebody who's kind of built their own thing with with AI as a tool here, like if someone listening to this builds an AI product this weekend, what would make what would make it something that you would actually entertain investing in?
Play from 21:04 what what would be the things it would need to have or be doing or intentionally positioned for you to take it seriously in that regard. Let me answer ~ maybe with something that isn't related to what I'm like ~ officially doing on a day-to-day basis. I have built ~ a number of products for my own personal use. So I've built personal software, if you will. One of them ~ is like a like a HQ or like a command center.
21:13The AI Command Center Pomp Built for Himself
Play from 21:32 And so it has a to-do list, it has what other people owe me, it has a bunch of our business information, it has my calendar integration as a CRM, like all of this stuff where I can basically operate. On a day-to-day basis in this software. If I really wanted to, I probably could figure out how do you productize it and go bring it to other people. I'm not very passionate about it. Frankly, I just built it because I needed to solve some problems for myself.
Play from 21:57 As I was building that, the single most important thing to me was iterating the product rather than drafting up some massive product scope. So 10, 12 years ago, I was a product manager at Facebook. And what we used to do is we used to sit down, ~ product spec documents was a thing. you would ~ draw it all out. We were still pretty fast. this Facebook, right? we were very data driven.
22:11Building With AI Through Fast Iteration
Play from 22:23 We we were able to, ~ get moving. But now what I did is I basically said, hey, this is the general direction. I just mind dumped. It built something, it wasn't very good. And then I just started using it. Say, hey, can it do this? Can you do this? Screenshot this, fix this problem, blah, blah, blah. And just quickly, you just iterated, iterated, iterated. That velocity of shipping or improvement is to me actually almost more important than what the actual features are or what the thing is.
Play from 22:50 So one of the things I've said is I'm not a very ~ avid public market investor. I've got, you public some public stocks right, but don't day trade. I don't spend a lot of time evaluating individual stocks other than, I own certain companies because I either like the founder or I, you I believe in the company mission or whatever. And I you maybe I'm ashamed to say a lot of times I buy the stock and don't even look at like what the financials are. I'm like, if it goes to zero, I I I I'm down for the ride and ~ Yeah, I'm mission driven, right?
Play from 23:17 I I'm I'm in. Like Elon, I don't even care what Tesla's revenue is, I'm in, man. Dude, let's let's go. I'm I'm ~ it's my philanthropic gift to Elon. Hopefully he makes me some money, right? ~ but but ~ If you think about public companies, I actually think one of the aspects that would be most indicative of what companies get disrupted by AI versus which ones benefit is just if you could put a quantitative measurement on what's their shipping speed. So if you could somehow go into these companies and ~ measure the velocity of their GitHub commits, ~ per employee or per engineer, I bet that that would actually be a really interesting data point that would be predictive of their future ability to succeed, right?
23:29Shipping Speed as an AI Investing Signal
Play from 23:59 I would be very, very interested in that. Because we we've we've talked about this. We've ~ poked fun at it. We both used to be in-house. I spent like a decade in tech, mostly startups, but in in SaaS specifically, it just doesn't seem like people are moving fast enough. Right. Like when when you and I and and Rory and like we can like even with the examples you brought up with Silvia can can launch something every day.
Play from 24:28 There's just ~ so much subjectivity and these committees that are a part of the process that it just continues to slow teams down. I just don't feel like these businesses are moving fast, especially SaaS. Because especially PLG side where it's like Calendly, right? ~ do I need to pay $150 a year for Calendly or can I just build my own? Well, shit, I just built my own and now I use that, right? There's that side of it, but even from the broader perspective, going after mid-market enterprise, they're not shipping fast enough.
Play from 24:59 They're not shipping fast enough. And oftentimes some of the product releases I'm seeing now, it's like, I've me or Rory have talked about like that, or we've even solved that like weeks ago. Like it's just so baffling to me, like how that's taken place. Are you seeing that same thing? Or what's your what's your take on that? I 100% agree. Now, Calendly is an interesting one. I am probably at a point in my career where ~ I don't use Calendly, but let let's say I did.
