This Week in Startups - How Replit is using AI to supercharge IDEs with Amjad Masad | E1814

Episode Date: September 22, 2023

This Week in Startups is brought to you by… LinkedIn Jobs. A business is only as strong as its people, and every hire matters. Go to https://LinkedIn.com/TWIST to post your first job for free. Terms and conditions apply. .Tech Domains has a new program called startups.tech, where you can get your startup featured on This Week in Startups. Go to https://startups.tech/jason to find out how! Supergut is the only nutrition brand clinically-proven to improve digestion, balance blood sugar, sustain energy, and manage weight. Save 25% on their delicious shakes, bars, and prebiotic mix at https://Supergut.com with code TWIST. * Today’s show: Replit CEO Amjad Masad joins Jason to discuss the latest developments in IDEs (3:04), leveraging AI for coding (11:48), Replit’s Ghostwriter (23:40), and much more! * Time stamps: (0:00) Replit CEO Amjad Masad joins Jason (3:04) The origin of integrated development environments (IDEs) uses and importance for developers (6:44) IntelliSense, Replit's Ghostwriter and other debugging tools (10:28) LinkedIn Jobs - Post your first job for free at https://linkedin.com/twist (11:48) Leveraging AI in suggestion tools and coding platforms (14:17) The role of autonomous agents in startup development (19:20) Replit's Bounties platform and Replit's mission to build artificial developer intelligence (ADI) (22:25) .Tech Domains - Apply to get your startup featured on This Week in Startups at https://startups.tech/jason (23:40) Amjad Masad demos Replit's Ghostwriter and other apps (31:46) Getting more people to become developers and low-code platforms (36:35) Supergut - Get 25% off with code TWIST at https://supergut.com (38:06) “Make Something Wonderful” Steve Jobs in his own words (43:56) The human mind and the origin of Amjad's passion for computers (48:22) Embracing AI and the resulting changes at Replit (55:27) Paradigm shifts in Hollywood and tech (1:03:06) thebrain.com * Check out Replit: https://replit.com Follow Amjad: https://twitter.com/amasad * Read LAUNCH Fund 4 Deal Memo: https://www.launch.co/four Apply for Funding: https://www.launch.co/apply Buy ANGEL: https://www.angelthebook.com Great recent interviews: Steve Huffman, Brian Chesky, Aaron Levie, Sophia Amoruso, Reid Hoffman, Frank Slootman, Billy McFarland, PrayingForExits, Jenny Lefcourt Check out Jason’s suite of newsletters: https://substack.com/@calacanis * Follow Jason: Twitter: https://twitter.com/jason Instagram: https://www.instagram.com/jason LinkedIn: https://www.linkedin.com/in/jasoncalacanis * Follow TWiST: Substack: https://twistartups.substack.com Twitter: https://twitter.com/TWiStartups YouTube: https://www.youtube.com/thisweekin * Subscribe to the Founder University Podcast: https://www.founder.university/podcast

the technology industry. These guys were all on acid and mushrooms and coming out of the hippie-zippy movement. And they really were opening the doors of perception. And then they were with computers at the same time. Yeah. And, you know, imagine you're tripping on acid. And, you know, you're also like using new technology for the first time. You're like, yeah, I have some interesting ideas here.

We should make a mouse. It's like, what? A mouse? What? Yeah, it's like a little thing. You move it around. It moves the pointer. Like, literally a lot of these innovations happen from six.

these radicals, you know, coming up with really interesting ideas. This weekend startups is brought to you by LinkedIn Jobs. A business is only as strong as its people and every hire matters. Post your first job for free at LinkedIn.com slash twist. DotTech domains has a new program called Startups.Tech, where you can get your company featured on this weekend startups. Go to startups.com slash Jason to find out how. And SuperGut is the only nutrition brand, clinically proven to improve digestion, balanced blood sugar, sustain energy, and manage weight. Save 25% on their delicious shakes, bars, and prebiotic mix at SuperGut.com with Code Twist. All right, everybody, welcome back to this weekend startups.

As you know, AI has had a blistering year or two here in Silicon Valley. And one of the reasons is we have so many amazing developers and the tools for developers to build interesting stuff have gotten better and better. And one of those tools that you hear about all the time is Replit. R-E-P-L-I-T. In fact, I decided to get into coding

and I was talking to my friend Sonny, and he said, oh, I just open Replit, do this, blah, blah, blah. And so I started playing with it, and I started taking their 100-day Python course got to date three or four and decided, you know what? I don't have the time to do this well. It's going to be incredibly frustrating for me.

But it's super inspiring to see how many people are getting into coding and how these IDs, and we'll talk about that in a second, have really made that possible for many more people. I'm John Mossad, is the co-founder and CEO of Replit, and he joins us today. I'm John. How you doing? Good. I'm doing great. Good to be back on the show. we actually did the show a long time ago.

We did? That's the first time this has ever happened. When did we do the show? In 2018 or 19, we're... Oh, my Lord. Baby company, small startup. I mean, at the time, there was like tons of skepticism, whether what we're doing is useful,

whether people want to use it, or this idea we're putting out that a lot more people would want to learn how to code, a lot more people would want to make a career out of it. Oh, right. Now I remember you came on during the pandemic. Yeah. Right.

You just raised your Series A. Oh, yeah, 2020. Early on, 2020. It's 2020. Yeah. So IDs. Why are these important? You know, when you're, for any type of job, there's an environment that you need for that job.

That's both in physical life. You know, if you're doing something with your hand, you need a workbench. and software, you know, if you're an accountant, your environment perhaps is an Excel spreadsheet or any form of the new spreadsheets now coming out. If you're a developer, your environment is either some kind of homegrown combination of a Linux terminal and an editor

and homegrown scripts and all of that, which traditionally been the case, or what's called an integrated development environment. And that's basically bringing together a lot of tools, that make it easy to write, debug, test, and run software. And a lot of features are starting to get added to that, I think since the last time we spoke,

co-pilots. You call yours Ghost Rider, and training, obviously. A lot of different features get added to this. So before these IDs, people would do what? They would have just a collection of tools on their desktop? Yeah, basically it's, it's, you know, the traditional view on this is that I just need an editor. I just need a place to type code on such a hardcore coder.

I just need Vi, you know, Vim or Emacs. And then, you know, I'm just going to like run my code and run compilation of my code. I get whatever errors out of that. I'll go back to my editor and fix that. And that workflow kind of worked for a lot of people. It's sort of hard. It's hard to figure out, it's hard to figure out all these tools. but that became for whatever reason the industry standard.

And at some point, I think Microsoft would probably, I'm sure some history buffs can correct me on this, but Microsoft really mainstreamed the idea of an IDE, and Visual Studio was one of their bigger products early on. And so they put a lot of effort into making the development experience on Windows. And that's something, by the way, it's missing in the history of Silicon Valley, like how much of success of these companies, Apple included and Microsoft, is about really building the best development in the world.

