Why I'm Building an AI Writer

Jesse, there are people now who are creating a Jesse writing bot. Essentially, it's taking all of your internal calls and taking your voice memos and stuff like that and turning it into social content. I don't understand that. And let me, let me explain why.

Wait, sorry, who's doing that? You mean we're doing it internally?

We're doing it, yeah, internally. You're spending money creating internal software. And here's the part that I don't understand. There are all these other people who are working on this software. There are tons of apps already. Why not wait till they figure it out? Why spend your own money creating And if you are, why not turn it into a SaaS? I don't know why you're doing this internally for yourself. This AI writer.

Who, who's doing it? That's why I asked you on Slack.

Lee. That's what, that's literally what we're paying Lee to do.

No, no, no, no, no. I know we are doing it. Who is— you mentioned these, these mystery companies that are doing it and I don't know who they are.

Ooh.

Okay. Uh, you know what? One of the problems is they suck so badly that they're easy to forget.

Okay.

I don't even, like, get involved in them.

I rest your—

I rest—

I rest my case, Your Honor.

Yeah. But do you really believe that all these people are spending money on it, that you're going to figure it out with one person internally just for yourself? Well, I mean, beat them. Why not make it a bigger thing?

Maybe. I think the— I have a couple of thoughts on this, and I'm curious what Adam would say. One is like, is I have a dream and my dream is that I can have conversations like this. Or yesterday I was at a conference and I had all these really interesting conversations and I'm a unique human and I have unique worldviews and all these other things. And I want all of that. I don't wanna ever write a social post. And I want, I want, you know, the set of agents. And by the way, I showed this to a bunch of people yesterday and everybody felt blown away. I showed them Lauren's output of just like the categories that came from the Q&A we did. And then the actual posts that were written, and then how I rewrote them a little bit. And everyone was like, oh, wow. And it was the idea that, like, there was somebody pulling out a piece of it, pulling out the themes, and there was a different bot, you know, different agent that was writing them, and then someone else could edit them. And, like, it takes pieces to do this well. But I think that the notion of the specific idea I had of, like, I am a busy person meeting interesting people, giving advice to entrepreneurs all the time, and I want— I never want to write a piece of social content ever again in my life. I think Adam and team have run with that to try to solve that very specific problem. And I think that, I don't know that there's a software that's doing it, or you just said they're all shitty at doing it. So that's one piece of it. The second piece of it is I believe, and I think, Andrew, this is a big shift in your worldview. And I think you've been a little slow on the AI thing, if I may say that, to fully appreciate it, right? I mean, you were not a lot, and you weren't, yourself using AI because, because the softwares are shitty.

A few months ago, I was addicted to it as a consumer of certain things for sure.

But you've been very resistant to, to leaning in and adopting it.

And okay, I think I, I would like to hear more about that later, but let's keep going with this.

But I think one of the worldview you still think like you think in the old paradigm, which in the old paradigm is software. Software is the thing. The new paradigm with AI, I think one of my, one of my opinions is because AI is fundamentally bespoke to an individual person's situation, because that's the promise it holds to basically automate this thing that's very specific to Jesse Pujji and his, his orientation as CEO. Then it's, it's, I think there's going to be, we've talked about like a trillion-dollar services industry buildup around it because you actually have to get in and understand my very specific situation. You actually have to understand, you know, things about me that are unique. And then you're going to build me a piece of software effectively called an agent that only works for Jesse Pujji and only should work for Jesse Pujji because it's so powerful for me. Because the challenge, like old SaaS software was just too generic. It didn't, you know, and then it had to be customized or you had needed a person to operate it. So the other, the other reason, Andrew, is I think this is actually the way that a lot of AI software, whatever you want to call it, tools will be built. in the next paradigm here, and it won't be through software.

Custom created for individuals. So you hate your email software, there will be an email software created for you, the front end. You hate your—

When there's a business case for it. I don't think your consumers are going to go, no, my wife's not going to get an email app built for her. But I think Dollar Shave Club's email app will be different than Hulu's email app will be different than Bootstrap. You know, they're all different and they're all agents that are specific. And now Will there be a software that you customize to this? Maybe. But I actually think, like, in many ways, the platforms like OpenAI and others, they eat up a lot of the, like, you know, people like Adam, CTOs, and my old engineering— they always use the word plumbing. Everything's plumbing, plumbing, plumbing, plumbing, plumbing. I need data. I got to pull data. And that's plumbing. The interface, Jesse, is very straightforward. Well, ChatGPT and even the ability to create plumbing through writing code, but just the platforms themselves have I think obscured a lot of the plumbing and will continue to obscure it so that it's gonna be much easier to, to create things that are custom on the surface of one of those platforms.

