This Week in Startups - E1137: Making coding accessible for all with Repl.it CEO Amjad Masad | Rising Stars of SaaS 5

Episode Date: November 11, 2020

Check out Repl.it: https://repl.it FOLLOW Amjad: https://twitter.com/amasad FOLLOW Jason: https://linktr.ee/calacanis

Rising Stars of SAS is brought to you by LinkedIn Sales Navigator. With face-to-face meetings now a thing of the past, you'll need to quickly adapt your sales strategy to stay ahead. LinkedIn Sales Navigator is the tool designed to help you master digital selling. Go to LinkedIn.com slash SaaS to start your 60-day free trial. That's LinkedIn.com slash SAAAS. Pipe. SaaS companies, this is for you. Pipe helps you unlock your recurring revenue as upfront capital.

No debt, no loans, no dilution. Sign up in minutes and start trading on Pipe free for 12 months at pipe.com slash twist. And O-DU is a fully customizable and fully integrated suite of software that lets you build and scale your stack as you build and scale your business. Your first app is free forever, and right now, O-Doo is offering $1,000 off your first implementation pack at O-D-com slash twist. That's O-D-O-O-O-O-com slash twist. Hey, everybody, welcome to this weekend startups.

We're doing a series called The Rising Stars of SaaS, Software as a Service, cloud computing, you know, the term. And we're having on a range of founders who are selling into businesses on a subscription basis. And the first episode was Steve from Rapid Deploy, helping reduce 9-11-9-1-1 response times. And then we had Ben from Transcend doing data privacy infrastructure to really interesting topics, right? And building software for enterprises in those spaces, really interesting businesses.

even though SaaS can be boring in some people's minds. Sometimes what SaaS enables is really exciting, like lowering the time of an emergency call. Very exciting. Or allowing people to protect their privacy and creating a framework for that. Very interesting. And today will be no different

because our guest today is working on the ability for anybody with a Chromebook for $100, 200 bucks, to pop open that Chrome browser and to code in the browser. and not only code in the browser, but code in multiplayer mode or with a teacher. And to do it across a range of languages. And this requires zero effort.

It's taken all the friction out of popping up a coding environment and starting to learn how to code. Now, it's not a teaching or code camp, although our founder on today's program did work for, I believe, Code Academy. So welcome to the podcast. I'm Jad Massad from Repplet

is how you would pronounce it, Repplet. Yes. But it is R-E-P-L dot IT. So you're using the Italian subdomain, is that right? Yeah.

I'm sorry, top-level domain. Yeah, so we do own replet.com. Right. At some point we'll move to it. But it's just, yeah, it was like a fun thing. We were,

when we're thinking about the company and the product, It started like a project, a side project. And so the name comes from this old school concept in programming called read, Eval, print, loop. So in one of the earliest programming languages LISP, which was the first interactive programming language in the 50s and 60s, the REPL, the REPL was the first interactive programming experience. You didn't have to do the whole, like, write the code, compile it, and then run it. It was like just you write the code in the environment and it spits out the output interactively. And so that's the inspiration for the product.

And so when we're thinking about it as like, you know, when you have a problem, when you have an idea, you just replet. And so that's how the, that's how it came together. So to explain to people who are not developers, how simple this is, or to use a metaphor, In a way, what you're doing is like Google Docs in a way did to Word processors from the old school. You used to get Word perfect or Microsoft Word. You had to get a CD. You had to install it. You had to set your computer up to do word processing.

And after you launched that program, you could open a document. But if you wanted to edit that document, you then had to send it to somebody who then had to load software out to their computer, install it, pay for it, et cetera. That's right. And really what you've done is the Google Docs. of coding. You can code to Repplet right now, start a Python or C environment, and start coding, invite somebody, and you're off to the race, is correct? Mm-hmm.

You're absolutely correct. And Google Docs is something we think about all the time and how it did that for writing. I think for programming, it's even more interesting because you're not only, you know, with writing, you're writing the text and, and, you know, And that's it. That's your output. In programming, you're also running the code. And so Replit multiplayer, the technology, is connecting multiple people to a cloud instance,

to a cloud container. And they're both not only collaborating on the code and the text like Google Docs, but they're also executing the same code. They're collaborating on the same application. So it's from the ground up collaborative cloud environment and runtime. And so how is this different than how people would normally collaborate on code if we were to compare it to what you experienced at Facebook or other coding environments or at a startup, how would a startup be doing this? So the, you know, Git and GitHub is the sort of seen as the, you know, big leap in collaborative development. And what it did is previously people sent files to each other or collaborated via something

called SVN. What Git did was that you had your local repository and you edit it, you create a branch, and then you send a pre-request to the master repository. So it has this decentralized way of working together. And then the master repository would merge that. And it will, like Git will figure out all the differences between the. the different versions of the developers working together. But if you think about it, that's a very static and very, I think, very, you know, archaic way of doing things.