Apple puts a lot of work into Xcode, into their SDKs and their APIs, and Microsoft has done the same. Now, when the shift happened towards more open source and Linux and Unix, we lost a lot of that. And we went back to this hodgepudge of tools, figure it out yourself. People would use a text editor essentially. Vim's a text editor at a school, basically. It's like a stripped down word processor. Exactly.

And you could add things so you can script it, whatever. Some people enjoy that process. But it really sort of hindered a lot of people from accessing this field because the first thing you see is this blank terminal thing. It doesn't look intuitive. It doesn't look like anything you would want to do anything in. And so Replitt's initial product was that,

not only we're going to make this IDE on top of open source software, everything from the ground up is open source that we're building on, but also we're going to put it in the browser. We're going to make it so easy. You're going to go again like your friend. Anyone today is interested in kicking the tires doing something with software. You just go to Replit, sign up in a few minutes, and you have an environment and you can start coding.

Yeah. And the ability to have a co-pilot. and to debug has taken some of the harder aspects of being a coder, especially if you're a solo coder. It's changed things dramatically. So maybe you could talk a little bit about the impact of, I think, I don't know what you call debuggers that explain why this is wrong and how sophisticated they've gotten.

It's not just reporting back like, here's the era. They've gotten more sophisticated of that. So maybe a little bit on both sides of that, the ghost rider, the co-pilot, and these more sophisticated debugging tools. Yeah, again, going back to Microsoft and actually some microsystems and if you know, it is an investor in DrapLet right now,

but there's a lot of these bigger companies, what they've done with Java or C-sharp or these development environments, is they built what was called Intellicence at the time. It's sort of this classical AI that figures out the symbols in your application and gives you these autocomplete tools that help you kind of type less.

and for the computer to help you kind of debug your code and give you some intelligence. Of course, it's all classical intelligence, so it's all symbolic, right? So it's all algorithmic. We're trying to find, you know, we're creating this parse tree.

We're trying to find the structure between the code. It's really, really hard stuff. And I worked on this stuff for a long time at Facebook and other companies. The thing that was missing this whole time is that actually, you could represent code statistically. And you can apply the same methods we apply in natural language,

so NLP, natural language processing, on code. And this was not obvious a lot of people for a long time. There was a paper that I think was sort of a seminal paper in this era, called on the naturalness of software in 2012. And it presented the idea that you can actually look at software as sort of like natural language. It's very repetitive. It has certain structures. it has certain semantic.

And then you had the deep learning revolution. People started applying natural language processing, applying deep learning on natural length processing. And then you get into what happened with GPT, starting with Transformers, BERT, and then GPT2. That's when, to me, it started looking obvious that, okay, these tools are going to completely disrupt how we write code.

It's going to change the mechanics of how the machine can help us write code, how we can become more productive. And so we actually started building on top of that really early on. And we were like the first major company outside of Microsoft to release a co-pilot type product. And we did it in a way that we didn't actually rely on Open AI. We actually did an open source model. And we trained our model and we open source it as well. And that technology now has become accessible such that really anyone can do it.

So how does it change software? Well, for one, our autocomplete tools become a lot more powerful. You type a little bit of code, you type even natural language command and AI based on its training, based on reading all of GitHub and all of open source code out there. It knows that every time you are importing this library and you're kind of doing this function that way, that you probably wanted to write the rest of the code in this way. So it'll present a suggestion. So that's the sort of co-pilot interaction, but it doesn't stop there. So that's really a typing aid at the end of the day.

It's really helping you type less. Creating a job and finding qualified candidates. It's so time-consuming. Don't I know it? I'm trying to add five positions right now because things are going so well at the launch fund. But you know what? I have a secret weapon. and I'm going to share it with you.

Lincoln Jobs. They're about to hit a billion users over at LinkedIn. So just think about all the insanely qualified people that are there looking for work. You just go post your role on LinkedIn and you will be 100% certain. That's got access to the most qualified candidates. And guess what? First one's on us. That's right.

First drinks on us. Go to LinkedIn.com slash twist and post your first job for free. You got nothing to lose. And that will give you that purple ring on your profile. You see that? Everybody's got the purple ring now. That means everybody in your network knows you're hiring. They click on it and you'll get those friends of friends, right? Those are the really high quality leads that you want applying to your job because there's someone in your network who can vet them. So I did that for launch and I do it for inside and we found so many amazing people at our company. When you think LinkedIn jobs, I want you to think better candidates faster. Let me say that one more time. Better candidates faster. LinkedIn jobs will help you find the qualified candidates you want to talk to faster. Push your job for free at LinkedIn.com slash twist. That's LinkedIn.com slash twist to post your job for free terms and conditions to apply. And for people who are not coders, you experience this now when you're using notion or Microsoft Word or Gmail. It just here's the next possible words. Would love to and it's

invite you to whatever and it's studying what you've been writing emails about. So it's for, it's precisely for you. And it's based on the whole corpus of, you know, people writing. on the open web. So fascinating how much more impactful it is in coding, right? Because coding is a higher value typically than what people are writing in email.

And it's been harder to apply this technologies on code. So it really took until GPT2 and GPT3 for this to actually become possible because we've had these suggestion tools for a long time for natural language, but not for code. But once these models

got big enough and intelligent enough, now they're able to complete code, because code needs to be structured, needs to be parsed, needs to be run, needs to be correct, whereas natural language doesn't have a lot of these properties. But it doesn't stop there. And I think that's useful,

and it really made a lot of programmers a lot more productive. But typing is like perhaps 10% of what we do as programmers. Most of the time, we're staring at error logs. Other times we have a crash in production where we get an email, we get a PM kind of coming to us and saying, hey, like, you know, the software's crashing and it fixed that.

We get a bug, whatever it is. And this is, I think, where it gets really exciting because this is where we can start to get multiples more productivity. I think with a co-pilot interaction, at best, some people are saying it's 80% making me 80% more productive. But can we get 2x, 5x, 10? 10x more productive. This is where leveraging AI to help you through the entire software lifecycle comes in.

And this is where Replit, I think, is very well positioned because Replit is fully bundled environment. So from your first line of code to your first deployment, to your scale deployment. So we take the entire software lifecycle and we package it in one place. Now we can inject AI exactly at the right places to be. be able to help you in this crucial moments when you get it, when you get this crash in production. Now, the next stage after that, just to mention the next stage after that is agents. And, you know, agents are still primitive right now. But we, there's a, let's define, that was exactly

where I was going. So let's define an agent for the audience and how that might work, you know, if you were building, you know, Uber today or you were building Airbnb today. Right. So the way we use AI today in this co-pilot fashion and this auto-complete fashion, is that I am doing the driving as a human. I am doing the thing, and there's something that's augmenting my behavior, such that it's making it easier to code or type or do something. Agent inverts that a little bit,

and you give the agent one instruction, one higher level goal, and the agent is responsible for breaking down that task into subtasks, and for going and executing those tasks, in an autonomous fashion. And what that commits you to is that the agent need to actually be able to recover from errors. It needs to be able to recover from a lot of edge cases that you face in the real world.