25:16Should You Build It or Keep Paying for Software?
Play from 25:26 I'm probably not gonna go rebuild a separate, you know, kind of better version to save 150 bucks. Now, if it's for some reason I can't do something that I wanna do and it's like becoming a a like a friction point, okay, I could be that. But there's like there's some threshold dollar wise. Where I would say, okay, it's not worth my time, even a couple of hours to save 150 bucks a year. Now, with the other side of it though, is if there was some bug that kept like screwing up how the calendar, you know, was being sent, or I was mi I missed meetings or something.
Play from 25:58 Like I don't care how much time it takes, right? It's like missing the meeting is so painful that I'll go solve it. So I think people are trying to like make this like trade-off calculation of just because you can build something, is it worth doing it? Right. And ~ and again it goes back to if, ~ I I don't want to pick on Calendly, but just any company, if they're iterating and improving their product at this like ridiculous speed, then you're kind of like betting on them as a user that it's going to keep getting better and better.
Play from 26:24 And so like you're willing to stick with them versus a company that is degrading. So again, I it's no perfect like quantitative, you know, formula for this stuff, but I do think there's some interesting ~ decision everyone comes on of like I could build this, but do I want to or is it worth it? Now, where I find for myself building the most amount of things is one of two things. I want a piece of pro like a piece of software or product that does not exist, or at least I have not found one.
26:38The AI Projects Pomp Makes Time For
Play from 26:51 And so I need to build it. The ~ Pomp Command Center or whatever whatever you want to call it, that is a perfect example. But also, I have built a number of websites that are ~ intellectually like fun to do because you're using the technology and you're building and all this kind stuff. But They may have some downstream effects. So for example, ~ I built an entire website for the pomp podcast, pomppodcast.com. And when you go there, what you will find is a bunch of articles.
27:06Turning Podcast Episodes Into Articles and Clips
Play from 27:18 But what I am doing is I am taking every podcast episode, and then I am basically finding each individual topic, right? I am then writing an entire story that is very SEO optimized, et cetera. And then we are using another bot to go and clip the podcast. To find the exact thing, then embed the video. you go through all these different things, right? And I don't think that this is going to be some massive effort that, like, you know, we're gonna hire people for.
Play from 27:44 I don't even know if it's actually gonna be that crazy successful. But you could see right here, here's a Silvia, you know, sign up at the bottom of every article. Sign up for Silvia, right? There's, you know, all of these little things. And also, by the way, I got better at using all of this technology by building this. This took me two or three hours on a weekend. And it was like intellectually stimulating, right? But then it's like, can can we find traffic source to drive to the things that I'm like are my day job?
Play from 28:14 But also, you know what? It's nice for the podcast audience. if they want to go and and they don't want to watch the podcast or they want to learn more, get context around a topic that was discussed on the podcast, hey, here you go. Can we keep track of all the guests? Can we treat keep track of all the companies we've talked about on the podcast? all this kind of stuff. So if you bring it back to this idea of building stuff. I don't know, man. It is that like a really painful problem that I had? Not really. Right.
Play from 28:40 Is it like a nice to have? Absolutely. And I joke sometimes with with my friends that ~ you know, George Bush and Hunter Biden both became painters. And that was like their creative outlet, right? For their for for their their brains. Maybe this is like my version of like I I like enjoyed doing this. It was kind of like a nice, you know, three hour project. But I'm not confused. I'm not gonna go, all of a sudden have a ~ a company where we'll create you a website for your podcast or something, right?
Play from 29:11 It it's like a one in one. But but these do the these do have value, right? Because it's like at least you know what's possible, what raw walls you run into just by by putting in the reps, which ultimately matters, right? And and maybe it's like There has to be some sort of value to the rep, even if it is like I I think for me, going back to that Calendly example, when I did that, it was such an early first build thing for me that cause this was, I don't know, over a year ago, right?