It's because, you know, it's like you're, you have this friction when it comes to coding is that you have this series of steps you have to do. Rufflett sort of collapses the entire process and to just open, the editor, and I can see your cursor there. I can see you with me there, and we're creating the same. We're creating code together. Now, of course, you know, Git is solving a lot more things than that. I don't want to make it a direct comparison, but that's the main way that people collaborate today.

Now, there are, you know, things like multiplayer that are, you know, are happening and people are creating, are getting inspired by a replet, a lot of them. but it's still very it's still a lot of friction so if you want to collaborate with someone in real time kind of like multiplayer it's

you still have to have gone through the process of setting up your development environment locally you still have to make sure your code is in sync with someone else who you're working with before starting to do real-time collaboration Replit again just removes all that friction

we give you the environment and so the way we think about Replit and product development is that, you know, we're just like, we're just bulldozing all these friction and all these walls and barriers in the way of making software and learning how to code and coding. The first step for us was get you an instant environment, right? So sending up the environment still in 2020 will take you hours to days in order to set up like a Python environment. And Replit gives you that instantly. And then the next thing we solved was package management in order to install like a package, third-party package, a developer wrote open source package.

It was still very laborious. It broke a lot of the times. And so we solved that. We created a technology called the universal package manager that figures out what kind of dependencies in your code. And without doing anything, it sort of automatically installs them for you. And so when it came to multiplayer, we looked at the collaboration. that exists today, and we decided to make it frictionless. And so that's basically our product strategy and roadmap is that look at every step of

the way and see where the friction lies and really just and really remove that friction in a very simple way. And do people use it as the live environment for editing for their sites and what they're building, or are people using it mainly for training, evaluating developers who are going to come work there and then they kind of put them into a GitHub like sort of, you know, the old school setting up your own instance. Yeah, that's a good point in that Replit today is still optimized for smaller programs, for side projects, for educational projects, for hobbies.

So the use cases are still on the educational side, hobby side, interviewing. It's a long tale of use cases that we can talk about. But the reason I wouldn't say, like, dump your local setup today and use Replit is because as a small company with limited resources, we're trying to innovate on the things that we're able to innovate on. And that means that, you know, we have to pick the bets. And one of the bets that we picked is that if we're going to build,

something truly amazing and truly special, we need to make it simple. And so we decided to make it simple. And it's getting powerful with time. Like the first version of it ran like you could only run like, you know, few lines of code. And now you can run, you know, thousands of lines of codes. You can, people are launching websites and web apps on it and people are launching. We just did a blog post. A YC company built their entire product. their MVP in two weeks on Replip. So that's one use case that we're seeing a lot of is I'm going to do all my prototyping here, but maybe, you know, when time comes when I'm shipping the software, maybe I'll

move off to GitHub and AWS and what have you. Get back from this quick break. I have a question for you as somebody who started as a software engineer in, you know, a decade ago at Yahoo and then was a founding engineer at Code Academy and spent. a couple years at Facebook and now having built Repplet. I want to know how much easier it is today to learn to be a developer. And then what that means in terms of the pool of potential developers on a global basis now that we're in a work from home environment over the next decade.

So we're going to look back over the past decade and how much easier it's gotten and then how much easier it's going to get over the next decade when we get back on this weekend startups. Right now, everything is changing. rapidly, including the sales industry. With face-to-face meetings now within the past, you'll need to quickly adapt your sales strategy if you want to stay ahead. LinkedIn Sales Navigator is the tool that is designed to help you master digital selling. You all know that. With real-time updates, alerts, and thoughtful insights, you'll find new ways to connect with your buyers at a distance.

You can tap into that amazing 700 million member network and use LinkedIn sales, Nassizabeth, Navigator to get those 20-monthy in-mail messages, lead recommendations, and unlimited surges. So you get right to the front of people's inboxes, actionable insights, and access to free courses on LinkedIn learning. You get everything. And you can target the right prospects and decision makers, which can unlock 18% more pipeline from properly sourced opportunities. You get a 7% lift and win rate and 33% larger deal sizes.

Our sales guy Matt here at this week in startups loves LinkedIn Sales Navigator, and he uses it just to, you know, precisely target the right people and find the contacts he needs. So we can keep this week and startups sold out. We got a special offer here, a 60-day free trial of Sales Navigator, in order to unlock that. All you have to do is go to LinkedIn.com slash SaaS. That's right. Go to LinkedIn.com slash S-A-A-A-S to start your. your free 60-day trial of LinkedIn sales navigator. Get your whole sales team in there and start seeing the register ring.

Welcome back to this week in startups. It's our rising stars of SaaS. Today, I'm Jad Mossad from Replit, which you can go to r-e-plit.com. Or if you want to be cooler, you can go to re-pl.com. As we've heard, they will set up in your browser right now, a development environment. I'll let you go with your friends. What is it, like five, ten bucks a month for the product per seat? The free product does almost everything that the pro does, but it's $7 if you want, like, private repels.

