Main Topic: The podcast episode discusses the impact of AI on software development and the future of coding. The guest, Amjad Massad, founder of Replit, shares his insights on how AI is changing the way people code and create software.
Key Points:
Participants: Amjad Massad, founder of Replit, and the host of the podcast.
Takeaways:
Recommendations:
A small minority – likely less than 1% – of the world can code. Yet also widely known that the skillset tends to yield outsized returns, with developers generating some of the highest paying salaries out there.
But the field is quickly shifting, especially with the advent of wide-scale AI. In this podcast, we get to chat with Amjad Masad, founder of Replit, about these foundational shifts.
We cover how Replit has integrated AI into its platform and the implications on both current and future developers. It’s easier than ever to learn to code, but is it still worthwhile? Listen in to find out.
Timestamps:
Resources:
Replit: https://replit.com/
Replit Ghostwriter: https://replit.com/site/ghostwriter
Replit Bounties: https://replit.com/bounties
Find Amjad on Twitter: https://twitter.com/amasad
Stay Updated:
Find us on Twitter: https://twitter.com/a16z
Find us on LinkedIn: https://www.linkedin.com/company/a16z
Subscribe on your favorite podcast app: https://a16z.simplecast.com/
Follow our host: https://twitter.com/stephsmithio
Please note that the content here is for informational purposes only; should NOT be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security; and is not directed at any investors or potential investors in any a16z fund. For more details please see a16z.com/disclosures.
Stay Updated:
Find a16z on X
Find a16z on LinkedIn
Listen to the a16z Podcast on Spotify
Listen to the a16z Podcast on Apple Podcasts
Follow our host: https://twitter.com/eriktorenberg
Please note that the content here is for informational purposes only; should NOT be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security; and is not directed at any investors or potential investors in any a16z fund. a16z and its affiliates may maintain investments in the companies discussed. For more details please see a16z.com/disclosures.
Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.
Host
We're seeing people making software in 30-minute increments. Someone the other day put a request for a landing page up, they attached a Figma file, they got it back in 30 minutes. That's really never been done before. And why shouldn't it have existed like that before?
It's hard to find concrete numbers, but some estimates predict that less than 1% of the world knows how to code. Yet, it's also widely known that the minority of people who know how to not only consume, but create software, yield outsized returns, with some developers generating some of the highest paying salaries out there.
But even this field is shifting quickly, especially with the advent of wide-scale AI. And in this interview, we get to chat with Amjad Masad, founder of Replit, an integrated development environment that allows you to code live in the browser.
Here we chat about how Replit has tackled the difficult problem of making coding fun, but also how it's now integrating AI into its platform via Ghostwriter and the implications of these shifts on both current and future developers, in addition to the applications that can be built.
As a personal anecdote, I actually taught myself to code in 2018, and to this date, it's one of the best decisions I've ever made. And despite screaming this from the rooftops, still many other people find it extremely daunting. My journey took around 300 hours. Yes, I tracked it. But with the advancements in tooling and technology, it may actually be easier than ever to learn to code. And I was surprised to hear from Amjad just how quickly he thinks people can get up to speed today.
Let's get started. As a reminder, the content here is for informational purposes only, should not be taken as legal, business, tax, or investment advice, or be used to evaluate any investment or security, and is not directed at any investors or potential investors in any A16Z fund. For more details, please see A16Z.com slash disclosures.
Amjad, thank you so much for joining the A16Z podcast.
Amjad Masad
It's my pleasure. I've been listening to this podcast for a really long time, so it's great to be finally on.
Host
Well, we're happy to have you. So just to set the tone for listeners, you are the founder and CEO of Replit. I think you're also the head of engineering now, but we'll get to that. Can you just share what Replit is? And I'm probably getting a little ahead of myself, but how it differs from maybe other similar products or development environments.
Amjad Masad
I like to think about Replit as the technology that reduces the distance between an idea and a product. The moment any person in the world gets an idea for a piece of software, the distance between having that idea and making something in the world is typically very large. The internet has really reduced things down further. We think Replit is the ultimate answer there. We want to get to a place where the moment you get an idea, you just Replit it. That's where the name came from.
We do that in a number of ways. One way is that we simplify the development process. Crucially, we don't make it stunted. A lot of tools think that to simplify is to reduce the power. We actually keep the power while also making the process a lot more enjoyable and learnable. We also superpower it with AI. If you don't know how to code, we have various courses that can teach you. If you don't think coding is for you, then you can hire someone from our community to make the program for you.
We have this thing called Replit Bounties where you can put a price on a project you want to get done, describe the project, and then a combination of an AI and a human being will get that done for you. It's ultimately about getting from idea to product and helping everyone in the world have access to that superpower we call software.
Host
I love the way you put that because you've represented the different modalities depending on where someone is and their proficiency in terms of being able to code. Regardless of where they sit, getting them to that end product, that end vision that they want. We'll get to Ghostwriter, we'll get to Bounties. But I want to give people a real sense of some of the stories behind Replit because that's something that's really gripped me as someone who's been following the company.