Okay. This is, this is one of those things really that's been hard for me to accept. And Adam, I wanna bring you in on this too, but I still see a world where SaaS exists. And what you're saying is it will exist, but for a lot of business needs, we will have our own custom software. And the example that, that comes up a lot that does make sense to me is CRM. Everybody hates their CRM because it never does what they wanted to do, have custom CRM. And it's such a vital part of the business and it produces money. And so it's worth spending some developer time.

But let's use actually a real, another real life example though, that is the most common. What is the, what is the most common bespoke software tool in America right now? Andrew.

But bespoke meaning it's made just for me?

No, no, no. We, we all have bespoke tools we've built inside of it. And everyone knows how to use it. And billions of dollars in transactions— my hint— billions of dollars in transactions take place in this because of this software every day.

I can't think of a single thing that is actually bespoke that I'm using. Adam?

Excel. I was going to say Excel. Yep.

You have an app called the Bootstrap Giants P&L. You have an app called Personal Finance of the Warner Household. You have an app called Calculating metrics for my business. That is, it's just bespoke to you. It's, it's a very completely bespoke thing that's unique.

I've used Excel to create a bespoke experience that's unique to me. I get it. I do remember that even in college.

When you open Excel, Andrew, what, what does Excel want to create? Is Excel saying, great, we're here to make a P&L, or we're here to model this? Just says, hi Andrew, what do you wanna do today?

And then here's cell A1.

What would you like to put in it? And, you know, now you take these like SaaS platforms, it's the exact opposite. Has Gmail ever asked you how you want it to show up? Has, you know, how much can you change? Oh, there's like, there's a settings page with 15 different toggles that you can, you know, turn on and off. You could change the background, but it's a big, crusty, hard platform not designed to change. There are tools like this that are designed to say, hey, what would you like to do today? Go to, you know, Lovable, Replit, Base44, these like site builders. They can rebuild any website that you've ever seen or touched. They can build CRMs, they can build landing pages, they could build marketing pages, they could build anything. They don't— when you open it up, they don't want to do anything. They're just ready to do whatever you want to do.

And the future— but I think the future, like when you— if you think about Excel's If you percentage of tool or software automation, you've got these, again, you've got these first-party softwares depending on how techie your company is, right? I'm on the board of Schnucks and like, I think 10% of their software they've actually written themselves for their own business, which is kind of cool for a grocery store.

That's awesome.

DoorDash is probably like 50%. I mean, you know, there's techie companies that have written, then there's third-party software, which has been the big wave of the last 20 years. There's horizontal SaaS, there's vertical SaaS, right? Um, but then there's Excel, and I, like, I would argue Excel today is probably 10 to 50% of the software inside of every company. And, you know, the hint I gave you guys— people joke about this on Wall Street, and I worked on Wall Street— is like, dude, billions of dollars transact on Excel software.

What do you mean by that?

At the heart of every single major M&A deal is a custom piece of software that is the model that was built For that deal, and probably there's three of them. There's like the buyer's model, the seller's model, and it's it's software. Like someone goes in and says, "Well, tell me what happens if if we cut this cost," and someone goes, "Oh, here's what it is," right? Like so I think we this that's software that is automation that is tooling. When I when I worked at McKinsey, we got hired by a hedge fund to look at Verisign for them, and I personally built a model. I probably I don't know if I I have it, but like. It was trying to predict domain name growth in the next 5 years. And I had this insight while I was building it that because of my own domain buying, which is once people buy a domain, they very rarely let it expire. And so then we took business growth in the US, we took small business growth. And by the way, small businesses grow a ton, but they churn a lot. But we had this killer insight and it was my insight. I'll take credit for it. And this hedge fund made, you know, hundreds of millions of dollars for it. that they're never going to let go of their domain. So, so at first, because we ran the numbers backwards and we're like, wait, but all these companies get started and they go out of business, right? And, and, and we're like, but the domain growth historically doesn't show that. It's as if these guys are still around. I'm like, well, that's because once I buy a domain for $10 at Jesse's Coffee Shop, I'm never, I'm never going to let that thing expire.

That's—

I'm going to hold up— my heart's going to hold on to it forever.

If the business goes away, I hold on to the domain.

Haven't you? Do you all have it? Have we all done that? Don't we all? You have like—

I hate that I do it. But I do, I hoard them.

It's shameful, but yes.