Play from 29:38 It was just like, I want to see if I could do it, And then it was like, what problems would I run into in doing it? and then having gone through that process from start to finish, right? Just the reps, the experience that you build within that is great context to have in the future. Right, in terms of what's possible and how you went about it. But Rory, I I'm I'm over here laughing, right? Because ~ Anthony, we should trade notes because we just we just threw up ~ our podcast website literally what, Rory, two weeks in the same exact way.
Play from 30:06 Yeah. looks sick. This is ~ it's done in a very similar similar way. all the episodes. Yep. But like this is you you guys ever heard ~ I think it's Rick Rubin, in what's the book? Creative Act. He I think he talks about this idea of like ~ when an idea's time has come, it will find a vessel into the world. And I've always thought that was like a pretty interesting way to to think about it because maybe now when the barrier to building drops.
Play from 30:38 People are like, hey, I have this asset in the podcast. I should also use that content over here. You guys did it. I did it. We we didn't talk about it. If two of us did it, I bet you a lot of people are doing it, right? And so it's like it's the idea's time has come. And so then it's like, okay, well, what else can you do like that? Right. ~ I I think is a a very interesting heuristic to overlay on the world as we kind of experience change when new technology is coming out.
Play from 31:05 I love this the the personal asset side of things, like the website, right? ~ there's something there's something addictive about it when you get in there because it's yours and it all comes from you. ~ the all of our stuff, it's all transcripts. It's everything is just created from what we've already done, right? And then you're just like, what else can I do? and every day it's like, what else what else can I add to this? What other random, agent teams can I spin up to think of crazy ideas for this? And c can it go build Cloudflare agents to work in the background and track, what AI tools are, hitting us for AEO or actually training data.
Play from 31:38 now we have a we have like a log, so we're seeing, you know, every day when we get, you know, twenty thousand hits from Meta that's training on our website. Right. Like we yeah, scraping. Like we know what's going on in the background. It's like that wasn't an idea I had. That was a agent team idea, which I find to be just like you know, it's it's fun, it's interesting, you know, 'cause you you're not doing things that you would normally do. It's like outside of your scope. But like, you know, from that from that standpoint, it sounds like, you know, probably on your personal time. You're you're doing some you're doing some tests, you're doing some other stuff.
Play from 32:06 what where do you find yourself when you have, let's just say you have thirty minutes ~ on a night or something like that. Are you working on stuff for the podcast? Are you trying to build new tools or ~ dream up new features? what's a what's a random Grok bot session, look like for you? Yeah. ~ I would say let's use your example of at night. I got basically three things that I'm doing. I ~ have learned in my career peop people, I'm a menace at night because now all of a sudden no one's calling me and I'm on offense.
32:24Pomp's Late-Night AI Building Habit
Play from 32:39 ~ man. Yeah. I work with people, they're like, you're calling me at 10 30 at night. What the you gotta chill, right? I last night we're talking about something and ~ I knew the CEO of the company that we were thinking about working with, and I just sent him an email. I said, Hey, can we jump on a call at some point to talk about this? He responded in two minutes. He said, Yeah, call me right now if you're free. It's like now maybe nine forty five at night. I called him, we talked for 20 minutes, right? Then he put me on text with his COO and I told her, You call me whenever you're free.
Play from 33:06 Right. And she was like, tomorrow, you can call me right now. I don't care. And so the calls is is a big thing. The second thing I would say is ~ a lot of email. So, I try to do as much email during the day as I possibly can, but I do a lot of it at at night. And then the third thing is I spent a lot of time in Grokbot. Gro Grokbot is probably the thing that if you said to me when I'm not like offense or or defense on communication with people.
33:16Using Multiple AI Agents to Find Opportunities
Play from 33:32 And what I think I have dialed in pretty well is how do you in a multi, you know, kind of agent way just produce? And that production is across a lot of stuff. So How do you do analysis on growth? How do you do analysis on SEO? How do you do analysis on product? How do you do it on, retention? How do you do it on creative? How do you do so? It's like it's ingesting tons of information, identifying opportunities.