And some people talk about agents with a human in the loop. But for the most part, what we want is agents to be fully autonomous to be able to do work on our behalf. And for example, I should be able to tell Siri like a high level goal such as, you know, my wife and I, you know, today's our anniversary date and I want to go out to a nice restaurant. You should be able to figure out, okay, I need to go research nice restaurants in Palo Alto. You need to figure out like the budget on those.

I need to call them one-way one and get a reservation and then report back to my user what exactly happened and the flow that happened and whether there were errors or what have you. So this is where agents get a little more exciting than the way we're using AIs today. And in the case of building a business, building a startup, it could be, hey, I want to build a marketplace. Okay. Buyers and sellers, take rate. It would know a lot of these fundamental items.

And then it would know what kind of code. Oh, you need a profile page. Oh, you know, what's on the profile page? Describe it. Oh, it's a home. Okay, what are you going to do with the home? Sell it, rent it. You know, do, you know, home maintenance tasks.

You would have a whole bevy of possibilities that the agent would be then able to go build. How far, is that even on your roadmap to start thinking like that? You're building an app and you're dealing with non-developers. One milestone. Yeah, one milestone we're trying to head this year is to be able to tell an agent, I want like a login page or something like that. just one feature that's very concrete that developers do every day. And for that agent to be able to fork off your environment,

which is very easy to do in Replit, and for you to be able to watch it, and go and try to build that and run into errors or recover from those errors, and then come back to you with the code saying, here, I finish it. And then there's a cursive loop where you say, okay, this is good, this is not good.

It's sort of similar to how you're dealing with a junior engineer. Yeah. So let's try to automate that really simple, like junior engineer interaction you have and I think from there it could progress really rapidly the thing you talked about

where I can tell it go build me a marketplace startup feels very far away. It feels it feels probably five years away. But who knows? I mean the way these things have been moving is it's you know what I tell my team

is that we just need to be very reactive to the changes in the AI space. And we're at the bleeding edge of it and we make a lot of the, some major contributions to the field as well. But who knows? Like, maybe there's an unlock later this year or next year that the vision that you just put out could be possible next year.

But my estimation currently is that we're going to get asymptotically closer to that over the next five years. Yeah. The thing that's really obvious, you had user registration is, hey, like, there are some best practices here. One of the best practices is using a magic link. Like, I think Slack was really pioneered, or maybe it was Craigslist before them, did that as well.

Phone number. Do you want to use Google, Facebook, Twitter, X, you know, whichever, to do those logins? And so it's something you're uniquely qualified to do because you get to see, I don't know how many millions of developers, hundreds of that. I would suspect hundreds of thousands of developers on Replit now. Millions. Millions of developers in Replit, which is pretty impressive. You know, that alone gives you an insight into, oh, yeah, people are doing the magic link thing.

Oh, they are using phones. And everybody, yeah, go ahead. And also we have the Bounties platform. I remember that you've also checked it out. Yeah, I did. Pretty amazing. Explain the Bounties program, yeah. Yeah, so, you know, because Replit just on board.

this huge number of people from all over the world onto programming, one of the things that I found challenging when I talk to these developers is that landing their first job is actually quite difficult because people want to experience on all of that. But a lot of these people, when you look at their profiles, when you look at their replica profiles, that are actually better than a lot of Silicon Valley engineers. So what is the best way to show that they're better

is to create a platform where they have real work for real customers or real clients available on that platform. At the same time, they can make money and maybe that could be a business for us that we could make money out of. But that's really not the priority here. The priority is to showcase that we have

a huge number of developers around the world that are AI powered now and I can build amazing things really fast and really cheaply. Yeah. And so we put it out and at a matter of months, it reached a million annualized GMV. We should have sort of head on something. It's basically we had this backlog of developers that are hungry to do work and show

that they're able to do work. And that created this explosion. Yeah. And you have surplus cycles. There are developers who have, you know, maybe it's a Saturday and, you know, whatever their plans fell through. they got four hours and you put up a little challenge and I'm like, that's easy. I'll take that 500, it's free money. I'll pick it up off the ground.

Yeah, one of the things that was surprising was I didn't expect like big tech employees to be engaged in that, but one of our biggest bounty hunters is this sRE at VMware that does it on the weekends and on the plane and it's a game for him and he's making all this all this money. But for us, it's also a very valuable source of data to be able to make our AI better as well. Like we, you know, when you're able to match project description to output of work, you know, we're going to be able to make the agent thing work a lot better. And also, we want the developers to the bounties developers and developers on Replit to be able

to use that to increase their earning potential and to become more productive. So it is, you know, one thing from the start, you know, at Replit, we've, we've, we, we've been very clear on is that we're not one of these companies trying to build AGI. We're trying to build ADI, artificial developer intelligence. We're trying to augment you as a developer, as a solo entrepreneur, as someone who wants to build a business online, someone who wants to participate in the global digital economy. We want to supercharge you such that you, it feels like you have an army of developer, developers working for you.

Stop what you're doing right now. I got great news. dot tech domains is giving twist listeners a new platform to show up their startups. And that platform is this podcast. This week in startups, it's a new segment we're calling pitch J-Cal. I want you to go to startups.com slash jason, startups dot tec slash jason to apply. There's only one rule. You have to have a dot-tech domain name to get featured. This week I received a great pitch from hivepower.com. That's hiv-e power. tech. That's hiv-e-power. tech. which you can see in the video right now if you're watching, Hive Power is a Swiss startup.

They're working on transforming energy management and their software platform helps businesses optimize EV charging, distribute energy more effectively and more. Energy prices have gotten out of control. They're volatile. So Hive Power is saving businesses money by optimizing energy efficiently, like all these different times when you can get discounts on your energy. It's a huge market and they got a great domain name. Check them out, hivepower.tech.

And if you're interested in getting feature on this week in startups, go to startups.com slash jason and apply today. That's startups. tech slash jason. It's interesting. While you were talking, I was like, huh, I wonder if chat GPT could actually, you know, build a plan.

And so I asked chat GPT, please build me a feature set for a modern day marketplace for folks to rent vacation homes and cars. So certainly, here's the features. I have in my instructions with chat GPT4 that I like things and tables with links. And it says, here's the feature category, here's the feature description,

here's the stakeholder, and here's your data source or API. User registration, enable users to create an account via email, social media, or Google. And to your point, you'll have that done by the end of the year. I actually believe you'll be able to do that. User profile, allow user to upload photos,

ID and basic information. Search engine with filters like location price rating amenities, calendar availability, payment gateway, listing creation, pricing enable dynamic pricing based on demand location and time.