And so, dude, using that, we basically were able to take automation of small business growth in America, layer it onto domain name growth, and we were 100% right. And VeriSign had a .com, .net monopoly on the business. And this hedge fund who paid McKinsey $1 million probably made several hundred million off that insight because the market didn't realize that. The market just didn't realize how much domains would grow. But in any case, that's a fun side story. But the point was I built a unique piece of software that could take inputs and then could spit out outputs and had all this cool logic built into it. And I was 22-year-old Jesse sitting there, you know, building. But Excel is such a huge part of custom software inside business. Now, Adam was talking about Replit and all these other guys, whatever the tool is, I think that that is going to, you know, 5 years from now and 10 years from now, there's going to be so much of that software tools, whatever we want to call it, inside of everyone. In fact, I think it'll be the biggest part of every organization. And by the way, I think it's a massive services opportunity, but It's just a massive opportunity. It's just going to change everything.

What doesn't Excel do that made the space for those big platforms to come around? Why am I not using Excel as my CRM if it can build everything?

Yeah, that's a good question. I mean, I think you may know better than me, Adam, in terms of the actual technical answers of like how much memory they can store in terms of like, you know, everyone has the Excel model that crashes their computer every time they open it, right? Um, I mean, I would tell most founders, start with Excel as your CRM, right? Name, when did you last talk to them, like just key— or a Google Doc or whatever, right? And that's where it starts. And then at some point it's like, I want automations, I want something to ping me to tell me to— I want a drip campaign, I want enough data stored in here, I want a place where multiple people can use, can update the record. So I think there's all these, these other things that come up to run a CRM well feature-wise that Excel just basically can't allow for. Although I bet you, dude, if we went out and we should tweet this, show me the best version of an Excel CRM you've built, it would kind of blow our minds how far someone got building it. I like, it would be better functionality than some of the, some of the like SaaS ones.

Adam, I feel like you're getting at something with that question and some, some kind of depth that I want to get to. Yeah.

I mean, I, you know, I wonder what's, what's going to happen next, right? Like what, like if, if, Excel can do everything.

Yeah.

I think LLMs can do everything. They do a lot of stuff then really poorly. And I, I think there's a problem with a tool that, that says, hey Andrew, what do you wanna do today? Well, I don't wanna have to figure out the whole problem. Like do some of the work for me tool. And so what is it that Excel couldn't deliver that made the space for these platforms to come around? And then what are these platforms not delivering that's gonna make the space for these LLMs? Like, Is this just like a cycle or is this the next step?

Like, well, I mean, one, one simple one would be like, where, where we started this whole story, which was, I don't think it's, I guess Excel could do some content creation, but not that well. You know, you could concatenate a bunch of phrases that Jesse uses and then make a formula that would spit out, but it, it can't, it can't generate generative AI. It can actually generate words and sentences that are quite amazing, relatively speaking.

Well, with Google Sheets, you, You've got the AI, you can run Gemini on any cell that can reference all the other cells.

So, I mean, pre-AI, like, yeah, pre-AI though, like that would be a limitation of Excel. And you can do some stuff, but I think that the generative nature is a unique thing. And then I think stuff like you mentioned, Lovable, like the fact that you can use words to then generate like website objects and colors and the fact that it can speak to something that can then generate those things.

Yeah.

feels like another level of— I mean, there was that— Adam, maybe you can try to explain the thing you and I talked about in my office the other day, the funnel thing that—

Yeah, there's a word people use to describe LLMs and LLM tools that I think we should all be putting in our vernacular, which is amplify. The LLMs amplify whatever you put into it, and the system around it maybe multiplies that amplification in one direction. Maybe it's a little more specific, but what happens is you take the LLM. Let's say we're all just like talking to the base OpenAI models. Well, there's no tools connected. It can't see the internet. It's just the knowledge it had on, you know, in May 2025 or whenever it was trained. And there's amplification there. That thing can type at 12,000 words a minute. I can type at 70 words a minute or whatever. And cool, I can do more with that. But I can't do all that much more. So then you take a platform like ChatGPT and what is ChatGPT like? Andrew, do you know why you are addicted to it as a consumer?

I have a— there are a few things. It answers my questions quickly.

Uh-huh.

Um, it lets me talk to it and it draws me out. There are a few thoughts. Theories like that.

Yeah. Why is it quick? Do you know why it's quick? Do you want me to tell you why it's quick?

No, tell me.