Play from 33:59 And I think of it almost like how the tide goes in and tide goes out? We're ingesting all this information, and then it's like, watch out, and then here we go. And then here comes all of the the the the insights back of go do all this stuff. you get too much. How do you how do you handle when you get too much though? You're like, I got 10 great ideas here. I have probably time for two, right? It's like, how the what am I gonna do?
Play from 34:25 But Grokbot is pretty crazy. I mean, ~ I I like the word menace, by the way. ~ late night. I mean, because because, dude, you're okay, you're a dad like us. I think you've got four kids. Am I correct on that? Correct. So I mean, it's like, you know, there's a certain Part of the day that it's just like you got to be a dad too, right? And so, you know, how do you sort of sprint during the day when you got the kids at school?
Play from 34:50 You know, you got kids back home, right? Be a dad, kids go to bed, sprint again, right? So it's almost like you're running two different races the same day every day, trying to fit as much in. And it kind of feels like, I don't know, at least from my perspective, it just feels like we're in this very weird window of time where you almost want to maximize all the time that you have as long as you have the energy. energy to entertain it. ~ do you ever get burned out in a sense, right?
Play from 35:17 Cause we've talked about that too. ~ I think I myself have probably had more little one day or two day bouts of burnout, so to speak, over the last two, three years than I had prior in my entire career. Like are are you feeling that from AI and just interacting and do you know having all these ideas flood you and reading all the the output that comes back from the LLMs, right? It's, it's, it's a lot. Think that it's a burnout.
Play from 35:43 I ~ the reason why I I jokingly like I'm a menace when when I have ~ any moment of free time is ~ dude, I love this stuff. Like on the weekend I'm building that thing because I like genuinely like, yeah, this is fun, right? You know, I got friends that go play golf, they do this, whatever. Like, I I don't do any of that stuff, right? ~ I pretty much hang out with my family and then I do something with my fingers on a keyboard.
Play from 36:09 And I know that that's probably the least healthy thing in the world, but I don't know. Maybe our ancestors ~ went to the bar, smoked cigars, and fucking chilled with their buddies. I, whatever, right? And so ~ from from that perspective, yeah, yeah. They were like the human grock bots, right? It's just like they all meet up. ~ but but no, in terms of ~ with my family, I am very intentional about when and how.
36:28Protecting Family Time When You Love Building
Play from 36:37 I hang out with them, which I think is ~ it's kind of like having like a good diet, right? There's like a lot of ways. So, like some people are like fad diets. So whatever the coolest and the latest thing is, then it never like sticks. Other people, they think about like what can I not eat? So they spend all day like, I don't don't eat the ice cream, don't eat the ice cream, don't eat the ice cream, right? And then like they eat the ice cream. But if instead you approach it and you say, you know what? I'm actually gonna think about what do I eat?
Play from 37:04 I eat healthy. they identify with it, right? James Clear, atomic habits, I'm a healthy eater, right? Then it's like I love eating steak and chicken and broccoli and blah blah blah, whatever. they're dr they're running to something rather than running away from something. Well, the same thing happens in in my mind when it comes to how do you spend your time with family, friends, etc. It's like there are certain moments in the day, usually in the evening, where I have it carved out, I am I am in my dad hours.
Play from 37:31 ~ I I want to spend time with my kids, right? And then on the weekends. Saturday is family day. Sunday, I work. Now, do I do work on Saturday? 100%. Do I do as much work as I could? No. Do I do probably more than I want to? Yes. But ~ my kids know Saturday, it's game time, right? Like, let's go. What do you guys want to do? Sunday, especially Sunday morning, they know they're coming to the office with me.
Play from 38:01 Right. And ~ we've made a whole game out of it. I have a bunch of stuff and they, drawing and coloring, blah, blah, blah all this and I'm integrating it together. Is that the best thing to do? I have no clue. But what I know is it works for me, it works for my kids, it works for my wife. And that's how we're choosing to live our life. And so I I think it's less of there's one prescriptive way to do stuff. And it's more just ~ are you intentional? And are you and whoever else in is involved all signing up for this structure?