That feels like something actually, a pricing engine could be something that could be built into Replit, inventory management, a review system, that's a layup. Star ratings also a layup. Dashboard, give administrative overview,

booking to revenue, fraud detection, that would definitely, I mean, there are party services that do that. You can plug in analytics, third party services. I mean, that's after one prompt. You know, if we sat here and we started prompt engineering this and then connected it to Replit a year from now, I don't know, maybe half this stuff gets done or at least some portion of it. Actually, today, even Ghost Rider, I mean, I'll,

I can show you a quick demo of what it looks like today. So here, I just started Node.js. Ripple, what we call projects. And basically, I have just one file. You can start by writing code, but I can also generate code. Let's say, make a login page with Express. And I'll start writing the entire function for me. It's still not an agent in the way I describe it,

because I'm looking at what it's doing and I'm reviewing what it's doing. What we want is for it to actually be able to do all of that, run the code itself, be able to deploy the code, and be able to get feedback on it. But right now, it just made this login page right here, right? And okay, it's a start. And we can iterate on that from here.

It can just say, let's, I can right click, edit code, just say, add some CSS to make it nicer. Yeah. And I can iterate with it, but we want to get to a place where it does that sort of iteration on its own. Yeah, it's pretty slick already when you think about it. And there are so many problems in the world, you know, everybody gets to like, oh, it's going to replace developers. It's like, yeah, until we have a solution for every problem that humans face, I think developers are going to stay busy. What this really means is the typical startup is going to release features faster and make them better continuously and faster.

Just like open source was a catalyst, just like IDs were a catalyst, just like cloud was a catalyst. All this stuff just makes everything go faster. And the average app today compared to the average app just 10 years ago, you know, in the app store, in the app stores only whatever number of years old it is now. maybe it's 15 years old. When did the App Store start, 2007 or 8? I think 2009, because 2009, you know, remember Steve Jobs didn't want an App Store.

He thought the web would be enough of an app store. Yeah. Yeah. Not quite. By the way, we just added this feature to be able to deploy your app right from Replit. So this means you're in competition with AWS.

When you do this, you'll be like a hosting company in addition to, yeah. Our vision from the start is to do the end-to-end process. And again, this is where we can do interesting things with artificial intelligence, such that the AI can deploy. The AI can figure out the errors in production and iterate on them as well. But, you know, I think in the future we might even partner with AWS right now with partnered with Google Cloud. So we're building a lot of these features on Google Cloud.

And I think the space enough for a lot of players to, compete. What we care about is really that polished end-to-end developer experience. And we've seen no one do that. Everyone picks a niche to be really good at, and we want to be good at the end that thing. I mean, eventually, you could be cloud independent or multi-cloud. And when you click deploy or you want to put into production, they could compete for your business. It could give you three different prices for this, right? And you could list them in rank order of what it will cost to run these here, right? And then I believe there are, you know,

I get pitched on this startup all the time, hey, we'll examine your cloud bill and we'll optimize it by putting you on, you know, I don't know if they call them second tier servers, but maybe not as powerful servers and stuff like that, that are unnecessary for your job and just optimize your costs while not compromising your performance.

All that's a possibility, yeah. Yeah, and that's it. We got a deploy going. And you get some nice analytics and logs and everything built right in. And, you know, the deploy button could also go to your VPS or to your internal deployment. I think once we go into the enterprise in the future, if they have on-prem thing, we could also package a container and put it on their on-prem as well. just mean you compete with like GitHub and those kind of repositories now

and or you know let me show you the community in addition to a lot of the bounty stuff that we talked about which is here you can see all the tasks you can start doing there's also a lot of you know

a lot of apps inside the replica community that you can check out right like you can go to this app It's called Oracle GPT. It's a chat assistant. I can go look at how the developer has built it. I can run it or I can fork it and change it. So it adds a layer that GitHub never really added, which is how do you actually run this stuff?

And not just view the code and download it locally, but I can actually bring it into my environment and mess with it and publish it and we have this community where people vote on things and tip each other. And so it adds a layer that, you know, I think GitHub never really added.

What about Hugging Face? Are they a competitor to yours now or co-existor? How do you look at hugging phase? Are you a lot of developers, you know, sort of starting there? Hugging phase, I think they're on Replit. They have some examples on Replit that they built and you can like fork them to play with them.

You know, Jason, the way we think about competition and the way we think about just the landscape in general is that we really want to be the place where people create. And so if you help people create, we want to partner with you. So if you're Hanging Face and you're helping people create AI by hosting models and by training models, then developers on ReplaS should be able to use those models right from HangingFace. So in a lot of ways, Reppler becomes this positive sum playground for everyone to play in. And that's really our vision for this company. Getting more people to become developers and making it easy to onboard, have you given thought to that?

I know, I know you had this great, like, I don't know if it was 30 days to learn Python or 100 days or whatever. Yeah, it's 100 days. Yeah, 100 days. And we have it in Hindi, too. Oh, wow. That's fascinating. You know, because when I went, I did the first couple of days,

I was like, wow, this is incredible. It is literally walking me through it with a video, with the code. It's almost like a little bit of a game. Is this working to getting more young people or even people who are transitioning into being a developer? Is it actually increasing the number of developers in the world? Or do you kind of think, like some people do, hey, developing is a certain type of person digs it and it's not for everybody. Yeah. So I think about it as sort of these concentric circles, right, where, you know, at the very

inner part of the circle, you have those, let's call them platform developers. Typically, you're living in Silicon Valley or, you know, or Seattle and you're working at it on AWS or Google Cloud or one of these like, you know, hardcore kind of platform companies. You're writing low-level code, right? Then you have a lot of product engineers. Those product engineers are founders, are people working at startups, or full stack engineers.

And I think that that's a lot more accessible than people think it is. If you're a product engineer, you are basically, your day-to-day job is waking up in the morning and translating some user demand or user story or something that PM told you to do or a designer told you to do,

or you're a founder, you're trying to meet some, demand in the market into code. And code is sort of incidental to your goal. Your goal is to put software out into the world. Yes. And I think a lot more people can participate in that in a lot of different ways. One is we can make code more accessible and more learnable, and that's what we're doing.

Two, I think there's a lot of opportunity for AI to help with this, such as AI is writing the code for you. Yep. Three, I think some version of no code might work. I'm not a big fan of the way people are doing it right now because they're basically recreating the programming environment using blocks and things like that, and that's not very scalable.

But I think someone will figure out a no-code environment that actually makes sense. Interesting. So let's double-click on the no-code piece. Right now is somewhere between a whizzy wig, what you see is what you get, type environment, you're dragging things around, and then it abstracts the code,

and then it writes the code on the back end. You don't get to see it. Nor do you need to. And then there's like sort of scripting, hey, if this happens, do this, and it's kind of light, but it is building a whole facade on top of some code base. Yes, is I guess how this is all working.