I'll call up two, but then I want to play a quick game so that we can share more of these stories with the audience. Two things that I remember. One is recently shared a 13-year-old kid that set up a sizable crypto mining operation via Replit. Another example that I saw a while back was these two kids, they looked maybe six years old, crashing a computer science teacher conference and sharing their startup pitch.
So I've come up with five or so different scenarios. I have no idea if these scenarios have actually happened on Replit. I haven't shared them with you prior to this, but I think they might stimulate either other stories that have happened on Replit or you can tell me if these things have indeed happened. So let's give it a shot.
The first one is someone has started learning to code via Replit. They haven't gotten a computer science degree. They haven't done a bootcamp. They've learned within Replit and since been hired as a developer at a FAANG company. Has this happened yet?
Amjad Masad
I would say it happened. Yes. We've had someone who learned to code on Replit primarily on their own, became a very productive community member and applied to work at Replit. We thought they were very talented, but at the time, there wasn't a really good fit for them. I helped them apply for other companies and they ended up at Google.
Host
That's awesome. Let's move on to the next one. Someone has used Replit to build over 50 projects. You can use your own definition of project here. These don't need to be full-blown startups, but have you seen someone build that magnitude of projects on Replit?
Amjad Masad
Yeah, I think that's a fairly reasonable number of projects. There's a 16-year-old developer. His name is Rayhan. He's one of our most prolific programmers. He actually built a bunch of interesting projects where he reverse engineered how Replit works. He built an unofficial API. One of his unofficial APIs is a security program that searches people's repls to find Discord tokens.
A lot of people build Discord bots on Replit and they copy and paste the tokens in clear text as opposed to putting them in our encrypted service, the secrets manager. He would find those tokens, he would invalidate them because Discord has a service to invalidate those. Then he would send them a notification saying, "Hey, we found that you've exposed your token."
He is one of the most prolific bounty hunters. He built one of our earliest bounties, which was a startup that wanted to build a stable diffusion-based t-shirt generator. He would generate a t-shirt based on a prompt and then get it printed and sent to you. There isn't one week where I don't see Rayhan producing a new piece of software. I would say 50 is probably on the low side of things.
Host
Yeah, I undershot it, I guess. Let's rapid fire through the last three. The third one is someone has built an app on Replit and through that app has since made a million dollars or more from it.
Amjad Masad
A million dollars? I don't think that happened yet. There are people that are on trajectory to doing that right now. There are bounty hunters that are making thousands and thousands of dollars a week. I think that'll continue to grow. There are startups that are starting entirely on Replit today. There are a couple of AI startups that have their entire stack built on Replit. Presumably one of those projects will maybe get to a million dollars.
There are earlier examples of startups that prototyped their projects. For example, Fig is this command line autocomplete tool that this YC startup built. It's used by tens of thousands of engineers. The original product was entirely built on Replit. They're starting to sell. I don't think they're making a million dollars, but that's another example of maybe being on the path to a million dollars. I would say it would be a success if an individual person made a million dollars. I think that's really my dream and I hope we can get to that this or next year.
Host
That's awesome. And I should mention, there's a time element to this. I believe Replit was founded in 2016?
Amjad Masad
That's right.
Host
So we're around six to seven years later. Let's see how things progress. The next one is someone over the age of 90. I went a little crazy with this one. Someone over the age of 90 has used Replit. Has this happened?
Amjad Masad
Someone who ended up working at Replit, their grandpa in India is a physics PhD. They do a lot of physics writing. In one recent paper they published, they used Replit to write a physics simulation and Replit was published as part of that paper. I don't believe they're 90, but they're definitely 70 plus.
Host
Okay, final one. Strangers have met via Replit, maybe it's the forum, maybe working on a project together, and have since gotten married.
Amjad Masad
That would be awesome to know. I haven't heard of it if that happened. I know there are some community members that recently got married. I don't think they met on the community. I've certainly seen some people date and things like that. But married? I haven't heard that yet.
Host
Well, I had to ask it. I hadn't shared these with you beforehand, so I went from what I think are maybe more ordinary scenarios to more extraordinary ones. One story, let me mention, there's a story in the news about the world's youngest Microsoft Azure AI certified programmer. Naturally, someone looked them up and they found that they're on Replit. This kid is six years old.
Amjad Masad
Oh, my goodness.
Host
If you go to the Replit profile, they have a thousand followers on Replit. They have games with thousands and thousands of runs. They're completely prolific. I can't wrap my head around a six-year-old doing this every day. It's pretty amazing.
Amjad Masad
That's incredible. Six years old, that's grade one. In grade one, I was learning to trace letters so that I could write the alphabet. That's incredible. Are there any other stories worth mentioning? Any top of mind scenarios like a six-year-old who's coding on Replit?
One of my favorite stories is one day I wake up and I see my Twitter blowing up. I go on Twitter and I find that this Indian mom and dad are tagging me and saying that I corrupted their child. Their kid is totally addicted to Replit. He's supposed to go to IIT. As you know, IIT requires a lot of preparation, but that kid is not interested in it. He's not studying at all and just programming all the time.