Often ChatGPT is really great for consumers because of its memory tool. It knows you, it knows your style, it knows your preferences, it remembers kind of who you are and the way you want things to show up. If you go and talk to just the LLM, like we're using the API version of these LLMs, No tools. Every time you talk to one of those, you're starting from scratch. You know, Big Bang, new world, everything starts from scratch. You, you have to give it all the information that you want it to do. So, all right, what's one level above the LLMs that's a little bit more amplified, but also a little more specific is these like chat platforms, ChatGPT, Claude.com. You know, you're actually going and talking to an agent, not the LLM. It's got all these extra tools. Now it can do more, Mm-hmm. But it can also do less right now. You can't do some of the things that I could do. You can't run 50 requests in parallel and kick 'em off and do all this. All right, then you take, let's say, Lovable. Like, Andrew, you could build a, a website on ChatGPT. You could ask it to write HTML and CSS and make it pretty and use your design reference, but it's gonna take a lot of chatting. It's not really designed for that thing, so it's gonna be hard to do. Plus at the end, you've still gotta like copy and paste all that code out into something else. What are you going to do with that code? I don't know what you're going to do with that code. Now you go into Lovable. It's still just— Lovable is built on top of the Anthropic models, I believe. It only wants to build websites. Still a chat. You could still say anything you want, but it only wants to build websites. Well, now you don't have to tell it to make the boxes pretty. You don't have to tell it to use nice colors. You don't have to tell it how to host a website. You're not copying and pasting the code. One more layer on top of the ChatGPT-type agent is a purpose-built agent for like building websites. But like Jesse's saying, cool, now it's way more possible. You could build a website and host it in 10 minutes. You, Andrew, like we could do it right now before the end of this call where you've got a brand new website for yourself. And Jesse did this once while we were sitting together and that thing searched about him. It posted it up. That's great. That's like really, that increases the capabilities. But now that you have those capabilities and you say, awesome, Take access to my email. Here's all my customer data. Here's access to our internal file systems. All right, host that website, make this CRM for me, and give me a signup. Uh-oh, is all that stuff available to the public internet? Do you know? Do you have any control over that? Do you— how could you check that? And often what we see happening there is, is people hosting private data, customer data, and there's all these scrapers. People are like, ooh, I'm gonna go and find every lovable site. Yeah. Because I know 50% of them have some proprietary data that like the person using the tool had no idea when they uploaded it in there that it was going to be public. Now, ChatGPT, you're uploading custom data all the time, right? But you're generally not sharing those publicly. So it doesn't really matter. You know, OpenAI has it and we could debate if that's good or bad. Lovable, there's a big publish button. It's so easy to click that publish button. Oop, now all that's on the internet and you didn't know. It's not obvious to you. The agent might even lie to you. and say it's not accessible. It's like, as you grow, the amplification— any of these LLM amplifications, they don't just make the thing better and easier, they increase the risk, but they're also more specific. Like, you couldn't easily process huge amounts of data in Lovable or do deep research. I don't— they don't have a deep research feature.

Adam, where, where does this end? I mean, it makes sense, right?

You—

yeah.

It's almost like any supply chain, right? It's like I buy— I can go plant corn in a cornfield, and then somebody else farms the corn, and then they sell the corn to someone, and then someone beats it into cornmeal, and then they sell it to the bread guy, and then the bread guy sells it to the grocery store. Then somebody, a restaurant buys it from the grocery store and then cooks it, and I get a sandwich. I mean, we forget how much of a supply chain in the world exists today.

Yeah.

So, it reminds me, I mean, that's an analog in my head. But where does, like, when you were talking, one thought I had was like, you know, there's like, back to SaaS versus services or how it's all gonna play out. Like, do you think, let's make up a dumb example, like municipal governments. It's like a really niche category. Is someone gonna, is Lovable gonna solve the problem of them making and managing their websites? Or is someone gonna build like, you know, should we go build municipalgovernmenthostingwebsite.com, right? Munisites.com. Munisites.com. And it's all powered by AI. So it's like a way better version of Squarespace or whatever, and then that's going to be the tool that everybody uses. Or our services business is going to come in, and it's going to be a mix of everything, of course. But like, is that— I guess, is the municipal government website idea a good idea, or is it a bad idea?

Do you remember, Jesse, do you remember when you tried to build your lovable personal website? Do you remember how that went?

Yeah, it was, it was like blah. It was meh. It would have taken me really like an hour to make it probably what I wanted, maybe more.

Uh-huh. Um, I think I agree. And why, like, why, why, why was that? Why would it take a whole hour? Like, if this tool can do anything, like, why would it take an hour? What was it wrong at? What did it, what did it miss?