And so you're proposing what? Yeah, so the way it works is that they have these pre-plog programmed visual components that you are kind of composing together, broadly is how it works. And you can double click on one of those components and edit some of its features or even write code and this is where you get into some low-code platforms. The problem is that what you want out of your computing environment is to be Turing-complete.

So Turing-completeness is the idea that is the idea of a universal computation, is that the computing machine can run any other program that any other computing machine. machine can run, right? So this is a very technical concept, but it basically means that every programming language today that we have that's in use is Turing Complete. So Turing Completeness is a necessity for this thing to actually encompass all sorts of features that you need in order to build a real world software. Most of these environments are not Turing Complete. If they try to be Turing complete, it becomes really hard to manage. It becomes really wild. It's actually the visual components actually get into the way, in the way of actually producing the software.

So it's a scalability problem. So you end up either simplifying too much and you're not churing complete or if you want to be cheering complete, you run into a scalability problem. So that's really, it's sort of a, that's why you see a lot of these No Code platform grow really fast and then they hit a ceiling. And then they stop growing a lot, even as a business. It's because the demand, they can't. They can't expand the demand much more than that because their users are heading these ceilings with their products.

You have heard me talk about my weight loss journey and the role that SuperGut has played in it. SuperGut is just an amazing product. They were a huge hit at the All In Summit. I love their bars. I love their shakes. In fact, here it is. I was just drinking this shake. My wife makes a beautiful shake for us every morning. I love the chocolate brownie bar. And they have unflavored prebiotic mix that you can add to anything. My wife puts it in my coffee. You know, she's trying to keep me healthy. She wants to keep me around for a little while. I can't blame her.

I'm a fun time. There's a reason why SuperGyde is so good for weight loss, okay? The SuperGut product, you may not know this. They mimic the effects of OZemec by boosting your GLP1 hormone. And this will help quell your hunger, and it's going to boost your metabolism. That's the combo you're looking for. Products taste great, tons of great flavor, strawberry, chocolate, all the stuff you love. But here's the thing.

This is all proven scientifically. the Sultan of Science himself, David Freiberg, is the co-founder of this company. They conducted a placebo-controlled clinical trial with Stanford. You may have heard of that university last year. And that's been published in the medical journal, diabetes, obesity, and metabolism, the results. Amazing. Participants lost weight, just like me. They lowered their blood sugar and they improved their metabolic health.

And also helped with digestion. So it tastes great. It's going to help you lose weight. If you want to improve your gut health, you want to drop a few pounds, or you just want to feel sustained. energy throughout the day. Do what I did. Go to supergut.com and use the promo code twist to get 25% off. Supergot.com use the promo code twist to get 25% off.

What a great deal. So making coding more like talking to the computer, I guess, is where this is going to end up. And using a natural language known as English to program is how many people in the AI space have been proposing this will end up. Have you used a code interpreter? I have not. You should check it out. I think you would like it.

So Chad GPT has this mode called code interpreter where you're talking to chat GPT, behind the scenes is writing code and running it to help it to help it answer questions. I have used it, yes. I've uploaded CSVs to it and then asked the questions about the data that was uploaded because you can use it for that function as well. Right. Yeah. And that shows you the promise, right?

I mean, there is something there. It still hallucinates a little bit and does some silly things, but it seems possible. And I think talking to, you know, I've been reading the new Steve Jobs book, make something wonderful. I didn't know about that book. Who wrote that? So Lorraine Jobs put together all of Steve's writings for all his lives and in a almost chronological way

he kind of lays out. Oh, wow. His life's amazing. It's phenomenal. Oh, that's a free e-book. Look at this. This is amazing. Yeah.

Steve Jobsarchive.com. Yeah. Make something wonderful. Steve Jobs in his own words. Wow, it's available in Apple books or you can just download it for free. I had no idea about this. Okay, so continue.

What did you get out of it? I remember like a big student of computing history and Silicon Valley history because I think to understand where we go from here, we have to understand the past and how we arrived here. There are a lot of amazing things about how we made computers and technology. There are a lot of drawbacks as well. One of the drawbacks I'm trying to fix is that

we've created a generation of computer consumers, not computer programmers and users, right? Think about how most people use computers today. It's basically a consumption device. You're watching tech, talk and YouTube and maybe it's a communication device. Maybe you're talking, texting. But the original vision for computers is that it augments our intelligence.

It actually computes for us. And we've really lost a little bit of that. And so I think Steve Jobs was phenomenal as a designer and entrepreneur and really the goat in a lot of ways. But some of the decisions they made were obviously guided by the. this idea that people can't compute or people can't program. Actually, there's a line in the book that I found, which I kind of went looking for, which is initially they were having trouble explaining the value proposition by the Apple

2 because you buy this thing and you open it up as just a terminal. You can't really, you don't know what to do with it. And you have to learn basic in order to program it. And what Steve Jobs arrived at is that actually people shouldn't learn how to program and we need to build UIs and GUI, because people should just use computers as opposed to program computers. And that decision put us on this history path of actually users and programmers are separate instead of being one.

And I think there's a vision of the future is a lot more exciting, where anyone can be a creator, anyone can use computers to their full power. Yeah. And it's, I mean, you see this in generative AI, right? So we were talking about coding this all time. But, you know, if you wanted to be an artist and make a logo, you wanted to make the next logo, speaking of Steve Jobs, you know, this is something that you went to art school for a long time.

But if you start talking to stable diffusion and some of the other tools out there, Dolly, and you said like, hey, what are the most iconic logo, or yes, chat GPT, what are the most iconic logos of all time? Then you start using stable diffusion to make something just based on the creativity in your words, you might start to find things or ideas that you wouldn't have thought of. This is where I like the writer strike,

when you look at the writer strike and they're saying, stop, we want to ban chat GPT, I'm looking at it and I'm like, no, you actually don't. You're all using it anyway. And if you go into any of these language models and you start just brainstorming ideas, it's just a more efficient way than reading 10 books,

you know, about science fiction, let's say. You can just see, see which, you know, tickles your fancy in terms of ideas. And it's going to come up, like you said, with hallucinations or bad ideas. But that's how you get to great ideas. You get through all the bad ones, right? And you have these jumping off points where, you know, something sparks your creativity. Oh, yeah.

What if Decker is a replicant? Oh, that's a great interesting idea. You know, and Blade Runner, like, oh, wow, spoiler alert. You know, these ideas typically come because people are brainstorming, smoking a joint, or freaking some beers or going for a walk. and, you know, something just pops into their head. Like, it's almost like the hallucination of these language models. Is a feature.