I felt pretty shitty about it because it started going viral and a lot of people were pretty negative about the parents. Apparently it hit a nerve in some discourse in India about how parents push their children really hard on IIT. I reached out and tried to help. I tried to talk to the kid and tell him, "Hey, you should listen to your parents."
A few months after that, India was hit pretty hard with COVID. That kid wrote the application that first responders were using to find equipment, to find oxygen, to find all these things. It went totally viral. It brought down our website. In my mind, I was like, his parents should be really proud now.
Then something even more fantastic happened. Out of that experience and him going viral using his skills, he got a job. Now he gets paid more than his entire family. You go from getting blamed for someone ruining their future to actually doing a huge favor to their future. I think that's the power of software and the internet.
Host
That's beautiful. Something that also stands out to me from that story, which we'll touch on throughout this interview, is this idea of coding being fun. This kid really wanted to code all the time. As someone who taught myself several years ago, prior to learning, I thought coding was really dull, something more prescriptive instead of creative and artistic. After I learned to code, I saw it in a different light. This idea of making coding fun is really important. How have you been able to design Replit to actually enhance people's creativity and make them want to come back?
Amjad Masad
Most things that are fun tend to be devoid of a lot of drudgerous routine work. When you're playing a video game, you're not building the video game every time or setting up the TV or doing some rote IT task. When you are doing a sports hobby, you're in the flow. You're doing the thing you're excited about.
The problem with coding is that a lot of the maintenance around the development environment, the packages, and the integration of all the different components was what engineers were spending most of their time doing. The moments they were coding, they were in absolute bliss. But those moments were actually very little in terms of the pie chart of what it meant to work as a programmer.
The first thing Replit does is remove the need to do all this setup. That in itself made programming a lot more fun. Then there's the collaborative aspect. A lot of what we find fun in life has to do with other people. We're social animals. If I can share my program with you with just a link, that's really fun. That's what makes Figma fun. That's what makes any collaborative tool fun. I can just send you a link and you're in there with me, or you can play it and try it out.
Finally, there are explicit gamified elements of Replit. We have a built-in currency called Cycles. Cycles allow you to earn from bounties. You can spend it on AI or compute, or you can cash it out. It's built with a mindset that feels like Roblox. It feels like one of those in-game currencies.
Host
I like your analogy of the setup or the admin that goes into coding at times being a lot more overhead. If playing soccer required me to put on my shoes and shin pads for an hour before every hour-long game, I'd be frustrated. I just want to get on the field. That ratio is actually important.
Let's take a left turn and go straight to AI, which is the theme of the last six months with many companies. Give listeners a quick overview of what Ghostwriter is and how it works.
Amjad Masad
Ghostwriter is like a pair programmer that is an AI. It helps you in every aspect of software creation, whether that's typing code. We provide great text completion to either complete lines of code or entire functions or entire classes. You'll see that happen as you're typing. You can also generate entire programs. You can do right-click generate and give it a prompt in the same way you could give Stable Diffusion or DALL-E a prompt to generate entire images. You can give it a prompt and generate entire programs that run entire applications.
You can highlight a piece of code and right-click and transform it. Finally, the ability to talk to Ghostwriter is the final piece. It's our number one priority. It'll be like a chat box inside Replit. It'll chime in when things happen that it could help. For example, if you get an error, it'll be like, "I can help you with that." Like a Clippy thing, hopefully not as annoying, but Clippy was way ahead of its time. It'll be like, "You got an error. If you want, I can try to help you with that." Or you can just ask it, "I wrote this code. Can you write tests for it?" It's really remaking the IDE as we know it.
Host
Something I've been thinking about as I've learned more about how AI is being implemented with code is the binary nature of code. When you get a bad error, it's just no, no, no, no, no, no. Then eventually you hit a yes. Sometimes that takes way longer than is comfortable. What I like about this idea of having this sidekick is the nuance. The AI can say, "Oh, have you tried this yet? Did you mean this by your code? This is what your code is trying to do. Is that what you intended?" Getting suggestions and having the learning curve be less like a no, no, no, no, yes and more like a gray that evolves into a yes. A gradient descent, sort of how you train an AI.
Amjad Masad
I 100% agree with that. I think it humanizes the process as opposed to this mechanistic trial and error thing. It allows us to iterate a lot faster on tools. Traditional IDEs are very hard to build. They're super complex. Some of these IDEs are super big. You download IntelliJ, that's a couple of gigabytes of stuff. They're clunky. They take up a ton of RAM. Try starting Xcode. It just consumes your computer.
The cool thing about the AI revolution is that the AI is going to be running in the cloud. You'll give it a prompt about what you're talking about, what your code is about. It'll be able to implement suggestions and features without having this heavy, algorithmic, hard-to-maintain piece of software. I'm really excited about that. As a startup, you always want to catch a new platform shift. With Replit, I feel like we're catching this platform shift. The older IDEs will not be able to adapt as fast as we can.