What did you miss? I don't remember, but it was, it was like, you know, the color scheme wasn't quite what it was. The way it had, it was missing. Remember, I think it was missing a couple categories of tabs I would have wanted on the site.

Yeah. It, like, I remember you were trying to also get it to Look, go look up all the podcasts I've ever been on and list those out on the page. And it just—

Yeah. When I think of the spec versus—

Yeah. Yeah. So, so I, I think the, you know, if you look at Lovable as an Excel replacement, uh, is there an Excel for municipal governments?

No.

Not that I'm aware of. Is there an Excel for anything that's not just Excel?

Well, I mean, there, there's interest. It's a good question actually. There are, there are several plugins. I mean, there's actually a nice little, plugin industry. And I think the most famous example, I forget the name of it, but it's an insane business. It's like a 60% EBITDA margin business doing like $50 million EBITDA a year.

Excel plugin business. All they do is Excel plugins. Yeah.

What is it?

Is it one specific plugin for one industry?

Yeah, it's like a PowerPoint plugin for consultants. I mean, I'm sure there's other versions of this. This is just what I'm familiar with.

Yeah. What the hell was it called?

I used it all. Think-Cell. Think-Cell. And it's actually, you know what it specifically does is it connects Excel into PowerPoint. And it's like a famous, it's like a bootstrap giant and it makes tons of money.

How is this something a regular person would use or only a McKinsey consultant?

Only, yeah, people in consulting or whatever. Oh dude, it's way bigger than I just said.

How much?

Revenue is $200 million in ARR. EBITDA 150, and the first article says Sinven considers 3 billion euro sale of ThinkCell in test of software.

Wow.

Boom. It helps build data-driven charts, automate slide layouts, and improve presentation.

I mean, I lived in it when I when I and it wasn't I didn't even live in it. I was in PowerPoint and Excel, and it was the one I'd like there was a menu in my Excel that I'd pull down and go ThinkCell.

Yeah.

Yeah.

I think this is probably a pretty good analog to answer this question of, well, you could have this plugin. You were a, you were a person at a place that had resources. You had this plugin, or there was probably someone you could have written that would've worked on the, on the deck for you. Or there's a consultant somewhere else that'll work on it for you. There's an agency, there's a software. The question in a little bit is like, where do you want control? Where do you wanna own kind of the workflow?

Yeah.

And Andrew, this, this is a little bit of a tie back to why would we build our own custom workflow and tools around content creation? Like, we want to own it like that. At the, at the core of the whole thing, you can just say we're interested in owning that because it's so important to us in our business that we would like to have all of the control and all of the upside. And I bet the reason Jesse didn't farm out this work to someone else, like, well, no, you wanted to push the buttons, right? Like, you didn't. You could have asked for help, but you wanted to be in charge. You just wanted a tool. Sort of.

I just don't think anything exists. Or Andrew, you sort of said it all sucks. So I—

I've sat in on meetings with consultants who said that they do this really well. And when I look at the results, I just don't even have the heart to tell them how bad it is.

Yeah.

So I talk about it internally. Okay.

I think there's that. But sorry, can I pull on one more thread on this?

Uh-huh.

Because my brain is just going now.

Like, I can see.

There's this thing, Andrew, where Adam, you know, or there was a post by Reid Hoffman where he talked about becoming voice-pilled. Do you know what that is, Andrew?

No.

It's this idea of, of like, because prompting is becoming so normalized in, uh, in AI, like, and voice is just a way easier way to write a prompt, right? And like, Adam showed me this. He's like, we're all going to be talking to our computers in 5 years or 3 years. And I agree with him. And even like, I've gotten Dee, my wife, Like, she's, she's a consumer of ChatGPT. And I was like, well, just, just talk to it. Like, tell it what you want specifically and you'll get way better answers. And now she does it and she gets better answers. I just— this is like a weird thought. Like, the municipal government— I don't know why I'm like— I had this visual of like, okay, there's some app for municipal governments. And every day the guy, the guy or gal or whoever— if you— if they had to write a prompt and do work, that would be annoying. But if they could just walk in and say, hey, today on the site, I would like you to feature the the farmer's market that's gonna be there. Can you go ahead and do that? And the computer's like, yep, it's on the site. Or like, hey, we need to add a calendar for trash pickup. And I don't know why the hell I'm talking about municipal city, but like, there's something really interesting about thinking about the world voice first and how much it changes the way we interact with software that just, it feels like another vector in all of this.

