Replit’s Path to Product-Market Fit—The $1 Billion Side Project

In the startup world, we often hear stories of founders who’ve dreamt of starting a company—any company—since childhood, with big business ambitions stemming from their days of running a lemonade stand. This is not one of those stories. 

Amjad Masad’s founder journey began instead with a singularly focused vision: bringing the power of software creation to everyone. Leading a business was more of a means to that end. “I always was attracted to the entrepreneurship track, but I’ve always ultimately been guided by my worldview when it came to creating technology,” Masad says.

He started with the premise that coding should be as simple as opening up a new tab in a browser. That might not seem all that groundbreaking in 2024, in an era when LLMs can now write code for you. But while collaborative, web-native tools like Google Docs emerged in the late 2000s, real-time collaborative coding in a web browser was still not widely available for a surprisingly long time. 

It certainly wasn’t available back in 2008, when Masad was a computer science student. If developers wanted to work together on a project, they typically needed access to the same physical machine. It made partnering on developer projects nearly impossible with anyone who wasn’t in your immediate vicinity — and forget about working with someone on the other side of the world.

This pain point was the genesis for the open-source project that would become Replit, which Masad often refers to as a “mad science experiment gone right.” He and a few of his university friends built the project, then known as Repl.it, as a preliminary way to get a REPL (a computer programming term for Read-Evaluate-Print-Loop) for your go-to coding language whenever you need one.

This “science experiment” was meant to stay that way — safely tucked away in the confines of a metaphorical lab, where it could be used more or less as a sandbox for other developers. In fact, after college he left it behind on the backburner, moving on to pursue a career as a software engineer and manager with stints at Codecademy and Facebook. 

However, Repl.it continued to grow and acquired a rabid fanbase of developers, all hooked on being able to collaborate in a coding environment in the web browser. Pretty soon, the early signals of product-market fit were so obvious, that Masad could no longer ignore them. What had started as a workaround for a college class, would eventually turn into a $1B side-project. 

Replit has soared in the past eight years since it launched as a startup. Today, it boasts 30 million users and has raised over $200 million in funding. And that pioneering streak has only continued. The company saw another big spike in growth after releasing its AI coding co-pilot Ghostwriter in 2022, which doubled its user base. Just this past month, it unveiled Replit Agent, an even more sophisticated AI-powered tool that can be thought of as a “pair programmer,” configuring development environments, installing dependencies, and executing code. 

In this exclusive interview, Masad retraces it all, recalling the early signs of customer love he saw for Replit as a college student, what finally persuaded him to pick the idea back up as a founder, and the contrarian bets he placed that turbocharged Replit’s growth. 

Replit’s story is a reminder that not just every company, but every founder’s path, looks a little different. So with that, let’s rewind the clock back to 2008 to the very start of Replit’s journey. 

CREATING THE EARLY PROTOTYPE 

As a college student at Princess Sumaya University in Jordan, Masad showed up to campus brimming with ideas. But reality quickly nixed these lofty plans. 

“As computer science students, the first thing we were faced with when taking a programming class was the initial slog of setting up the development environment. It wasn’t just a one-time thing, it was constant,” Masad says. Even after clearing that initial hurdle, the path ahead was still fraught. Once an environment was set up, it was nearly impossible to collaborate with anyone async. 

“If you talked to any engineer, they’d know the struggle,” he says. “I’d write some code, and it would work on my machine, but not my project partner’s or my professor’s. It made it incredibly difficult to code together.” 

This was back in 2008, when single-page applications like Google Docs were just emerging and more and more technology companies started shifting products to the cloud. Masad assumed that meant that a collaborative web-based code editor was just around the corner. “But when I looked around, there was nothing. There was no platform that allowed you to run code, share programs, and collaborate with others.”

So he decided to take matters into his own hands. With the naivety and ambition of a young developer, he set out to build the tool he envisioned. “I thought to myself, ‘How hard could it be?'” Masad laughs. “It turns out, the prototype was easy. I created a text box where you could input JavaScript, click a button, and it would evaluate the code. There was also a way to save and share it, and my friends loved it.” 

I immediately got feedback that what I was building was useful, which is what every developer wants to hear.