Host
Something that I've heard you talk about is the decision to build Ghostwriter on top of your own models. Something like Copilot is built on top of GPT-3, to my knowledge. That's a decision to be built off another platform. But you went a different route. Can you speak to how you made that decision and what kind of inputs led to that output?
Amjad Masad
First of all, how crazy is it that Microsoft had another company, whereas Replit built our own thing? There are multiple ways to answer this. One of them is UX. UX is inherently inseparable from the infrastructure for how a product works. Most people think they're separate things. But if you're serious about making products, there's a famous Alan Kay quote to Steve Jobs. He told Steve Jobs, "If you're serious about making software, you have to make hardware." That's why Apple is a full stack company because they think about everything from the transistor to the touch.
For us, if this is going to be a core interaction with our platform, we have to be able to optimize it. We have to get the latency down to the point that it's going to be a really great user experience. We weren't able to really get that when we're hitting something over an API. The latency would be all over the place. We couldn't get the caching right. We couldn't get the location right. We didn't have control about any of these things. That's a huge downside of being a consumer of a mere API.
The other part is the strategic part. If you believe that this is a primary platform shift and this is going to be a core part of your technology, then you have to build it. If you call yourself a technology company, that means you build technology. It doesn't mean you're just building glue code on top of existing technology.
Finally, we think that we have a bit of a data advantage. That data advantage will compound over time and will allow us to train more advanced AIs over time. All these three reasons made sense for us to bring at least part of it in-house. I should say that we still use OpenAI for a lot of the bigger workloads that require really large models.
Host
Something I found really interesting was Daniel Gross and Nat Friedman were on the Stratechery podcast. They talked about how Copilot ultimately got to the interface that it now is. Originally they wanted to create it as a chat bot. They thought people would run into an error and want to talk to someone and ask, "How do I fix this?" They'd get a response and implement it. But ultimately they pivoted to what you might imagine as a robot on your shoulder that only speaks up when it has confidence. It's the early days of Ghostwriter, but thoughts on how you got to the specific interface that is Ghostwriter today and how you linked the UX that people see to what's happening in the back end as you're building these models?
Amjad Masad
I think there's two modalities. One is pull and one is push. Pull is the human knows what they want and they're going to ask for it. You write a prompt, you wait a little bit, and you get it. Push is the robot on your shoulder that is continuously suggesting improvements. There are tradeoffs to both. The push model is actually fairly expensive because you're computing all the time. It needs to be fairly low latency, so that commits you to a smaller model. You can't use a super large model.
On the pull model, I'm asking something. I'm going to formulate my question in a way that the AI could understand it better. I'm going to be able to wait to give the AI time to think or to compute. It's not either or. You have to do both. I think that's the fundamental UX innovation we brought to this space. We call it society of models. Copilot uses a single model. Replit uses three different models of different sizes.
The smallest model is the model over your shoulder continuously giving you suggestions. Then a medium-sized model to do the transformations and things like that. Then a super large model, the kind of model that you would want to talk to, ChatGPT-style model.
Host
That's great. That shows a concrete example of the capabilities you get from developing your own models and being able to fine-tune them to specific use cases.
Let's zoom out a little bit and talk about where we are right now. We're at the beginning of this phase where AI is being applied to code. Copilot came out about a year ago. Ghostwriter came out recently. We're in the first inning. If we extrapolate to Ghostwriter v4 or Copilot v6 many years from now, I want to think about how you see the overall environment for developers emerging or evolving.
On one hand, I could see how people might argue that having this technology, being able to just tell a computer, "I want to create this app, go build it for me," is going to create a wave of really low quality developers who don't really know what they're doing. They're just relying on this AI crutch.
I could also see an argument though, if we're talking about suggestions within their development environment, this actually creating better developers. They understand a little bit more about how their code is being executed. Maybe they more quickly pick up new skills, new languages, new frameworks because they have this assistant. How do you see this all evolving and the ecosystem of engineers changing?
Amjad Masad
Anytime you make something more accessible, you get the entire gamut of things. Instagram made photography way more accessible and you get a long tail of crappy photographs. But you also discover people who would have never been photographers. Everything in the world is like that. YouTube, most of their creators don't get any views. But a few creators like PewDiePie, pre-YouTube would have been just a kid in Sweden, not a super celebrity. Making things accessible gets you a lot more people involved. The trade-off is that you're going to get a lot of noise, but you're going to discover talent that wouldn't otherwise be discovered. I think that's good for the world.
With software engineering becoming more accessible because of AI, we're going to get a lot more developers. Probably 10x more. Right now there's about 30 million developers in the world. There's probably 300 million developers by the end of the decade.
The way I see the developer market evolving is that there's going to be a bit of a bimodal distribution. Bimodal distribution means there's no middle end. There's a large tail on both sides. The middle end is the sort of glue code plumber type developer that we have today. I think that'll go away. The reason is because platforms are going to be a lot more expressive. They're going to be able to be programmed using natural language. Cloud platforms are building better abstractions. Things like Replit will make backends a lot more accessible.