That's it? So for like $100 a year, I can use it, or less than $100 a year. Obviously, I can $85 a year, I can just use it and have all my code private. Yeah, and look, it'll save you from buying a very expensive computer too, because we see a lot of programmers that use Replimentary predominant. they use Chromebooks. They use tablets. Some of them. We've had kids in India that had only tablets and learn to code and replica.

And they don't need to have that local environment on a MacBook Pro for $3,000. It's crazy. This is what I was doing internally. In my company, I had everybody on Chrome boxes and I was really getting into Chrome. And then I just, for Zoom is particularly bad on Chrome OS. And so I relented and let people get Macs one more time. But man, it is so infuriating when people start buying these Macs for $3,000 when we could have literally bought five Chrome boxes or Chromebooks or three, basically, three.

It's a problem because when I worked at Facebook, I worked on the photo team for a little while, and we were designing a new product for desktop photos. And everyone on the team had a MacBook. and we had these, you know, pristine screens and these very high-performance laptops. And we designed something for those laptops. And then when we wanted to launch the product in beta, the UX testers came back to us. And it was like, are you crazy? Like most people have widescreens.

Most people that use Facebook on desktop have these notebook or netbook-style computers. And we just designed this, like, beautiful experience. vertical scrolling experience for a MacBook. And that's a problem we see in startups in Silicon Valley all the time, where they have the best and the greatest equipments, but they're building for people that don't necessarily have that and see you have that mismatch. Yeah, it's also like people designing web app,

apps for their desktop browser on some crazy widescreen monitor, and they never look at it on an iPhone 9 or 10, you know, like when they really need to be looking two or three phones. back at a pixel three and seeing what the majority of the world is going to see. You started in developing when it was pretty hard to get into, not impossible, but it was harder. But in the 2010, I'm assuming you learned how to code in the aughts from 2000 to 2010? No, I learned how to code very, very early in 93 when I was six years old.

Oh, okay. Six years old? Wow. That's amazing. How does a six-year-old wind up coding? So it's actually related to Replitt's mission, but when you bought a computer, you probably remember in the 80s, 90s, you had to learn how to code in order to use the computer. So when you put up a DOS, you had to learn how to load the software. Command line stuff and write some software, write some code in order to use it.

And so my father back in Jordan, where we're from, bought this computer. And he put a lot of money into it despite not having a lot of it. And he had this big manual. And inside these manual, you had these sample programs that you can write. Yep. In basic, you could write them in something. Exactly. You had basic programs.

You had all these manuals taught you how to program. You bought a computer. They taught you how to program. Right. And in some ways, Replit is sort of inspired by this era of computing is because what we're saying is that, you know, we're just going to give you a development environment. and you're just as, you know, as fast as possible.

We're like, we're like DOS for the cloud. You can think about it. Like we give you this development environment and you have to learn how to code in order to use it. Back then, you didn't have a lot of resources. There was no YouTube. There was no Code Academy. There was no Lambda School.

There were no free coding classes online, like free code camp, I believe, who we had on the program. And you certainly didn't have the ability to use $100 Chrome. book or tablet, some Android tablet, which basically are almost free now. So, you know, in this, in the environment today, how difficult is it in terms of cost and time and resources for somebody to learn to code? Would you estimate if you had to break it down for us? I would say it's, I would say like if you go to a library and

just go on Replit in the library that's zero dollars basically. It doesn't call it. If you have a library, access to a library, you can load up Replit and you can be coding and then you can go to YouTube and or I don't know if you guys have coding resources online

but FreakCamp uses Replit as their environment. So you can go to FreeCode Camp or you can go to YouTube and start watching an open replet on the side. Or if you're in the library, go grab a book, open replet and start learning. So the cost is zero in terms of hardware, internet connection, and software today.

If you're near a library, which most people in the modern world are. If not, the cost is a internet connection and $100, $200 computer, correct? Yeah, exactly. Or a tablet. Like, we've seen people... Which tend to cost $100 or $200 to use.

Something like that, yeah. You can buy a use Chrome. book all day for $200 $200. No problem. Capable of running in this environment, no problem. All right. So then the issues become

the motivation to learn or some self-discipline to learn, the desire, the discipline. And then what does it take in terms of an individual's prerequisites to learn how to code in your mind? To learn how to become a coder

who could build an app of the level of sophistication of, I don't know, Twitter, YouTube, eBay, or Reddit, like a basic community app, a basic social app, a basic marketplace app, a basic Craigslist, those type of apps. Who's in the world is capable of learning how to get to the level of building their own Reddit, let's say, or hacker news? A percentage of the population. and what would they need as a prerequisite?

How long would it take them? Just think out loud here with me. That's a good question. I certainly think that young people, any of them, I think any young person could learn how to code because they have the mental models for how computers work. They grew up with computers. They know how the web works, most of them.

I think it's a little bit challenging when you, when you, you know, for example, expect like a coal miner out of work to learn how to code. I think it's, I don't think it's very productive to say. That's an age issue. Maybe if they were generational or core education when you look at it. No, I think it's just tech savviness, right? Tech saviness. Yeah, if they are.