Buoyed by the positive response, Masad dove headfirst into this new approach, compiling various languages to JavaScript — eventually expanding his project to include more languages. (He even attempted to write a Python interpreter in JavaScript, a challenge that consumed him for 12 months before he eventually admitted defeat.) But a serendipitous discovery changed everything. “One day, I stumbled upon a project by a research group at Mozilla,” Masad says. “They were exploring the idea of compiling different languages to JavaScript, treating JavaScript as a bytecode interpreter. That captured my imagination.”

To his delight, it worked. He developed a prototype that was the first to support Python, Ruby, and several other languages in the browser, and open-sourced it. This early project, which he called Repl.it, laid the foundation for what would eventually become the company Replit, Masad’s integrated development environment platform. But let’s be clear here — in Masad’s eyes, at this point in time Repl.it was still very much a side project.

BRANCHING OFF

While working on Repl.it on the side, Masad was also laying the groundwork for a career in tech (one that he hoped would land him in the U.S. someday). Upon graduating, he nabbed a job as a software engineer at Yahoo’s office in Jordan. But he continued to work on Repl.it on nights and weekends to keep up with the growing demand for the project, alongside his co-founders (one of whom, Haya Odeh, he ended up marrying.) 

Even though no co-founder was working on the project full-time, by 2011, it continued to grow in popularity among developers, particularly student developers — a strong sign that it was already on its way to finding product-market fit, with or without Masad’s prodding. “It eventually exploded on Hacker News and GitHub that year,” he says. “While Replit, which was the demo app, did well, the thing that really took off was the underlying engine, called JS REPL.

This surge in growth happened to coincide with a burgeoning movement toward Massive Open Online Courses (or MOOCs) in Silicon Valley.“The idea of being able to learn online was capturing people’s imaginations,” Masad says. It wasn’t long until several U.S.-based tech companies, like the edtech company Udacity, started reaching out to Masad and his co-founders. But there were two that Masad crossed paths with that accelerated his career in particular: 

Codecademy 

Codecademy, a prominent player in the space that had just graduated from Y Combinator, began using many of the packages Masad had written for Replit. In 2011, shortly after JS REPL became a fan favorite of developers upvoting it on HackerNews, the Codecademy founders reached out to him for help, initially offering him a contracting gig for $15 an hour. This quickly evolved into an opportunity for Masad to join the company as its founding engineer.

Faced with a choice between launching his own startup in Jordan, where resources were limited, or joining Codecademy and moving to the U.S., Masad chose the latter. “The second option was more exciting to me,” he explains. “It allowed me to come to the U.S., which had been a dream of mine for a long time.”

Facebook 

After two years at Codecademy, Masad joined Facebook in 2013. He was drawn to the company by Mark Zuckerberg’s vision for Internet.org, an initiative aimed at providing universal Internet access. Timing is key to remember here. In 2013, the boom for building mobile products was well underway, and the prospect of shaping this future captivated Masad. 

“The world was in the middle of a massive shift, where soon everyone would have access to computers via phones and tablets,” he says. “It was exciting to see progress in democratizing access to the internet.” 

At Facebook, Masad initially worked on the Android app but soon gravitated back towards infrastructure projects. He became one of the early members of the React Native team, which sought to bring web technologies to mobile. Masad was responsible for writing a lot of the JavaScript infrastructure for React Native and was eventually tasked with building the company’s first JavaScript infrastructure team. 

“We wanted to modernize the JavaScript infrastructure on Facebook. So I built a team, which also ended up being responsible for building Jest, Babel, and Yarn. We essentially built the modern JavaScript dev toolchain at Facebook,” he says.

RETURNING TO REPLIT

Somewhere along the way, Replit faded into the background while Masad devoted himself to his roles at Codecademy and Facebook. But in 2016, while still at Facebook, Masad contemplated what he wanted to do next. His career decisions up until that point were centered around three core beliefs: 

• 1. Universal connectivity, computer literacy, and computer access were the future.
• 2. There are going to be a lot more developers, in a lot of different flavors, around the world. 
• 3. Software development could be much better.