The front-end engineer is just going to get way more powerful. A front-end engineer will be able to build full-stack products because they have access to all those really powerful platforms. They're going to be able to produce a lot more, use AI in every part of the coding process, whether it's testing, CI, CD, design. Everything is going to be powered by AI and made a lot better, including quality control.
On the backend, low level platform engineering side, those people are just going to get a lot more powerful. Imagine John Carmack. John Carmack is what we call a 10x developer today. Imagine giving him an army of AI developers that he could delegate work to, that he could ask questions of. You're going to make him 100x, 1000x more productive. You're going to have maybe fewer of those low level 10x engineers, but they're going to be 1000x engineers.
A single company might need two or three elite engineers and then dozens of front-end engineers building all these products and maintaining what the customers see. But that core group of elite engineers, their impact is just going to be tremendous. They're going to be demanding a lot more money and they're going to be making a lot more money. If engineering is really your craft, it's not going away and you're going to be able to accentuate your power.
Host
I think it's funny that you mentioned the 10x engineer because a lot of people make fun of this concept. You don't see a 10x plumber because you're limited in the leverage that you can get with your time in a scenario like that. But we know that one of the reasons there truly can be 10x engineers is because software can give you leverage. If you enhance the power that software can provide, it's not crazy to imagine a 100x engineer or 1000x engineer. If you have these robot developers that you can rein in and apply in a specific direction.
Amjad Masad
It's scale. It's a concept of scale. This is what technology has allowed us to do since the dawn of humanity. You go from collecting crops manually by hand to using animals to using robots. Now a single farmer can maintain entire acres of farms just because of all the technology they're using.
The 10x denialism is kind of funny to watch. Every one of those people that say 10x engineers don't exist know in their heart of hearts that 10x engineers exist. They probably worked with someone they highly respected and admire. The reason they say that is just politically motivated. They just don't like the fact that some people can be better than other people. That's just a fact in life. It's uncomfortable to accept that some people actually do bring more value to different organizations.
Host
With the bimodal description that you've shared, I have to ask the question: Is it still worth learning to code? To me, it feels like it's more worth it today because it's a tide that raises all boats.
Like we said, the 10X becomes a thousand X, but a 0.1X becomes a 10X, right? So if you're someone whose impact with coding was previously going to be very minimal, now it's going to be meaningful because there's this rising tide. Previously, you learn a bit of coding, you learn how to plug together some frameworks and create some UI. Going from that to doing a little bit of parsing, for example, parsing some text, is very hard. Right now, you can just use GPT-3 to do the parsing. So you're doing the basic coding, and anytime you find something that's a little difficult, you can plug in GPT-3 in that place. I actually think that programming becomes more fun and impactful because of that. So it's totally worth doing, even more worth doing than before, because you're going to get more done. You're not going to get as stuck.
I think too, how quickly people can probably learn. I wonder if you have any data on this. When I learned, I decided to track how many hours I was spending from when I first started to when I actually felt proficient. For me, that number ended up being 300 hours. Maybe I was slow, maybe I was fast. But it kind of shocked me because if you distill 300 hours down to learning full time, that's less than two months. And now with the tools we have today, we can probably do it way faster than 300 hours. So any thoughts on how quickly someone can actually get to proficiency with the tools that we have?
Amjad Masad
So I know you're a fan of Balaji, and he was on your podcast. He has this beef with media, and he wanted to show that the New York Times is a bunch of bots. He put in a bounty on Replit. He was one of our early adopters of bounties. He wrote, "I want someone to build the GPT Times instead of the New York Times, but totally based on tweets. You give it a tweet and it goes and generates an article written in the style of the New York Times." It should be up now. If you go to gpttimes.com, that's the site that was built using the bounties. The person who built it was on day 80 of 100 Days of Python. 100 Days of Python is one of our programs to learn how to code. We say you can do a day of Python in basically 20 minutes. If he was on day 80, spending 20 minutes, how many hours is that? That's 26 hours. So in 26 hours, they were able to build an entire website using AI and earn $1,800 from Balaji. That's your answer. Maybe that guy was an outlier, but we're seeing it all the time where it's taking weeks for someone to get to a place where they can build things. I think that's really what matters.
Host
I think that's incredible because I think one of the best decisions I made was learning to code. To me, it really was a foundational shift in understanding the world around me. So much of what we do is digital, it is online. Even if you don't want to go and be a developer that's hired to build products, just having that understanding. It's kind of crazy when you think about it. If you were to position that to people and say, "Now it only takes 30 hours. You could do it in a week. You could take four days off from your job, go through this." I know there's a difference between 20 minutes a day and a straight shot, but that's pretty incredible.
The final thing I want to ask you about when it comes to AI is how this might shift, not just how quickly we can code or learn to code, but how this may fundamentally change what we can do with code. Let me give you two examples to shape the question. The first is when we learned that computers were better at chess than humans, we didn't just learn that fact. We also learned that there were all these other moves or modalities to chess that we had never considered. It reframed the way we saw the game. Another example is one that our games team shared recently. In the flight simulator game, you can fly and land an airplane in a digital world. It has a 3D model of the Earth that was built off of the 2D model from Google Earth. That 3D model was built with AI, and it only could have been built with AI. So those are two examples of how this technology not only made things faster, but actually made things possible that weren't available before. Have you seen any glimpses of that? What does it change to have that superpower accessible?