Play from 38:28 And if that's the case, then knock yourself out. Do it do whatever you want, Yeah. I think there's there's that I definitely feel this, right? I feel like this this anxiety, especially with the family and everything, where I'm like, we're we're in a period where this might be, ~ the first ~ generational change opportunity, if that makes sense. we have we're we're ahead of it. We've been ahead of it for years. We've been basically one step ahead of every client and every person I've come across, in most of these spaces.
Play from 38:53 It's like you feel like you have a limited you you have a limited time because of the pace of everything. Where it's like I have to I have to capitalize right now. But it's also like it's hard to take a step back sometimes and be like, ~ can this can my little whisper flow message to to Grokbot wait until tonight? It probably can, but there is this this weight on you to get it out and go build something and keep it moving. Cause look, as you said, ~ velocity is the I think feel like we all know that.
Play from 39:22 We all feel it. So that's why it's like this constant go, go, go, go, go. I don't know how healthy that is. ~ mentally. But it it is like for someone with, you know, a little bit of creative ADHD to just like go and build stuff all the time. I mean, do you do you feel does like the velocity ~ I don't know if is weighing on you is the best way, but just like does it feel like it needs to be like if someone who like needed to jog every day to keep themselves sane, right?
Play from 39:49 Like does it feel I feel like it's the same way that I need to just like build and ~ know what's going on and like try, 'cause I have to be ahead. And I I don't know, is do you do you feel something similar to that? Yeah. athlete, because you're trying to stay ahead of everybody else, right? And the competition is right there with you. Yeah, I like that. I ~ I am very surprised at how undercovered this entire idea of like people are working harder than ever, even though the technology is supposed to give you more free time.
40:06Why AI Saves Time and Creates More Work
Play from 40:18 It's like the human capacity for production will expand to fill the allotted time. Mm-hmm. So, like, cool, if you just give me more free time, I am going to then use the free time, right? And so ~ it here's an easy one for you guys. on YouTube, we get tons of spam as most people do, right? And so we have in the comment section, ~ every comment has to be approved.
40:36Training an AI to Filter YouTube Comments
Play from 40:47 Dude, we get hundreds of comments, you know, whatever, thousands of comments, whatever, however many it is, right? Now, the reason why that is interesting to me is unbeknownst to me, my brother, who's not technical, ~ he built an entire YouTube comment filterer. And then he had it set up in a way where for a week, all the comments went in and he went in and manually approved or didn't and trained it so that the and he was measuring the accuracy of its ability to say we should approve this one.
Play from 41:19 I was like, dude, how did you think of all this? Right? I and he goes, No man, I'm the one who approves all the comments. this shit, it sucks. Right. He's like, it's taking up hour of my day. And so I was like, you're motivated to do it because you're gonna get back the hour. Right. And so I think that is probably the one thing you one of you guys asked me earlier, what are you looking for when you're looking at these companies, these products, all stuff? Is like what motivated you to build this?
41:34How Pomp Spots an AI Idea Worth Building
Play from 41:46 Because there's something about if it's a really painful problem, you finally get tipped over to go and actually solve it. And my brother, I mean, we've been doing this for years. He finally hit the point where he was like, I'm done. I'm done with the YouTube comments. And he went, he built this really cool thing. And so I think that is ~ Something I always look for, but also then it's like, okay, for you guys, what are you gonna build? Well, what are the most painful things where you feel like you're wasting time on your schedule or or your day?
Play from 42:12 And it's like attack that first because you're gonna be most motivated to actually, get the solution. Definitely. And that just to just from a transition here, because I know we have you we are you are a very busy man and we are definitely on a a little bit of a time crunch here, but I wanna I definitely wanted to ask this because I I feel like Drew and I have maybe mentioned this from a level of the eight, the people that we talk to on our daily basis, right? ~ this this movement is so real, but it's also feels like a lot of it is performative.