You know, Jesse, it's amazing how many people still do, even in our space, even people who are addicted to AI, still do not do voice. You talk to Alex Lieberman and me, he and his team, Arman is the guy who runs his dev shop, they don't use audio. And it's because a lot of them are in rooms with other people. And I think it was Sergey Brin who had said in an interview with All In that he had never tried it, and because he was being asked about it, he said, you know what, I'll go back. But he said, I don't try it because there's a room full of other developers around me. I don't want to interrupt what they're doing. But it really does change things. All right, coming back to, um, The, where this is all going. Adam and I were talking about how you all have a thesis on AI. Is, is you trying it internally part of that? Part of you saying, hey, you know what? We think every company's gonna have bespoke software. We think they're gonna be these massive companies that build bespoke software for enterprise. Let's try it on ourselves and then build the next one in a collection of it. Is it that you wanna create SaaS that's built with it? I'm trying to figure out what it is. What is it, Jesse? What is it, Adam?

You go first, Adam. Yeah.

At minimum, it's a, it's a flywheel. You know, these tools are, are hard to learn. A lot of them suck. Some of them are really good.

And we're talking specifically about AI writing tools that will write in your voice. They also—

no, all of the tools, you know, I've tried video editing tools, I've done writing tools. I demo a bunch of tools every week. I've got a special credit card just for my AI, you know, R&D testing stuff so that I, you know, I don't accidentally spend all our money. And most of them are not good. What, what I'm trying to find is usually some combination of there's a tool that's actually good. There are a handful of them out there that are like, oh, this is a phenomenal tool. The tool on its own though, if you just, if you tried to use it tomorrow, you probably wouldn't get that great of results because most of the best tools, they're not just great at doing the whole thing all around, all on their own. That tool is combined with a workflow that may or may not work with you and your business and how you want to do it. You also might not know it. And these tools are bad at educating what the workflow should be. So you've got to figure out the right way to work with the tool. And then you have to figure out how to communicate with the tool. Usually that's prompting, sometimes clicking buttons, whatever. And when those 3 things come together, boom, there's a tool. So we want to be trying as many of the tools as possible, but we also need to be building it because in a lot of these cases, like the workflow is actually more interesting than the tool. And the tools are going to keep changing. I guarantee we're not going to remember the name of hardly any of the tools we're using today. 2 or 3 years from now. But so I'm not too worried about which tool is there and how good or not they are. I want us to learn the best workflows that get the most juice out of whatever the state of the art is in the tool for video editing or content creation or coding or whatever.

Because you're saying, look, no matter what the tool is, even if it's video editing, there's going to be an iterative process between the user and the software that someone's going to need to know intuitively. We want to develop that. I get that part. The part that blew my mind was you're saying, I don't think any of the software's ever gonna be great. You don't think that we're going to get software that takes this video and creates clips?

I, I think Lovable is like pretty cool right now. Like, does Lovable stick around for 5 years? I don't know. I, I probably wouldn't, probably wouldn't bet that much on it because 2 years from now it'll be a pretty big platform and they've got a lot going on and they're focusing on integration and enterprise partnerships. And as the LLMs are better, there's better research, better methods. Someone else is gonna really quickly create another version of Lovable, just as quickly as Lovable was created. And I wanna be on that tool when that tool comes out. I don't wanna be stuck on, you know, this Lovable tool.

Do you, but Adam, do you think Andrew's, Andrew asked a question, uh, uh, outta the side of his mouth just now. It was like, like video editing, for example. Do you think there will be tools in the next 5 years that will be as good as Michael, like as a video editor?

Right.

Yeah, I'm actually— I need to go and find— there's this tool, I think it's called Eddie AI, that even today is— oh man, I want to try it because I signed up a couple of weeks ago for Descript. I wanted it to work. I really wanted it to work. I've already canceled my subscription because in the first hour or so I spent all of my credits and really didn't get anywhere. It didn't, didn't really work. I wanted to like edit a video by editing the transcript and combine a couple of them together. Spent all my credits. So, all right, fine, I'll keep trying. I upgraded to the next tier, spent all of those credits, and now I was on like $200 a month plan. Okay, this still sucks. This isn't worth $200 a month. But I've been watching like some YouTube videos and some other people talking about this. Eddie AI seems to be an interesting tool. It doesn't promise everything. Descript, anyone off the street could walk in and edit videos. CapCut, anyone off the street could come in and edit videos. You kind of get crummy results. And if you want something special, I think you're going to spend all your credits and not get it. Eddie AI like plugs into Premiere and Resolve. It's your, it's your Excel plugin. It doesn't, it does some stuff on its own. It's like it can ingest content. It'll process it all. It'll help set up your timelines. And then it kind of hands off a lot of that work to the nonlinear editors that already exist. And this is, I think, is like one of those workflow things. If you want the tool to own the whole workflow, Like, it's not good today and maybe it'll never be that great because my workflow is my workflow. I want my artistic process. There's tools like Eddie AI, again, I need to test it, but tools like that that might enable me to scale my workflow dramatically. And I think we're going to see that like in a lot more places like this plugin idea. Like, I'm pretty bullish on like, heck yes, I want a lot more plugins that exist in the tools I'm already using that enable my custom workflows.