Amjad Masad
This is such a brilliant question. I haven't thought about it as much, and I think I should. Your example with chess also happened with Go, with the Lee Sedol game. Do you remember that move? The AI made a strange move that gave it an advantage in the near term and a huge advantage in the long term. They were so confused about it because it's almost like alien intelligence intruding on this thousand-year-old game, producing this fundamentally novel move. I don't think we've seen that entirely in programming yet, but I'd be on the lookout for that.
What I would say we've seen is that tasks that previously required a ton of laborious work are getting done like that. For example, the parsing question. GPT-3 is incredibly good at parsing. If you give it a malformed JSON, it will still parse it. Writing parsers is one of the hardest things you can do in programming. Writing parsers with GPT-3 is one of the easiest things you could do. You could spend 15 minutes in the OpenAI playground. So that goes from a task that requires hours and maybe days and weeks of building and testing to something that takes 15 minutes. That's a fundamental phase shift in how we do things.
I think there are ways in which we haven't totally explored how to use LLMs in programming. Can you create backend as a service using LLMs? So basically, Firebase, but entirely using natural language. Firebase allows you to have a backend without any backend knowledge. You just start storing data and retrieving data and it just works. Can you have a backend that's completely programmable using natural language? Can I describe my application and just write the front end for it and have the backend taken care of? If that's possible, that'll cut down on how many engineers you need on your team. That'll cut down on time to create a prototype. I think that will be incredibly exciting.
Your question was more like, what is something fundamentally not possible that became possible? I don't think we've seen that yet, but I do want to think a little harder about it and really open my eyes wide. Maybe one direction that could happen is the action model. Have you looked at action models at all?
Host
No, it'd be great if you could describe what they are.
Amjad Masad
Transformer models are the models underpinning large language models. GPT stands for generative pre-trained transformer. You take a transformer, a type of model architecture, and you throw a huge corpus at it and it learns a ton of things. Then you can program it using prompting. Now you can do a different type of transformer where instead of throwing a ton of text at it, you throw a ton of actions at it. What are actions? For example, actions could be all the mouse and keyboard events in a browser. You take this raw stream of data and train a transformer to do that. Now the transformer encodes knowledge about how to use a browser.
Host
That's wild. What can you do with that?
Amjad Masad
You can now instruct the transformer to go book an Airbnb for you, to do more complex tasks like find the place with the best weather in this time of year and book an Airbnb for me and my family. That wasn't possible before. If action transformers became mainstream and as powerful as GPT, then it'll unlock a new programmable platform. Now it's almost like everything has an API suddenly. In the past, a specific database has an API that you can plug into. But now anything in the browser, anything on the web could potentially be transformed into its own API without setting up the specific API yourself. The language model could actually figure that out.
Host
That's fascinating. To your point, we haven't seen these specific examples of "wasn't possible, now possible" yet because we're really early. But even just what we've discussed so far, the amount of foundational development that is required from engineers today that will soon be abstracted just opens that time up, that brain capacity up to apply to something new. It's a little bit inevitable for something to emerge. We're just not sure what that something is.
I want to ask you very quickly about mobile as well. Up until now, desktop has really captured most of development. I can't think of many developers who code on their phone, but it sounds like this might be changing. In fact, I heard you say that you think millions of people will code on their phone.
Amjad Masad
If you think about what we've been talking about with AI, the idea that your primary development experience will include a big portion of chat. The best way to do chat is on your phone. Everyone texts on their phone. Everyone talks on their phone. Being able to generate software by talking to your phone is going to be a very clear thing that will happen. At minimum, being able to instruct your AIs while you're on the go and review their work, that's obviously going to happen. As typing becomes less of an issue on the phone, making complete pieces of software will make a ton of sense. You're just prompting and reviewing, prompting and reviewing. That iterative loop is very clearly going to work very well on the phone. It could actually be more delightful on the phone because it could do a lot of swiping activity. With my team, we joke, we call it Tinder for code. You prompt the AI and it gives you a piece of code, you say yes or no. Say no, then it gives you another piece of code and maybe you give it another prompt. That iterative piece of making software using AI lends itself nicely to the concept of a phone.
The other thing is that it's not just the phone, it's the tablet. How crazy is it that we don't have good IDEs on the tablet? It's kind of surprising that Microsoft hasn't made VS Code for the tablet. We're the first kind of major IDE for the tablet as well. You can do everything that you could do on desktop on a tablet. You can attach a keyboard. You can go to any coffee shop in the world. Everything happens in the cloud. Your storage is in the cloud. Your AI is in the cloud. You don't need that much local capability and you can just write software on that.
Host
I haven't thought of that actually, but it's true. I have never seen someone coding on a tablet, just like I've never really seen someone coding on mobile. Let's see how that evolves.