Play from 42:38 ~ I feel like a lot of companies I come in contact with are like, yeah, we're totally AI enabled. The whole team is AI enabled, everything is, you know, our whole infrastructure is AI enabled. But you get in there and it's it's trash, right? it's like this is just smoke and mirrors. You just said that to me to get me interested. ~ from and I know you operate in probably a different a lot of different industries and we probably operate in. Do you do you f do you ever feel that? Do you feel like there's a lot of performative nature in it? Like people just need to attach themselves to AI to s either sound relevant or like not feel like they're, you know, behind in that in that sense.
Play from 43:09 Is that something you kind of like come across? Well, There are people who are ~ op ~ scheming, right? And they're like trying to take advantage or like capture the moment by like just frankly being full of shit. I don't worry about those people. Like those those people always kind of get washed out. I I joke that I lived in Miami for two years and I moved down there before kind of a bunch of people had moved there. And I thought I was gonna be there for, you know, a couple of months and then all my friends start moving and all stuff.
43:12Spotting Real AI Adoption
Play from 43:35 And so I was very fortunate to watch this because I kind of saw Miami, even though it's the pandemic, I saw Miami for about three or four months. As like Miami, not like a bunch of tech people are moving there. And then I saw the transition. And my number one ~ observation was all the guys who had the ~ rented Lamborghinis and fake Rolexes and all that, they scattered ~ rats when the lights get turned on when the guys with real money moved there. Right. They just they disappeared.
Play from 44:01 Cause all of a sudden they would go to a dinner and they'd be like, Yeah, I'm this blah blah blah blah whatever. And then like the guy next to him is like actually a billionaire and they just look stupid because they got all this like fake jewelry on, right? that's the guy's got a t-shirt, right? So ~ it was it was fascinating as a psychology or sociology study. It was incredible. But I think a lot of that is same with AI. you have the pretenders, they just get washed away and they run away whenever the real people are around, right?
Play from 44:27 So don't worry about that. What I think is true though is there's a lot of people who are like, I use AI and they genuinely mean it. But what they're talking about is they are replacing usage of Google with ~ a LLS. So it's like that that's the first step for sure, right? But they're just dipping their toe into what is possible versus you guys built the whole website for your podcast, and I saw you had the audio integrated in and an embed and like there's like work done there, right?
Play from 44:56 When I hear people talk about like I use AI a lot, my first question is usually what is the thing you've built that you're most proud of? And whatever that is, it could be something not really that complex and but kind of stupid. But then you can see them light up and be like, I had this problem and I solved it, blah, blah, blah. And you're like, okay, ~ you're in the mindset and you're obviously using this as how do I get the most value or the most juice out of the technology? You ask some people and they're like, man, it answers all my math questions on my homework.
Play from 45:25 ~ by the way, like that, it does do that, right? the you you should use it if that is a use case for you. But I just think what you're trying to measure is like, where are they on the adoption spectrum? And when you find the people who are actually solving their problems and using new technology and building and doing all that stuff. What becomes very obvious to me is those people going back to this idea of the iteration speed inside of a company, personally, they have a really high iteration speed. So a new voice model comes out, they'll start playing with that.
Play from 45:51 A new this comes out, they start playing like they're just using everything as it's coming out. And so right now there's eight hundred thousand people, I think is the last public number that's using Grokbot. 800,000 people out of almost whatever, 70 billion people in the world. So what you're It's bubble, right? Yeah. seventy two. Yeah, yeah. you're on the edge, man. you're you're you're you're you're tipping over the the the cliff there, right?
Play from 46:18 And so from that standpoint, it's like, okay, well, those 800,000 people, they're living in a different reality than people who, you know, I I always joke, don't know how to spell AI, right? Or the people who are just using it as like Google replacement. And so again, there's no right or wrong. I have immense respect for the like quintessential boomer who's got a flip phone. I'm like, dude, you might you actually might be have figured out life, right? Bring it back.