It's a feature. It's like mimicking these weird moments when people are tripping out, you know, from their cold bath or from taking mushrooms or going on a hike or taking a shower, whatever got their creative juices flowing. I think that that might be what's literally happening is where, you know, as we recreate this intelligence, starting to realize, like, maybe it's not that. unique, you know, how our brains work. Maybe there's some magical aspects to it, but I almost feel like we're figuring out how our own brains work when we build these things. One of the reasons I got into computers as a kid was because the, you know, I've been always into computers, but the thing that really attracted me to become a programmer and eventually

want to be able to do AI is because I wanted to understand the human mind. Like, it was, like, it's one of those questions that I can't stop thinking about. What makes us us, what makes us thinking beings? And I think once you understand all these things, you kind of start to potentially understand the meaning of things, which is, have a very existential drive in me to actually try to understand, you know, what makes humans humans. And I think there's what's called in philosophy, the hard problem of consciousness. we can't really probe consciousness

because it is inobservable from a scientific point of view. You can have a complete description of the brain, but you can't really understand what's going on inside. Whereas now that we have these machines, we can actually try to probe them from the inside in an interesting ways.

Actually, a lot of neuroscience now is making its way back into AI. And one of the ways people are playing with AI is what happens if we turn off this neuron? What happens if we turn off this layer? What happens? And I think they're called ablations. It's kind of like you're running this process to be able to judge performance

based on different ways the neural network is structured. And I think you're right. That starts to give us some understanding of how intelligence works. And I think that's really exciting. And it'd be great. It'd be really great. If philosophers, if a lot of people from the liberal arts side of humanity is actually interested in these things, because that's an exciting moment.

It's kind of disheartening that a lot of artists are offended by it or are not really excited by it, but hoping they would come around to it. Yeah, and we have some corollaries in psychology and brain science. You know, you have these, what are they call it, split-house? hemispheres where people have had their hemisphere split or, you know, twin studies. You have all these like interesting, you know, people have different damage that has occurred to their brain tragically. And we started to figure out just through those biological circumstances, you know,

some of the oddities of the human brain and consciousness and how we perceive the world, right? Have you seen this really wild video where, you know, if you have epilepsy, one way to cure it as you sever the hemispheres. Yeah. And so that way the epilepsy doesn't travel back and forth. And there was a video on YouTube about a guy with a severed hemisphere, and they were trying to kind of probe and understand what happens to humans when we actually sever the left and right sections of our brains.

And the experiments were fascinating. So basically, they gave each eye different video feed. I think the right eye is connected to the left hemisphere and the left eye is connected to the right hemisphere. One of them is the language center, I forgot which, but you basically give an instruction to the non-language center, go to this side of the room and grab this other thing, and you don't give that instruction to the other side,

the side is able to talk. The subject actually stands up and walks over to that side of the room, and the experimenter will ask him, like, why did you go there? And he would hallucinate. Wild. He doesn't have in that side of the brain, the talks, doesn't have the information to be able to say, I was told to go there. And so he hallucinous, oh, I want to go to the bathroom. Yeah.

And that's really fascinating. That's wild. Yeah, he's basically taking the next guess like we were talking about with Auto Complete. Like, why would I get up? the bathroom's in that direction and so that's the next possible word you know, I'd love to invite you to dinner, the movies,

like what are the possibilities here? Could be getting a glass of water, I could be going to the bathroom. It's probably one of those two, yeah, it's really fascinating how quickly this is all happening. I'm curious, your thoughts on like, what are you

seeing internally with your team? I don't know how many developers you have on your team, but you got a front row seat, you guys are obviously all true believers in using, you know, these ghost riders and co-pilots and using AI. What has the last year been like in terms of your team's productivity and how much are you embracing AI and running Replit?

You know, the, again, on an individual developer perspective, we see our developers are a lot more engaged, a lot happier and a lot excited about writing code because, you know,

it just removes a lot of, lot of the drudgery of writing code on a kind of a company scale level, I don't think we've seen a huge productivity improvements. And again, that's because the thing that is lost between actually writing the code and actually deploying a product, there's everything that happens in between that AI hasn't touched yet. And this is why we're really excited about it. Because even if you make the individual developer slightly more productive by making it easier to generate a code,

it still doesn't translate very well for it to be perceptible from a CEO point of view. In a solo entrepreneur perspective, it 100% is. And what we're seeing of our customers, I mean, we have this guy on Twitter actually, he built 250K ARR business on Replit, and he didn't know how to write any code coming into this. He just used ChapchipD and Replit and Rupplet and Ghost Rider and other tools to kind of figure out how to write the code himself. He doesn't even know how to use Git.

And he built this huge business for himself doing that. And obviously, that would not be possible without AI. We're seeing all these stories of sole entrepreneurs or small companies getting a lot of benefit out of this. But for it to actually move the GDP or even smaller move, like productivity of the tech industry as a whole, there's still a lot more work to be done. So people enjoying the job more.

Maybe some products are shipping faster. Are you seeing that yet? The products are shipping faster or maybe not? We were known for being super fast already as a company. Yeah, the product is nuts over there. Yeah. So, you know, it wasn't that perceptible. There's definitely, it definitely feels like we're prototyping things a lot faster.

And there's, there's a lot of prototypes flying around. More shots on goal, maybe. More shots on goal. That's right. And, you know, instead of getting in a meeting and talking about an idea, you get actually a prototype pretty quickly. Oh, that's very, well, see, this is so powerful that maybe we're like, you and I are agreeing so quickly on it, but it's such super profound.

If people who are coming to the meeting instead of saying, we should build this for saying, I built this. And then two other people say, I built this. Well, now you as the CEO or the management team, we're going to look at the three ideas and say, okay, let's test this one. Then we'll test this one. Yeah.

And then our designers are shipping these replet links between each other. And I see a designer, instead of waiting for the engineer to do this thing, they'll just like post like, here's the new replet link for my change, for my CSS change or my visual change. Interesting. So definitely seeing that velocity of, ideas and prototype inside the company. It's almost like if we were running a restaurant,

you would be having people instead of like coming and say, hey, I want to try a duck dish or whatever. They should be coming and be like, here's the finished dish or here's like a approximation of my new deconstructed, you know, peaking duck. And then everybody just tasted it. And yeah, you know what? We should keep pulling that string and see where it leads us.

I think that's like, if you think about the nature of startups and finding product market fit and product velocity, that's really a game changer. That's perhaps more important than shipping fast, to your point. The other day I was talking to Jack Dorsey came and visited a upload. He's a repleteer user. He's a really excited about it. And he said something that in the early days of Square,

every time they wanted to trial and you design, they would have to ship the design to China to fabricate it and come back. And then they started investing in 3D printing. and that increased the product velocity so much such that, you know, designs were changing every day and they were getting the right thing faster. And he equated what we're doing to software as to 3D printing. Yeah, that makes total sense. And if you think, like, he's been such a great entrepreneur, I just had Ruloff from Sequoia on the program.