Mm-hmm.

to like amplify me because the tool doesn't amplify me. The tool amplifies itself, but the workflow, and if it can do a bunch of the work for me, now I'm amplifying. And in theory, I'm in this room because I'm smart and special and experienced and, you know, capable of figuring stuff out. So I want, I want the workflow to amplify like from there, not from what the tool thinks I should do.

Yeah, it makes sense. I don't— I mean, you know, the— my version of the answer Andrew is like, I tell my kids, when I talk to my kids about CDs, they like don't understand why CDs ever existed, right? Because, because they go, well, what, what was— how did you listen to music? I was like, well, there was these things called tapes and they were this big and you put them in a Walkman and you listen to them and then you rewind it and you forward. And, and, and then they're like, wait, why was this whole other form factor? Well, you could skip tracks fast and it didn't You know, there was these other, like, it was thin and, and like, but from their vantage point, they're like, well, everything just streams straight to like a, like a, you know, they're like that whole intermediate phase of technological innovation feels dumb when you look at it from their lens backwards. Now we all live through it, so we're like, well, yeah, it was incrementally like much better than a tape. It didn't get all whatever. I mean, we could come up with our reasons for it. And the CD industry was a massive industry for a long time.

Right.

Right. So I do think my version of what Adam's saying is like, I think human beings take a long time to adopt technology, longer than the technology moves faster than humans' ability to adopt it. I think there's going to be a lot of interim form factors and tools before we get to the ultimate version of this. But I will also say, like, I—

this—

I don't want to say this time it feels different, but because that's, you know, But like, this stuff does feel like it's moving faster. What we can say is that it's moving faster. It feels like a significantly more powerful technology than anything we've ever seen. And obviously, again, just like the, you know, this thing of like you can have this stuff write code. I don't know. There's versions, there's things about this that make it very, very distinct.

We can definitely say this time is different and it is different. But coming back then to your thesis, you guys are apparently raising money. And I only say this because NAC, This COO has asked me for some things that seem like it's going into a, into a deck. Um, are you raising money?

Maybe.

Okay. So then let's suppose you were raising money. I'm looking at your faces. I like that I've made you guys uncomfortable. I don't want this to be such a happy clappy, let's all get together. Let's get to the discomfort.

All right.

Let's suppose you were raising money and you needed to build something. What would you build? What's the thesis that you would bid, that you would use this money with?

I think, well, I think the— there's a couple of things that get me fired up. I think personally and professionally, I think one is, I, you know, I love being around lots of entrepreneurial people. And entrepreneurial people are usually really smart. They're very self-motivated. They're coachable. They're— there's like a real feeling, you know, when you're around someone who's an entrepreneur. And, and like, I think of my life dream and goals, like, I just want to be around lots of those people and hopefully help their trajectory and shift them in some meaningful way. And I think like, I live in St. Louis and I love St. Louis. And I think like having something here would be very, very interesting and compelling. And I think there's a lot of people from the Midwest. I think there's, I think there's tons of talent. You know, there's that some quote that talent is evenly distributed, but opportunity is not. And I think there's not enough opportunity for talented people here in the Midwest. Like, you know, my— one of my— some of my favorite quotes, like Andreessen went to the same college that Adam went to, UIUC, University of Illinois Urbana-Champaign. That's like this— it's the MIT of the Midwest, right?

Mm-hmm.

So did Max Levchin. So, I mean, you know, Larry Page is from Michigan. There's like— there's a really well-worn path, actually. And I think there's a big opportunity for more of that happening here. And then, you know, I think obviously if we were to build something, I think AI is important. You know, one of the observations as a marketer I've had is you go category by category. Well, AI is going to shift healthcare. It's going to shift legal. And when you say, how is it going to shift it? It's going to shift the way they do work. They're going to, you know, a lawyer to create the first document, it's easier now, whatever. I think that shift is going to happen for marketers too. It's going to shift the way marketers work, how to add creative, get made, how do the emails get made, how does content get made on social? But marketers have a double shift because in 5 years, if you're a marketer, not only is your way of work different, but the way that people discover businesses and brands is gonna be very different because of OpenAI and the other LLMs are becoming another channel. They're becoming what Facebook became 15 years ago, or Google became 25 years ago, which is the place where consumers start to try to get information about something.