Play from 46:44 like yeah, you might be exponentially smarter than me because you've already speed ran this whole thing and simulated it. Yeah. by the way, your brain's probably not rotting. You don't give a heck, you know, about whatever's going on in the news. Like, you know, people can call you. You probably got five numbers saved, right? You like call, you know, your kids, your wife, like whatever. Like maybe you're winning, right? So like it it there's no prescriptive kind of ~ outcome.
Play from 47:12 But I do think that if you're like, hey, how do I get the most out of this technology? You want to go and coalesce around the people who are at the cutting edge. Not just the people who, hey, I use it as a ~ Google replacement type. I a hundred percent. I I was gonna bring up this really interesting thing. I think this is in in your book. ~ so one of your, I think, chapters says, Amateurs built Noah's Ark, professionals built the Titanic.
Play from 47:40 Can you tell me how you think about that as it relates to AI? Because I love that. Well, I don't remember where I I put in the book wherever I found that, but I thought it was it was very good. And ~ you know, the way I I think about that specific phrase is expertise is not good or bad. It's all about fitting the expertise to the problem. So if you have a defined problem, for example, ~ when you do your taxes, you probably want a tax expert. I don't want just like my smartest friend to do it.
47:43When Expertise Gets in the Way of New Ideas
Play from 48:06 I want a tax expert. But if I'm doing surgery, I want a doctor. I don't want my smartest friend just like, hey, let me cut you open and see what happens. Right. But then those are defined problems that people have solved before. When you have a problem that no one solved before, expertise sometimes can be a liability. Now it could be an advantage as well. But for example, Elon wanted to build reusable rockets. He went and talked to experts, and they said it's not possible. And he had to put together a group of people who were misfits.
Play from 48:35 And they came together and they figured it out. So I don't condemn. expertise, but I do think that you gotta understand what type of problem is this. And is expertise more or less likely to be helpful here? When you think of AI, I think there's a lot of things that ~ people who have expertise tend to be older because they've had to accumulate the knowledge and experience. They're less likely to use AI to solve problems in the same way.
48:39Asking AI What to Do Next
Play from 49:02 So one of the things that I have to constantly remind people on our team is they have a question. I said, Did you ask AI? Like what? And I'm like, they're like, I don't know what to do next. What do you want me to do? I'm like, did you ask AI? And they're like, why would I do that? I'm like, because it's smarter than me. And they just look at me like, they're like a deer in headlights. what do you mean? And I'm like, no, dude, this thing is smarter than me. It has way more, it definitely has better memory than me, right? It's got, you know, better context, all the stuff.
Play from 49:28 So, like, I find it funny sometimes where I just ask Silvia, Grokbot, whatever, what should I do next? And I'm just like, It works. I'm like, my God, literally it owns me, ~ I I I work for the AI now. I ha I have a whole skill for that. It's called blind spot. I just type it in. It's like, okay, this is what you should be doing right now based on our chats. It's like, okay, buddy, Now here's what's funny, right?
Play from 49:56 Is this not a new concept? ~ use Google Maps as an example. Like you plug in an address and it tells you where to go. You literally drive your car where the machine tells you to go. So it's, you know, it's kind of an extension or an evolution of that. But the first couple of times, it definitely is a little crazy when you're like, you know, what should I do? Because first you got have the humility to be like, okay, what do you think? But then two is when it tells you the answer multiple times a day, I'm like, Yeah, I wouldn't thought of that.
Play from 50:23 ~ I you know, is it like ~ should should I hang up my cleats? You know, let l let's just you know hang up the cleats and just let this thing run the place. it's like that's where it's like, you know, you you you hear the conversation around okay, AI can't be original, AI can't be creative, but it can be if it's got enough context as to who's using it, right? Or what the mission is, or you've got enough raw material to prove from, right? So The way it's able to be original and creative in that aspect is because it knows the things that you've done.