Another feature you've developed that you alluded to earlier was bounties. Can you tell listeners a little bit more about what this is and how it's going so far?
Amjad Masad
Bounties is part of our portfolio of products that make it easy to make software. We realize that not everyone in the world wants to be a coder. The combination of the network effects on Replit, being able to discover a lot of software engineers like the guy who made the GPT Times for Balaji, the AI revolution, and what's happening in currencies, whether cryptocurrency or centralized currency, there's a lot of interesting things. It's a trifecta that allows us to build what I think is a fluid software labor market.
The theory of the firm is by this famous economist, Ronald Coase. The fundamental observation is that full-time employment is a bug, not a feature of the market. The reason full-time employment exists is because the transaction cost of doing something is really high. Uber is an example of something bringing the transaction costs almost down to zero. By doing that, it creates a huge amount of flexibility in the market. Anyone can enter the market, anyone can do the work, and they can do it on their own terms. There is no binding contract between the different parties.
Software has been something that has been very hard to contract out. When you do contract out, you get a lot of problems, a lot of issues running the software, a lot of quality issues. The fact that Replit is a fully integrated place with high quality software engineers using it allows us to be a place where someone can put a description for a piece of software that they need to get done. A high-quality developer can go and using AI can make that software very quickly. We're seeing people making software in 30-minute increments. Someone the other day put a request for a landing page up. They attached a Figma file. They got it back in 30 minutes. That's really never been done before. The combination of all the technologies we're building allows us to create this marketplace.
Host
How are you thinking about this marketplace among the existing marketplaces? Because it sounds like this is more peer-to-peer. I post a project, someone within the community puts their hand up, or maybe they're matched. When I think about the existing marketplaces for developers, their job is to vet a bunch of developers and say, "You have this skill set, you're this good of a developer," and also vet their clients and say, "This is a good project." Then they match people and take a commission. How do you see bounties as an ecosystem differentiating or providing something new?
Amjad Masad
I think it's already differentiated. The reason is because the development environment is built into the system. If Uber was a marketplace to connect you to people and then they have to go get their own car, you'd have to go meet someone at a coffee shop and then you and them go get a coffee. That's an absurd example, but that's what happens at Upwork or some of these marketplaces. I ask for a piece of software, you go make it in something that I don't know, then you send me a zip file, and then what do I do with that? On Replit, I just send you a link to a computer that's running your application. That's a fundamental innovation on top of that. All the services being integrated right there, your OpenAI API key, the cloud runtime, the database, just replicating this complete platform makes this process a lot more efficient.
Host
That's a great point. When I think about other scenarios where people hire developers, one of the massive gaps is the standardization. But it's also the AI tools that you're integrating. If someone gets a project and they can actually read the code, maybe they're not a developer but have some level of understanding of what they got back. In the past, they just get back this code that they can't understand at all. If they need to refactor anything or change anything or get a new developer to work on it, they really struggle. They also struggle to have some sentiment in terms of how good the code is.
Amjad Masad
Exactly. We want to create more market dynamics in the future that are more interesting. For example, this concept from crypto called staking. The idea that people can stake their money in order to say, "I'm going to build this better than anyone." I'll put up a bounty for a thousand bucks. You'll say, "I'm so sure of my ability to do that, I'm willing to put up a hundred bucks that I'll be able to deliver on time. If I don't, you can take my hundred bucks."
Host
I love that.
Amjad Masad
There's a lot of innovation to do in markets. You can also integrate AI things on top of that. For example, you can have an AI project manager where an AI takes one bounty and splits it into 10 bounties. We have this experimental product that I tweeted about recently. The moment you put in a bounty, we generate the scaffolding for the code. Scaffolding is basically the structure of the code. Every function that is not implemented could be its individual bounty. Once you add all these innovations together, you're going to get this super fluid market of AIs and humans where you can go from an idea to a product just like that. Crucially, the product you're going to get is something you're going to be able to iterate on in the future and get more bounty creators to contribute to. It's a living artifact that's working on the platform, as opposed to a zip file thrown over the wall.
Host
I love the idea of people being able to legitimately bet on themselves if they want to participate. When I talked to Balaji in our interview, we talked about this idea of evidence versus confidence. Someone can have a lot of confidence and say, "I have these requisites on my CV or I have this degree from somewhere." That instills confidence in other people that they can get something done. But what's better than confidence is evidence that you can actually get it done. If you can show, "I've done this project before, here's my work," and also put $500 down to say, "If you don't like it, you can have my $500," that is evidence that you can do the work and you're willing to put your money where your skills are.
With that said, let's move on to the very final section. We'll just rapid-fire go through these questions. They are questions that I've seen Sam Altman from OpenAI talk about as four important questions relevant to this advent of AI, especially as it progresses, taking the lens of not just Ghostwriter v1, but Ghostwriter v10 quite a ways from now. I'll just read you each question. I want to get your raw take on each of these.
The first question is, with that technology, how do you think that fundamentally changes society? Something I care deeply about is equality of opportunity.