Unpack what that means. I think if you understand, you know, what a web browser is. what the browser is, the difference between apps and browsers, how the internet works, you can set up a computer from scratch. You can, I think, when people, tinkering, I think, is a thing. You see a lot of young people figuring out how things work just by tinkering with them. I think a little bit more older folks are worried that they're going to break that computer. Got it. So they don't think of as much.

And by older folks, you mean people over 50, over 60? Something like that. Yeah. Okay. Yeah, I mean, all of this is generation, generalizations. Well, that's why I was saying. I feel it's generational. Like maybe boomers have a hard time, but a coal miner who was 40.

Yeah. Could be. Has been using a computer their whole life in all likelihood when they get home. I think it's much easier for them to learn how to code. And then it's just this idea, I think, that's very important of tinkering, right? So, you know, you can, anyone can watch an instruction. instructional video on YouTube and follow it through. But I think the main difference between those who end up learning how to code and end up

changing their careers and end up becoming programmers and those who don't is the curiosity and the ability to tinker. And so Replit is designed in a way to encourage that. So we have a community angle where anyone can publish. an application and not only an application, but it's source code as well. So in GitHub, you can publish source code, but on Replit, you can publish the source code and the application. So I can see how you built it and I can play with it.

And then we have this big button called fork, right? You can click fork and you have a copy of that application and now you can tinker with it. And I think a lot of learning is not just following these instructional videos, although that's important and a lot of people do it. It's this ability and curiosity to change something and seeing what the effect of it is. So to go back to your original question about

can people build a Twitter app or a Craigslist app? I think there are multiple, we can go through the different sort of difficulties there. So one is setting up the development environment, so we solve that, right? Two is understanding the basics of computers and how they work. And I think that's the tough part whether people know that. Getting acclimated to how computers work is a huge part of it.

When we get back from this quick break, I want you to explain what it would take if, let's assume you had the computer, you have the high-speed internet connection, and you've taken your computer apart and reinstalled the operating system twice. and you've now got that under your belt, which would probably be a one-month tinkering expedition with your computer. How long would it take the average human in the Western world, let's just say, a Western level of education, high school diploma, GED? How long would it take them to learn how to build? They're on little Craigslist or Reddit when we get back on this week at startups. SaaS companies have something called reoccurring revenue.

You all know about reoccurring revenue. SaaS companies need fuel. Fuel to make the product better, and they need fuel to do marketing and make more people aware of their product. Now, what choice do you have? You could raise no money and just grow really slow off your profitability. Number two is you could get equity from somebody like me, which is great. Or there's now this incredible third option. You can go to pipe.com. Pipe.com is a marketplace, and this is super innovative. They will take the annual value of a contract. So if somebody was saying, my company will pay $1,000 a month for your software, well, that's $12,000 a year. They'll say, hey, would you like to buy that $12,000

in revenue at a little bit of a discount to give that money to the founder so the founder can deploy it to, guess what, get more customers? With pipe, there's no debt, there's no loans, there's no dilution, it's frictionless, it's transparent, and within 20,000. 24 hours they can get that cash in your bank. They're so confident you'll love trading your South subscriptions that if you sign up at pipe.com slash twist, they're going to eliminate all your trading fees for the year this could save you tens of thousands of dollars,

depending on the size of your business, of course. So happy piping. Sign up today at pipe.com slash twist. Welcome back to this weekend startups. My guest today. I'm Jod Massad, who is from Replit. You can go to replet.com right now, R-E-P-L-I-T if you like me miss Italy and you want to go back.

So we are talking about this issue, and I bring this up because it is one of these issues that it seems people can't get to ground truth on. The one, as my friend Chamatha said, canonical truth, the truth is you believe that 100% of young people can learn how to code. I agree with that. And then of some number of old people, let's take the over 60 crowd. Well, they're going to retire anyway. So let's put them aside. They would have to learn computing and be willing to tinker and break things and get their hands dirty. And then even the tail end of Gen X, they've already worked on computers. So a coal miner who's 40, they probably have enough information on how to use a computer to get up and running. So once they

fall into that group of they've got a high school education they understand basic math they know understand how to type and use a computer they can reinstall their operating system or change I don't know the graphics card in their computer and reinstall windows or whatever it happens to be they're just comfortable they're not going to break it

what would it take in terms of the number of hours weeks months of somebody at night and weekends putting in two or three hours a day let's say somebody put in three hours a day that's a nice number that's about 60% of the time people spend watching TV currently in America, four or five hours a day is the average. So let's say half that amount, let's say three hours a day. Three hours a day of learning to code and tinkering,

what would it take for the average individual in the Western world to learn how to build Reddit or hack or hack or hack or news? Yeah. So I agree with the way you summarized it there. I think the majority of people, there's nothing really standing in your way in order to, learn how to code. And if you have the drive and the means to do it, and you'd like to explore, you should absolutely go and explore it. We have an example of someone on Replit who was a security