And they started with, you know, small to medium-sized businesses, even people in flea markets, It's just, you know, being able to take credit cards for the first time instead of cash and sending people to a cash machine with a $7 fee in some bodega or something. They're like, yeah, you can just take any credit card and fire that up. And then he was so bold. He's like, you know what I'd like to do is I'd like to create a second product inside this company just for consumers, cash app, boom. And now cash app is this gigantic business.

It's huge. Well, he's the third business, Jason, that nobody's talking about. They're Bitcoin business. Yeah, explain. Yeah. The last quarter, I think they saw $3 billion in revenue. Obviously, it's still not monetizable revenue, but $3 billion in Bitcoin. Volume, yeah.

Volume, which is wild. And now they're building their TBD product, which is still TBD in progress. But the idea is to just make it such that, you know, you can move Bitcoin across the globe really easily. And so that Square and Cash App. can sit on this Bitcoin rails and be accessible all over the globe. And I think that's a really huge vision. Yeah. And that really is.

Like shots on goal, you know, and that's the thing that's going to just keep increasing with tools. That's what gets me excited about technology. I mean, I think the tech industry got a little boring towards the last decade, towards the last part of the last decade. last part of the cycle, yeah, like 2017, 18, 19, everybody's making, like,

you know, like the next app that's already existed 20 times. Yeah, you see a lot of that. Yeah, that's thing. And it became very formally, and I think now we're back into a period of innovation where we're going to actually be able to lift

the common human being and be able to supply them with tools and intelligence and computation and really try to make the world a better place and kind of go back to this roots of like, you know, it just became cliche almost to say you were here and make the world a better place. But I think that's really why I'm here and why I'm doing what I'm doing at Replit. Yeah, you know, it's like, it's so wild when a new paradigm shift happens now that I've lived through so many of them. Like, I, I expect them to happen.

And then I just understand how they kind of go through the hype cycle and everything. whether it's mobile or cloud or crypto and now AI. And it's really like a chance to take things that exist in the world already and just make them phenomenally better because this new platform technology, technique, however you want to frame, you know, cloud computing, etc. And you think about something like YouTube, and I was talking to roll off about this as well in this episode.

That's coming out on the swing startups where, you know, he did that YouTube investment. And, you know, nobody wanted to invest in YouTube. Everybody thought it was stupid. People who had a lot of experience like Mark Cuban, we're like, yeah, I don't think Netflix and YouTube are going to work because it's going to flood the internet with traffic that it can't handle. And then, you know, the blind spot was, I think storage had gotten so much cheaper in cloud computing. And then bandwidth was getting more efficient compression algorithms.

Actually, you could let people upload videos and get as many views as they want. And just for a couple of low dollar CPM, be able to actually break even and then eventually make money, right? Even with the amount, think about the just insane. amount of uploads that happened on YouTube and how that was burning, just mountains of cash, mountains of cash being burned for storage, for Dropbox and for, at least Dropbox had, you know, a paid version and a cap, but YouTube didn't have a cap. You just keep uploading videos. It was a really sort of dangerous, crazy prospect. Now you look at AI, you know, it's like in Hollywood, there's a piece of software called Final Draft. I don't know if you're aware of that

where like people write screenplays and Final Draft. And, you know, had a startup. make a version of Final Draft. It's called Cyberfilm.aI. And they have this tool called Saga, and you can go make your own screenplay. And then make, what do you call them, storyboards and stuff like that? I told them they should call it storyboards.

And like somebody who feels intimidated by writing a screenplay, I can be like, I want to write something like natural born killers or reservoir dogs, but I want it to have an Indian cast and I want it to take place in India and I want it to revolve around this or that and kind of adapt it, make it your own.

I want to make a sci-fi version of reservoir dogs. Like, I'm down with that. Yeah, make me a sci-fi version of reservoir dogs. Let's, you know, start your creative uses or whatever. That'll be fun, actually. Yeah, we kind of actually... Sci-fi or a dog, yeah. I mean, this is what I mean, you look at Star Wars.

it was a sci-fi version of, you know, Kurosawa's work, you know, in samurai films and stuff like that. So, you know, why not? Let's go, right? And it's like a chance to just start again, over again with this new technology, you know, and you start happening on mobile too. I think Hollywood is getting remade

from the ground up with this technology. I just invested in this YC company called Kino AI. What are they doing? What do they do? you know, I'm going to explain in my own words because they'll probably use different words, but they actually make all the video archive at all these big studios

usable because they index them and they put a search engine on top of them. Amazing. They index every frame. And so you can go and search for, you know, every explosion and whatever studio and look at all the footage they had.

Yeah. And coming of this... It's a map software. Media, media, media asset management. So, like, CNN has all this footage from, you know, Anderson Cooper show, you know, and they may not have ever used,

they may have used 10% of it. The other 90% are sitting on a hard drive. It's not utilized. Yeah. Do you know Adam Curtis? No. Adam Curtis is really awesome, BBC documentary filmmaker.

Oh, yes. Century of the South, hypernormalization. Hypernormalization, I know, yeah. Yeah, basically this guy goes into the BBC archives, sits there for a year and looks at a bunch of footage and then cuts together a story with a voiceover and creates this amazing documentaries. And so imagine all this footage could be now accessed and indexed and repurpose and probably using generative AI also completed and remix and all of that stuff. So I think it's really exciting what's going to go on. If they will adopt it.

If they won't, they will get disrupted. If you haven't seen hyper normalization as a trippy movie. All this stuff is trippy. By the way, one about our industry that really is not, people have not really seen it, I think, is... Is it by him? Yeah, all watched over by machines of loving grace. And it talks about early optimism and Silicon Valley. talks about the cybernetic movement and about the ideas they had about how to structure

society and some of the cults they made and all that stuff. It's a really interesting documentary. And the name of that one is... All watched over by machines, all watch over by machines of loving grace. All watched over by machines of Love and Grace 2011. Wow. I didn't know about that one. I'm going to watch that. That's a perfect one for me to watch. Yeah. I mean, are we talking about like the history of this, like the technology industry. These guys were all on acid and mushrooms and coming out of the hippie, zippy movement.

And they really were opening the doors of perception. And then they were with computers at the same time. Yeah. And, you know, imagine you're tripping on acid. And, you know, you're also like using new technology for the first time. You're like, hey, I have some interesting ideas here. We should make a mouse. It's like, what, a mouse?

What? It's like a little thing. You move it around. It moves the point. Like literally a lot of these innovations happen from 60s radicals, you know, um, you know, coming up with really interesting ideas of what could happen.

I mean, Steve Jobs said that was major part of his journey was, you know, opening up the doors of perception. I don't know the full extent of his drug use or tripping, but it was not insignificant from what I understand. It was not. Uh,

there's a guy called, Bill Atkinson, who was part of the original Mac team, and he built the system called HyperCard, which is in a lot of ways, is early Replit. And we actually talked to him at Replit, and he lives here. And, yeah, he just talked about how it all started with a trip. Yeah, I'm reading his page right. He said he had an acid trip, and he built HyperCard. That is hilarious.