Amjad Masad
The ability that anyone in the world can contribute and build something is just beautiful. It's amazing. It allows us to include everyone who's willing to work hard. That's the real Americanism, in my opinion. That's what America is built upon. The idea is that if you're willing to do the work, you're going to get the chance at doing the work. If you do the work right, you're going to get rewarded for it. I think that's beautiful.
However, I think that comes with increased inequality. The pie is going to be bigger, but there's going to be differences between the 10x difference becoming a thousand x difference. That's going to create some political issues because it's going to create envy and accentuate some differences. That part is not going to be fun. Reasonable people should be able to say it like it is. The 10x denialism is a small symptom of a larger issue, the issue of being uncomfortable with people being rewarded more because they have a better ability of doing something. That's going to be a big issue that society is going to have to deal with.
Host
I agree. Exponential technology means potentially exponential gains from certain people and not others. That's not necessarily inherently bad, especially if the pie is growing. But yes, that's going to be something that is divisive.
That relates to question number two, which is how do we ensure that it benefits us all? That does not mean it benefits us all equally, but how do we ensure that this technology is something that widely benefits the people participating in society?
Amjad Masad
As much as possible, having competition is going to be important. A world in which just one company controls the biggest AI models is not going to be great, but neither do two or three companies. The lesson we learned from the last generation of big tech is that especially if there's a monoculture like Silicon Valley, you're going to have similar decisions made. How do we prevent that in the world of AI? That's very important. Open source plays a part in that. As much as possible, pushing this technology to be easy to run and easy to develop, but not something that's closed off where a few people can control it.
A lot of what's called AI alignment today is not really aligning with what the average human being wants. It's aligning with the Silicon Valley average sensibility, which I don't think is good. We should try to build as much as possible neutral technology. My bias is going to be more freedom, more decentralization. Ultimately, these models are a tool. The tool itself is neutral, but the application of the tool doesn't always happen to be neutral. Some people hate the fact that it can be neutral and think that neutral is bad. That's where we really need to push back. No, actually, it should be neutral, and the user has to decide what to do with it.
Host
I think that's a good modifier. I'm going to combine the final two into one question. Given the exponential nature of this technology, the evolved form of it, how do we deploy it safely? You can interpret the word safe in your own terms. How do we deploy it safely? And how does governance play a role in that?
Amjad Masad
I like the word safety. A lot of the problems in the world today and the way our world is shaping up is this concept of safetyism. This concept that companies need to ensure our safety, and then they encroach on our freedom to ensure our safety. There was a recent gas stove debate. A government official said they might consider banning gas stoves. That impulse of "we know better than you," the nanny state impulse, I think it's bad. Every time governments have created atrocities, they've come from that idea. The government wanted the American people to be safe from the Japanese, that's why they created Japanese internment camps. You can use safety to do the most abhorrent thing in the world. We need to be skeptical. Anyone talking about safety probably wants something bad for you. At least when they want to impose something under the guise of safety, I would be super skeptical.
On the question of governance, I would also be skeptical. Maybe I'm showing my libertarian impulses here. I don't think anyone is inherently responsible for governance of technology. Government has a role to play. Government moves slow, and that's good because it needs to learn what's happening. If the government moved really fast on regulating cars, they would have gotten it super wrong. Now regulations are making self-driving move a lot slower than it should be. The process in the U.S. is actually fine for regulation. Move it slow, let's learn what's happening in the real world, and then let's have a reasonable debate and discussion. A democracy at the end of the day will arrive at the right decisions if there's sufficient freedom. I'm a big fan of debate and vicious debate in order to arrive at the truth. But that's going to take time.
The problem now is that there are some people talking about AI takeoff being so fast that you need to react to it really quickly. I think that'll get us into trouble. If you're going to force politicians to understand this technology before it's even deployed, you're really asking for trouble.
Host
On the flip side of that, something beautiful that's happened in the last year or so is that this technology has taken off really quickly, but it has also done so in a way where it's gotten in the hands of consumers. If you look at the charts of ChatGPT, their speed to 10 million users was so much quicker than Facebook or Snapchat or other apps in the past. While it is inviting some skepticism to say, "Oh my gosh, this is happening so quickly, this is scary, we don't know the implications," I also think you're going to see a level of pushback. At that point, you're taking something away from people, this superpower as we talked about it.
Amjad Masad
It's important to notice that nothing catastrophic happened. This technology is now massively deployed and nothing as bad as the sort of AI safetyism part of the debate said was going to happen. It's important to recognize this because the people arguing for extra controls are going to paint a future picture of catastrophe. The question to ask them is, "OK, there are millions of people already using this. Nothing really bad happened. So what are you actually worried about in concrete terms?"
Host
Your time dimension is so important here. You look at technologies like cars, and they were worried in the early days about cars scaring horses. Having a layer of humility to say we don't know how this is going to shake out is key.
Amjad Masad
I'm super optimistic about the future of the technology and everything that's happening. It's the best time to be a builder. Thank you.