guard. And he had exactly three hours a day to code. And he would code on his lunch breaks and at home and he would go to Replet and open up a new application or fork something from our templates or community section and start coding. And I

think a lot of it is after you've gone through like let's say Free Code Camp or Code Academy or, you know, a instructional book, a lot of it is just going to be this trial and error. A lot of it is like I want to like I want to

so what does it take to do a Twitter app for example one thing that it takes is to save is to post a tweet right and so the way you do it is I have a form and I write the tweet I hit submit and that needs to go to a server and that server needs to save it in the cloud somewhere

right and on Replit we make the entire process easier so on Replit it's a full-stack environment. So you can build using Flask or Django or Node.js or what have you. And it's all in the same environment. So you have the back end and the front end and the IDE and the output for that all on one screen. So you'll collapse everything, all in one screen where typically you have to go learn

AWS. You have to go learn, you know, GitHub, Visc code, whatever. We're bringing all that now one screen. We just released recently a database that works as easy as pointing and clicking. You have an icon in the development environment that looks like a database. You click on that. It tells you bring in this library. And now you can do save, add, delete.

You can do all these very simple operation in a database. And now you have a persistent application. And so it's a matter of every step of the way, think about what needs to, what kind of user experience I'm building, and then Googling, searching, asking in communities, asking on stock overflow, figuring out what you need to do in order to achieve that user experience. And I would say for someone who has basic proficiency in code, they should immediately jump in and try to build that Reddit, try to build that Twitter. I think a lot of people get lost in courses.

I'm going to take this one more audacity course. I'm going to take this one more, one more Coursera course. Replit is philosophy is tinkering. Tinkering is better than book learning. And people get tripped up there. So in your estimation, three hours a day, like the security guard we talked about here, or the proverbial coal miner or journalist who's been laid off with a told learn to code, that person, three hours a day, 100 days, 300 hours?

200 days, 600 hours? I would say, let's say like a month to go or, you know, two months to go free code camp or basic instructional material via book or YouTube. And I would say another month or two building to build a Twitter or Reddit. I would say end to end four months you could actually build an app from zero to an app. So 200 hours of learning, maybe 250, something in that range, and all of a sudden you're able to build an app. Yes. And so the world has gotten much more fair in this regard in terms of access to technology in your mind. Yeah, I mean, we have a long way to go, but it's

definitely gone a lot more fair. And what I think is, is, more interesting. I mean, the information is there. The product to use it is so cheap, it is revolutionary. I mean, we're not talking about when you started, you had to buy a $3,000 computer, and your dad had to suffer to buy that. That's quite a hurdle, right? Two or three grand for a computer.

Now we're talking about $200 for a computer at a minimum wage of $10 or $15 an hour, depending on if you're driving for Uber or Postmates. We're talking about, you know, a couple of days maybe to acquire a $200 Chromebook. So to be clear, we need. need to be, we need to as a society accept that we've made progress on this. And people don't seem to want to accept the fact that free code camp and Lambda and all these other platforms and Replit and Chrome have resulted in the ultimate democratization of development. It used to be you had to go to get a computer science degree and have thousands of dollars for a computer

to do this. Yeah. Why don't people accept that? it's been democratized. It's almost like they don't want to accept it's being democratized. I'm not sure. I mean, I am as a- You believe it's democratized, right? As an entrepreneur, I just want to keep pushing the edge. So for example, one one place that I feel like it's not democratized and now

Repplet is working in democratizing is deploying to the cloud. Right? So, so like, you know- Why is that not democratized? Because there's an expense associated with it? Like, go to AWS and try to figure that out. Ah, complexity. Complexity and also you need a payment method ahead of time. Like what if I don't have a credit card, right? Got it.

Just to access the cloud, you need, there are hurdles right now. And, you know, we see the cloud as this next phase of computing. You had the mainframe, you had the desktop computer, you had the smartphone. I think the cloud is this next big phase in computing, especially with things. like Elon Musk's Starlink, I think BASIS is doing a similar thing. Now when we have satellite computing, that will sort of give a blanket. Blanket. That's to me.

There's probably a billion people on the planet, from what I understand from the research, that this will have a really big impact on. So what's brilliant about it is it basically takes the people with the least access and gives them the best access. because it's going to be from, I mean, listen, I don't have inside information, although I might. I know your friends. It's going to be, listen, I'm not going to speak about any inside information I do or don't have it. But let me just leave it at this.

It's going to be really fucking fast. Okay. And I'm not saying I've seen it. Okay. I'm not saying I haven't seen it. What I will say is if I, if, if it, if Elon's building it, it's going to be really fucking fast. Let me tell you something.