I actually didn't know that. You should have him on. I think he would come. Yeah, that would be dope. I remember HyperCard. It was like, here's how to organize. It's kind of like a database and here's how to organize information. It kind of had the concept of hyperlinks in it too, right?

Like you could jump from card to card. And yeah, fascinating. Yeah, this idea was in the air. There was actually, like, the web ended up winning, but the web was actually the least sophisticated of all the different ideas around hyper media at the time. Yeah. People forget even the term hypermedia. but this idea that in a way

there was a very cool piece of software called The Brain. And it was mind mapping software. And if you go to TheBrain.com, it still exists. And I remember knowing the founders from the 90s. And the product was really simple.

Here I'll show you. This was like a really trippy piece of software and I used it for years. I wish I had mine because you would basically, yeah, it would make a mental map. You see it there on the screen?

Yeah. And, you know, AI would do a really wonderful job of this of just taking our conversation here and then any other conversation happening on podcasts. And if we hit on HyperCard or Hypermedia, making jumps between all the pockets, it'd be a great startup, by the way. You just take all the podcasts in the world and make a brain map of them. And the idea here was people with Quick Keys and Shennie Jardin and I used to play with the software and we would sit and work and put it up on the screen

when we had ideas for conferences and stuff like that and we'd make these maps where you would jump from like product design and then you click on something over here and click on something over here and it would just load web pages associated with each of the neurons if you think about this like a neural net or your brain and it's kind of forgotten mind mapping but jury that's Jerry Colonna, Jerry McCowski from Lease 1.0 which was Esther Dyson's newsletter.

Jerry McCalsky had kept creating Jerry McCalsky's brain and he would publish it publicly of all the things that were in his brain, which is just a mind map. You know, there was a recent resurgence around Rome research, the backlinks and things like that. And they're great, and these softwares are continuing to develop and get better. But one thing that I found unattractive for my usage is how laborious it is, like how I have to create these backlinks, right?

Like you have to do like, you know, double brackets to create a new topic. Yes. And then you have to do all that. I'm like, it's just do it for me. Just use D. I mean, the co-pilot should be doing this. Like, you're going to be better at it than I am. And you just look at this here.

It's like, this is Jerry McCalsky's brain. I don't know if he keeps up with this, but it's like news about using my brain, using starting points and understanding Jerry's brain. beginners, whatever. Here you can use Jerry's brain. Getting started points. Big questions. Okay, I don't know what's in Jerry's brain or on big questions,

but that seems interesting. That's very interesting. You know, and all these things would be, I don't even know if this works anymore, but he, you could see like you could jump off on all. It was almost like his bookmarks and stream of consciousness.

Really interesting. Yeah, people out there founders, like, we'll back your startup, make something interesting like this. Yeah, I mean, this idea of being able to generate a map directly out of a conversation seems like a no-brainer. Someone should go build that. Oh, you know what? Look, it just did it.

It must have been his brain is so big that it wasn't working. And I just assumed it was loading. It was loading. Look, I clicked on different types of capitalism. And he has like climate capitalism, clean capitalism, coin flip capitalism, cognitive capitalism. time of democracy and gap. I mean, it's crazy. You know, the cool thing about this is like his grand,

grandchildren could actually like instantiate this in some kind of like AI that can talk to you and actually be able to talk as if it's Jerry. And in a way that creates something that could outlive you that could, you know, create your wisdom encapsulate your wisdom for people to see in the future? It's crazy, yeah.

I mean, it'll be when Neuralink comes out and you can actually put these categories together and it just, your brain kind of maps itself automatically. I mean, scary, but it's definitely the possibilities there. All right, listen, I can talk to you for hours, and I did again. Everybody, if you're interested, go to Replit, if you want to post some bounties and have it done quick, cheap, faster, better, cheaper.

You know, it's a very few things that are faster, better, cheaper. And if you're not using Replits Bounty Program or their ghostwriter, you're a dummy. This is how your startup could go at least 50% faster. So instead of just like hiring a bunch of offshore developers, nothing wrong with that, or, you know, sitting there wondering when you're going to get this latest bit of code or feature done, put it up on the bounty program, see how you do. I think you're going to do great. Or go learn how to code.

or learn out of code, right? Everybody go learn out of code. I mean, I got three kids. I was like, literally I was like, oh my God, if I go down this rabbit hole, this is going to be 20 hours a week. No, no, no, I need to sleep. I need to spend more down with my kids.

I got the guilt of like, oh, this is like some crazy video game that's going to take over my life. And you got three podcasts and all that. Yeah, I get two podcasts. And I'm trying to get the podcast to be less per week, but with even better guests. Like, I was doing the news every day.

And like, I was really getting into it. But then I was like, you know what? The news is transient, and I'm not enjoying it as much. And so I really got focused on... Oh, you stopped that news thing. I do like news once a week on All In. And then I do a little bit of new... I'll do a news program on This Week and Starves,

but instead of doing news every day, I was like, let's just get, you know, three times a week, you know, legendary or emerging founders. And then one day a week do the news, right? And then it's maybe a little bit of a better balance. But I also have to enjoy doing the pod and, like, talking to you and then just, jumping off points and talking about the future and then meeting new entrepreneurs. That's really the heart and soul of this weekend startups. And I kind of reverted it back to also for my energy.

You know, like I have to want to come do this. And we're at 1800 episodes. Like, I don't know if I'm going to make it to 2000. If I'm not super compelled by the conversation, I need to be compelled by the conversation. So you got to. That's an interesting. I don't know if it's about getting older, but I started reasoning a lot more. from this point of view of like, do I want to do this for 10 years?

Right? And a lot of decisions that DRAP will become a lot easier because I'm like, oh, I want to build this company forever for as long as I can do that. If I want to do that, I want to make decisions that optimize for the long term. And there are things that you will get pleasure from and then things that you will consider chores and then things that you will consider painful. And when it's something that delight you. you run to it. I couldn't wait to get on the call with you. When it's things that you don't like,

you avoid them. And then it's chores, you do some combination of avoiding or doing them poorly. And that's where, like, being a great manager is just hiring people, you know,

if you're a founder, who look at your chores or things you hate as the things they love and they run towards and they get joy from. So, like,

I have people who get joy from, you know, you know, managing complex operational processes and that, gives them fulfillment. God bless them.

Not me, but there were people who love ops and they just thrive in ops or people love sales, you know? And I think you got to also like optimized for you wanting to get to the office every day and,

you know, get on Slack every day and enjoy it. Right. Yeah, totally. Yeah. Life is short. Yeah. All right.

We'll talk to you all next time. Bye bye.