50:31Giving AI the Context to Be More Creative
Play from 50:51 It knows your domain expertise. It knows the things that you're working on, knows how you're thinking about things, right? It knows the things you like, you don't like. So it's like by context dumping to that point, you really can tap into net new original creative relative to the domain expert that's that's operating within the LLM or within the environment for that matter, right? I I think that's right. It's ~ it it it's very ~ it's very interesting to to watch all this play out.
Play from 51:16 And may maybe l one of the last things I'll I'll leave everyone with is AI, at least for me, has allowed me to reinvigorate curiosity. And there are certain moments where I've come to conclusions. So, Bitcoin, as an example, when I decided I'm gonna hand this to my grandkids, whatever happens with the price happens. You ain't gonna shake me out. It's like, you know, who says whatever happens, happens.
51:19How AI Changed Pomp's Curiosity and Delegation
Play from 51:44 And my are gonna think I'm a or my grandkids are gonna think I'm a genius or an idiot, but what they're getting Bitcoin, right? Okay. With AI, the equivalent to that was when I realized it's smarter than me. And I'm cool with that. And so the second that happened, what do I ask it? How do I give it content? I work for the AI. And I think that that was a zero to one moment for me with the technology. And so what I tell people is, you know, are you yet at a point where you can admit to yourself that it's smarter than you?
Play from 52:14 And by the way, there's some experts, a surgeon is like, no, it's not better than me. They're right. It it is like the surgeon is still better. But in what I do on a day-to-day basis, if I give it a hundred percent of the context I have, I don't know, man. It is ~ it's pretty good. And sure, maybe I can, you know, play air traffic control and direct it in certain ways or, you know, start new project, whatever. But like knowing where to press your advantage versus when to defer is a skill that humans have been trying to, you know, kind of perfect for centuries.
Play from 52:47 Well, I'm a better human manager and a better delegator with our people because AI has taught me how to do that. Very fascinating, right? Very, very, very different than I think what people are are used to. It's definitely it's definitely an eye opener. Definitely helps. Definitely helps in a lot of ways that you probably, unlock things that ~ locked unlocked things for me. I I don't even know I could do that or cared about that or liked that.
Play from 53:14 so it's it's a you there's there's a lot of there's a lot of net positive, I think, obviously. But no, we probably gotta probably gotta wrap this thing up here. But ~ if we're gonna Drew, should we do our our normal sign off here? We do probably ~ a two seconds if we're gonna if we're gonna roll with it. So stupid. Anthony, is we try to think of the most obscure demographic possible to ask for likes and subscriptions. It's just a it's just an inside joke that we've been doing since episode one.
53:20The Fast Hours Sign-Off
Play from 53:39 So ~ D ~ likes and subscribes. Yes. We're just kid if you've been here for an hour, ~ it's like, subscribe maybe ~ you guys either one of you guys married? Yes, both. Both with kids. wife's boyfriends for ~ for for license? There we go. Yeah, let's let's also tap into the the child demographic. If you've if you it if you've got age appropriate kids that are on YouTube, subscribe.
Play from 54:06 We need we need to build the child demographic of our subscribership. Well, the the the the opposite is that's an inside joke from our podcast 'cause whenever I would do ad reads and I would just start rattling off tell your brother, sister, mother, father, uncle, sister, whatever, like wife's boyfriend and people would be like, Wait, what what did that guy just say? We we we talk we we talk to everybody. We talk to dog walkers, we talk to TSA agents. It does we don't. What's the most obscure one you guys have ever done?
Play from 54:32 a some cake baker some bakers. you know, we've gotten we've got really weird TSA agents, ~ we've got you you do. you don't wanna you you don't wanna walk in and Yeah. Yeah, yeah. Well well so so yeah, it's my friend. dude, thanks thanks for joining the show, man. We know you gotta roll. We appreciate your time, your perspective today. And ~ yeah, we'll see you guys next week. Thanks for joining us. See ya.
Play from 54:58 Thank you guys.