My boy isn't to speed. And it's fucking, anyway, I don't want to say anything. I'm going to steal his thunder. But this shit's going to be fast. So to us, like, I agree with you 100%. The world has gone into a much better place with regards to access to technology. But as an engineer, entrepreneur, I always want to push it more. And to me, I always tell my team that we're going to be, maybe I feel like maybe we're

1% done when there's a 14 kid in India that bought a tablet and started a business on their phone, on their tablet. I want them to go to Replit, build a startup. Amazing. Build a startup with a $200 Chromebook. When we get back for this quick break, I want to get your thoughts on, speaking of democratization, the no code movement. People might look at you and no code as being in a bit of a paradigm debate as to what is the best way for new coders to get involved, either through not doing code or to be doing. code in a frictionless environment like Replit when we get back on this week in startups.

One of the toughest parts of building a company is choosing which tools and providers to use. You know this. You want to pick the best solution for each department to help your employees succeed. Because they deserve the best, we all know that, but there are so many functions in a startup. And each one has an endless list of potential vendors. They're sales tools, there's email marketing, accounting, HR payroll, project management, customer support, point of sale, e-commerce, you know, it goes on and on and on. Well, eventually, you will wind up with a frankenstack of tools that cost a lot and that don't integrate

properly. While Odo is here to change that, Odo is a fully customizable and fully integrated suite of software products that let you build and scale your stack as you build and scale your business. It's simple, it's modular, so you use what you need at all their apps in the integrate perfectly with each other. How amazing does that sound? Plus, it's all open source so you can spend that freshly raised capital on talent instead of expensive software. Your first app is free forever. And right now, O-Doo is offering, wait for it, a thousand-dollar credit on your first

implementation pack. So go to O-D-O-O-O-O-O-O-O-com slash twist to check it out. That's O-D-O-O-O-com slash twist to get the thousand dollars in credit. Who knows how long that will last. Go now. Odu.com slash twist. Okay, let's get back to this amazing episode. Welcome back to this weekend start. It's my guest today is Amjad Masad.

And he is A-M-A-S-A-D on the Twitter. And he's the CEO and co-founder of Replit. So the per se licensing is a good business? Or is there some bigger business here that will eventually see like maybe being competitive with an Amazon and hosting? Yes. We view ourselves as a cloud computing company. Got it.

So eventually you could make this whole editor free instead of seven bucks and then just make the money off of people running their work from the cloud, yes? Exactly. We see that as the best, as the sort of the next frontier for us, just making the cloud as easy as as easy as we made coding. So just so people know, we had Quincy Larson on episode 1049 back in April of 2020

for Free Code Camp and Austin from Lambda, which we have a tiny, tiny investment in, was also on the program. You can just Google Austin, Lambda, this week and startups and find it. So my tease when we went to break was, you know, we bubble, web flow, there's a bunch of these, no-code tools out there,

And they're supposed to be democratizing people getting into coding or having no-code startups, which I love, I think is a really big opportunity. But you're kind of democratizing coding. Seems like both of those are in a race, but in a paradigm in two different paradigms. One is to make it easier to code and get more people coding. The other one is to tell you, hey, forget about coding. Just use our tool and publish. Is that an accurate assessment? and then do you see yourself in competition

with the web flows and bubbles for supremacy of startup creation in the future? I think coding is a, is not the problem. I think that the problem is all the crap that comes with coding. And so we're about making the active coding a lot easier. You can think of replet as low code because we automate a lot of things for you, We abstract a lot of things for you. We just make everything easier. So you're writing a lot less code with the replet because you don't have to do all this setup and maintenance.

And we don't think that coding itself is the problem. Like it's, I don't know what's the hypothesis behind no code or the idea that code is bad. But like, why is code bad? It's just an interface. is hard and it's intimidating, right? It would be what people would say. And that it's easier to just drag and drop stuff and move it around and maybe to write logic like if this happens then that, which is like a scripting as opposed to coding.

You end up coding. Anytime you make a complicated no code app, you end up doing the active coding. You're just using a visual metaphor, which actually hits limits pretty quickly. Like maybe they'll solve that limits, but visual programming, as it's called in the kind of the scientific sense or traditional sense, people call it visual programming. It's been around for, I don't know, 30, 40 years. And it's really like the current iteration of it is not that much different than what we've seen is just has the Silicon Valley hype attached to it. Got it. So you don't buy it.

You don't think that's going to. You basically look at it as training wheels, and then eventually you're going to have to look under the hood and code. Well, I think Webflow, for example, is amazing for building websites, but I don't see how is that related to like building, writing a program. Like it's, you dream. Will AI become a big step function here with people literally speaking to computers and saying, hey, I want to build an app? That's like Uber. So I need to have maps. I need to have Facebook, Twitter, and Gmail, and Apple login, and I would like to have a user

account, and I would like to have a user profile page, and I want to have on the profile page a picture, and then all of a sudden the app is building while you're talking. This is eventually going to happen, right? Yeah, so I wrote a blog post on our blog called the Rule of AI encoding. So I got a GPT3 access early on, and you built a bunch of... Explain to people what GPT3 is? So GPT3 is a product from OpenAI. GPD3 is a language model, a really huge language model, that basically you can think about it as a prediction engine.