Mm-hmm.

And I think, you know, I, we wrote this post recently about OpenAI is gonna be a $300 billion marketing business because it's got the search aspects. of Google where someone tells you what they want, and it's got the social aspects of understanding you extremely well. And as a marketer, that means very high CPMs effectively. So for me, I think you put all that together and you go, hey, Jesse wants to have a bunch of smart entrepreneurial people where he lives and focus on the things he knows, which is marketing with this new AI shift. And that's, that's like where my excitement and energy is right now.

So would it be A company that buys a bunch of marketing companies or a company that builds marketing companies that all help companies buy ads on ChatGPT? Or what are we talking about more specifically?

Well, I think, yeah, I mean, this is a fun topic that Adam and I have been jamming on a bunch. I mean, I think it's some, it's a studio accelerator. It's starting companies. That's my joy is in the moment that, you know, that the—

is this—

could this be an idea? Is this gonna work? Oh, I see it.

I think you still want to start companies. It's not a pain in the ass dealing with people like me, the chaos, the frenetic everything. Adam is—

No, I, I—

you're saying I'm a pain in the ass?

I don't know.

I mean, you're sick of dealing with me being a pain in the ass.

Well, I think I enjoy starting the companies. I think we need to do it in a way that does— like, is probably a little more scalable and does— takes less of our, our bodies and more of our brains.

But it's starting companies At the intersection of AI and marketing, and they've got to love St. Louis.

They've got to come spend some time here, be from the middle, you know, they don't have to be from the Midwest, but we want, I want people here. That's like more of like a, that's just, you know, if you're going to go do something you really, really want, do the thing you really, really want. And that's how I feel about it. But, but, but Andrew, you started going in this like creating ads things for, you got to really start to peel the onion on how distribution is going to work on these platforms. Right? So, there's like the— and I watched this happen with Facebook and I watched this happen with Google. There's level 1, and level 1 is pretty straightforward. It's like these AEO companies. Hey, you're a brand, you want to show up on the thing, I'm going to help you do that. I'm going to make content. Those are good businesses. I think they're going to be compelling. There's like a level 2, which is like, well, what happened in Facebook? There was this category of mobile measurement partners.

Mm-hmm.

Kochava, AppsFlyers, all like multiple billion-dollar businesses, which were smaller than Ampush at one point. That all became bigger. That's gonna be an industry in, in AI analytics built around cross-LLM analytics, zero-click analytics. There's gonna be a whole analytics infrastructure get built around it. But let's keep peeling the onion and kind of go, what happens when people don't really need websites anymore? Right? So, so let's pick a, pick a random vertical travel. Okay. Well, easy. Well, if I, if I can book right through OpenAI or Google's new, whatever Google will do something. Okay, the website, whether it's from Expedia or Kayak or American Airlines or the FourSeasons.com, like that shit doesn't matter to me anymore. What matters to me is I've come up with the perfect vacation itinerary and then I want to pick through a few options and I want to select them. Well, what's that going to create? Well, that's going to create an opportunity for data tooling for all those companies to be able to feed into the, into the MCPs correctly. Someone's going to be optimizing to go, well, how do you make sure that ChatGPT thinks to pull from your, you know, your feed versus somebody else's feed? Well, that's going to be a whole nother, another business. There's going to be an inventory management system. So you have to start to think of the world many, many layers deeper. And if these things become distribution platforms, a whole new set of businesses that we can't even fully imagine right now will come to be. And so part of the reason I love this trend is because I, again, I watched this with Facebook and like it went on and on. And I mean, there was just things like, you know, there's this whole set of companies, Andrew, you might remember them. Adam was probably not even in college. Yeah. Like Buddy Media, Wildfire Social. They were all like hearsay social, all really good businesses and good outcomes that were trying to help you manage your Facebook page and help you manage like you know how you commented and liked and whatever. And that was a whole industry. I mean, there's still like Sprout Social, and there's some companies that are still actually big businesses that help you do that. But but like that was really the version one of social. I mean, it was a long long time ago. So I think part of the reason I love it is 'cause I think there's like just gonna be so much runway and so much iteration in how these ideas take place that we think that, so it's not just like a, can I do ads on ChatGPT? That might be an idea, by the way. I'm not saying that's a bad idea, but it's just a very, you have to go down to the basics of, or like very first principles thinking around it.