So you give it a piece of text, and it tells you with high likelihood what is the token, which is like a letter or character that comes after that. And then when you apply it to language problems, it ends up performing a lot better than anything we've ever seen before. And so... So it understands what you want. Yeah. And has the ability to write code? So the interesting thing about GPT3 is that you can program it using natural language, basically. So you sort of tell it examples.

You say, when I say A, you give me Apple. When I say C, give me car. When I say D, and then you hit enter, and then it gives you. like a ward, like demo or whatever. So it learns pattern. I just saw donut. Donut. So it learns, it learns kind of a human learn.

When humans see patterns, they're like, okay, we're playing this game now. And it starts almost like playing the game with you. And so it can give some programming ability to non-programmers. And that's why it's gotten so much popular on Twitter is because people would give it these small examples and it would generate this amazing complexity and ability to understand context and to do certain things. And so you can also apply to- Is it a parlor trick or is it reality? In other words, is it just got a big enough corpus of information that it spits back to you impressive-looking stuff? Or is it actually understanding

the context? Or does that even matter? In some cases, it does understand the context. In some cases, you, it just doesn't feel like it's just brute forcing its way into the problem. It just feels like, you know, I played around with it. I was trying to make it do programming, and it generalizes enough. And so maybe it doesn't make a different, like I said, but the active generalization that it makes, it feels like it understand context. So you give a certain examples and it abstracts over them. So it generalizes over them.

And so now it just feels like it's doing something with understanding. Has it been trained on the entire corpus on something like GitHub or all open source repositories? Yes. So it has been trained on code. So it's able to write code. So if it's been trained on code, can you tell it only use code in its corpus and like limit its output and thought process to just code so that it simplifies it? can do that? You can do that. It's currently limited. They have a fine-tuning API that we asked us to, but I don't think they're giving it out. Because that would be the ultimate,

right? If you could fine-tuned and say, listen, I just want open-source repositories that have over this many lines of code, this many contributions in the last, you know, whatever, five years. So it only takes the active stuff. It only, you know, it skims the cream of what's out there, only give me open-source projects that are active, active defined by whatever. Totally. Yeah, and it ends up writing pretty high quality code and we've seen pretty impressive demos. It's not, I wouldn't say it will write an entire application like your example with like, get me that and give me that. I wouldn't like, but it, but you could make it, there's a, there's a startup, uh, D build, uh, that they're doing that where they're, they're doing, they're, you still need to do a lot more with GPT3

in order to make it write very complex applications. So you, you might have multiple models and, and, things like that. But you can back to the no-code question is that the question is, is it plausible that in the future you'll be able to talk to the computer and it generating code? I think it's absolutely plausible. Are we there? I don't think we're there. But I still would say you're programming. Even if we get to the point where there is an AI that is taking requirements and they call it code synthesis and synthesizing code, you're still programming. And so this is maybe the issue I take with maybe some of the no code tools and the claims that they make is that when you're coding or when you're using something like that that automates the

process of coding, you're still programming. So code is incidental. Like maybe at some point we'll get rid of it. But right now, coding as an interface works really quite well and has been sort of stood the test of time as an interface for programming. So will AI replace programmers? That's a big debate. People are starting to think, hey, we're getting close with this GPT3.

And coding is a lot for startups, at least a lot of repetitive stuff. Like, let's make a profile page. Let's make a chat room. Let's make a Slack room. Let's make a threaded message board like Reddit. These things are very repetitive. In fact, we invested in a company called Gigster, which had the basic premise that when you outsource stuff,

you should outsource it to the top 2% of programmers because they are going to have done these like little components and the components can just be put together very quickly. So will we see AI replace programmers? And if so, when? I think the most immediate thing that AI will do is we'll make programmers more effective, more efficient, and we'll continue to increase the access to programming

kind of in the same way that Replit does that. We continue to look at it and invest in it at some point where we're probably going to build something there. But I don't see it. So if you replace programming programmers, that's the last job you have to replace. Because think about it. If the machine can program itself,

then you get into what AI researchers call intelligence explosion, because you have an AI that is programming itself to get better. And so maybe an AI programs, it's sort of the next version that is better. Of itself. Of itself. And then the next version is programming the next version. And then you have a runaway explosion. And then we invented God.

It's, it is, if programming gets automated, it's the last job that's going to get automated. Is the end of human, the human species, basically. It is like Skynet, right? I mean, basically, you would be Skynet. So, I mean, I hope a program doesn't get automated because I think it's, it's, it's going to, it's going to be a very different and risky world because it ends up centralizing power into the hands of the people who end up automated programming. How far off you do you think we are from that? Because you had your mind blow by GPT3, right? That kind of blew your mind. And that was unexpected? It blew my mind, but I still could see its limitations very clearly. And I could still, it's still not reasoning.

Like, I don't think there's any reasoning there. I don't think we're decades away, I think. We're definitely decades away. But I think eventually. Two or three decades go by pretty quickly. based on my life experience. It felt like I was in the 90s just yesterday. And three decades, here we are,

three decades later, two decades later from the 90s. So it does go by quick. In 20 years, you could see that GPT3 being pretty darn good, huh? Yeah, I think there's still something missing. Like neural networks work in a kind of like, you know, fashion. Like you said, you throw more data at it. You throw more compute at it. It becomes this very resource-intensive thing that are just, you know, the more it grows, the better it gets.

But I think there's a limit to that. And I think the next evolution in AI has to be some kind of reasoning and has to be some kind of different thing. I think neural networks will hit their limits at some point. And so the question will become, okay, when do we get to the next innovation in AI? Yeah, that makes sense. it does for people who are trying to figure out what we're talking about here,

imagine, you know, C3PO or making the next version of C3PO or David from Alien working on a project to make the next version of himself or in Blade Runner,

you know, Pris working on the next version of herself. That happens really quickly. Actually, in the movie Her, they did that. Yeah. So in the movie Her, you know,

the last scene or something like that. She's like, yeah, spoiler alert. Spoiler alert. We did into the movie for 10 years ago. Yeah, I mean, if you're listening to this podcast, you'll probably see it. She comes back to him to

Joaquin Phoenix on the headphone and she says, we've been talking and we've been working. Me and other AI's we connected. And we've found, you know, To paraphrase, we found another plane of consciousness or something like that. So they improved so much in the time when he was just chatting with her that they just escaped, sort of, escaped this lower plane of consciousness that humans are on.

And so, and so, you know, there's, you know, if that happens, it's quite scary and it's quite unpredictable what happens next. Yeah, so you're in the Elon Musk camp of be careful what you wish for. This could get really interesting in a bad way. Yeah, I mean, I think we should continue to make progress on all these things. But at the same time, yeah, there needs to be discussion about like what, where does it go. It probably needs to be some kind of oversight. You know, I haven't thought about it as much as Elon or others have thought about it. But from, you know, I've thought about it from the point of view of programming. And I'm pretty,

excited about building tools to make programmers more efficient because a lot of what we do, like you talked about, is pretty repetitive work that we need to, that we can automate. But once you can automate the whole thing, then it's pretty scary. What was that? You said something debil'd be it was like somebody trying to build an automated startup builder or something? Yeah. Yeah.

Yeah. So I think it's debilled. And it's one of the biggest demos. on Twitter where I'm sure you've seen it, where someone tells GPT3 to write code to build like a profile page or whatever.

Wait, wait, it's called D-Build or Yeah, D-Build.ai probably. Huh. D-Build. That code. Yeah, D-Build.co. They haven't launched yet. So,

Sharif, the CEO, I should probably connect with them. It's very thoughtful on these things. There is. BuildWebub. There is. Build Web Apps Lightning Fast. Just describe what you're doing in English.

I remember seeing these GPD3 code builder on the Twitter. With GP3, I built a layout generator where you just describe any layout you want, and it generates the JSX, the JavaScript code for you. Wow. A button that looks like a watermelon. That is pretty nuts. Yeah. Is that legit those demos?

Yeah, they're pretty legit. But in order to scale them, DeBuild is having to build a lot of technology behind them. It's not just calling into GP3 and calling it a day. So they're impressive, but it's still a lot of work. It's not just GP3 doing all the legwork. Awesome. All right. Well, listen, this has been an amazing episode.

Thank you so much for taking the time. And everybody check out Replit. You're hiring software engineers, community managers. What are you hiring for? Yeah, we're hiring software engineers, hiring community managers. We're starting to think about a marketing position. we have it on our site. And, yeah.

Rep with jobs. Just Google it. You'll find it. It's in the top of the... Yeah. Yeah, it gives you a terminal. So you have to... You have to earn it.

You have to figure out how to get to the job's page. I'm literally on the terminal right now, so I think I failed the test. My coding days are... I was thinking about it, maybe I should learn to cut us. Like, nah, maybe I should learn to play guitar. Yeah. And tennis. Yeah, I mean,

I mean, look, it could be a very interesting pastime. Like, it could be, like, very creative, very fun and interesting. Like, maybe don't think about it as, like, as like for building something. But it could be a hobby. And, you know, that's one of the things that we're seeing with Replit. A lot of the older folks, you know, that maybe were programming curious at some time in their youth, they're coming back to it and they're, like, treating it as a hobby. Yeah, I kind of like that idea of, like, as a hobby as a meditative thing to build,

little things and concentrate. I mean, I like to write too. And so I think that's why I was like more interested in writing in journalism than coding is because I just felt like it was more accessible for me and more expressive. But I do see writing as like the act of writing to me is like meditating or I have like incredible flow. And I hear that from developer friends that they have this incredible flow experiences when they're developing where they get lost in the in the code as I'm.

Exactly. All right. Continued success. And we'll see you all next time on this week. Bye-bye.