“Those ideas dictated how my career progressed, and I was attracted to projects that fit within that rubric,” Masad said. Guided by these same principles, it dawned on him to turn the page back to the very beginning. 

As it happens, he wasn’t the only one still thinking about the project. “Even though Replit by this point had been abandoned and parts of it were broken, developers still kept coming back to use the product because there was nothing else out there,” Masad says. “There were a few ambitious projects like Cloud Nine which was acquired by AWS, but nothing that was web-native.” 

This was a signal that Masad was on the cusp of nascent product-market fit for a commercial product. When he took a step back, he realized going all in on this project could enable all three of his core beliefs and this was the push he needed to return to his original project after a multi-year hiatus. It was an exciting time; with only the existing demo page from his college days, Masad was essentially working with a blank canvas. 

But if Replit was going to become a billion-dollar company one day, Masad would have to swing for the fences. Here were two risky moves Masad made early on after picking Replit back up: 

Moving to the cloud

Steering Replit away from browser-based execution and pivoting to a cloud-based project was a hard left turn for Masad. After all, that was the initial wedge that made the original open-source project so successful. But when looking at the project with fresh eyes, and more importantly, with commercialization in mind, Masad started to view the browser more and more as a “dead end.” 

“The idea of downloading tens of megabytes of JavaScript on a Chromebook in Africa was just a non-starter,” he says. “My thinking was that cloud, with universal connectivity and internet bandwidth improving, would be the way to go. That was controversial at the time because there were a bunch of competitors springing up that were doing browser-based execution.” 

By pivoting away from the browser, Masad got to work rewriting the project for the cloud. “Docker had just come out and made it easier to do cloud-based sandboxing. I started adding more and more languages, and the project started taking off again.” 

Going all in on web-native

Another change Masad wanted to make, based on what he saw competitors shying away from, was making Replit web-native software. All of the other popular IDEs at the time were desktop apps, but Masad had a vision of seamlessly deploying code in the browser, with outstanding web-native UI to match. This meant building Replit again from the ground up. 

“One concrete truth about building web-native software is that URLs are very important,” he says. “Figma is an example. You can share your URL with someone and they can hop into the same environment as you, and the product becomes multiplayer by default. None of the online IDEs did that at the time, so that’s what we tried to replicate.” 

But pursuing this path wasn’t so simple. “Building as a web-native software meant we were taking on a much bigger challenge. Monetization was going to be harder, professional developers were not going to be able to use it at work for a long time, and that we weren’t going to be very fundable,” he says. 

Finding funding and monetizing the core product

Masad’s concerns about funding were valid. Replit first applied to Y Combinator in 2016 and was rejected. They struggled to raise a pre-seed round, despite the traction in users that they were seeing. 

Eventually, Christina Cacioppo, founder of Vanta, wrote them their first check, followed by Roy Bahat, a connection from Masad’s Codecademy days, who invested $600K in their endeavor. 

Boosted by this injection of cash and with new determination from a more solid product direction, Masad and his co-founding team started building a commercial product for the education market, albeit reluctantly. 

“At this time there were about 100,000 people using Replit,” Masad recalls. “They were a combination of students who found it through a teacher and self-taught teenagers who found it on their own. A lot of web-native teenagers didn’t even know the concept of files and folders, and so they would google a web-based collaborative coding environment, and they’d find us. We had built up this massive footprint on the web through people sharing links to REPLs that had this viral growth that gave us significant traction,” he says.

Given this strong foothold already, going after students and educators as the initial customer profile just made sense, but Masad had his reservations. “I always thought of the education market as a stepping stone for us. I didn’t like monetizing it, because I wanted Replit to be universally accessible.” 

My goal was to give kids around the world better access to tools and education. But in order to fund that reality, we had to sell to people who were willing to buy it, and that was students at the time.

By focusing on this narrower ICP, Replit started to see revenue roll in for the first time, and a “significant amount,” at that, according to Masad. But even with cash flowing in, it was still difficult to scale the team. “It was a rough period for us,” Masad says. “We were seeing this massive growth which was great, but to support this meant that I was on call all of the time.”   

Masad applied to YC two more times, another time in 2016 and once again the following year. 

He kept getting rejected — in fact, they never even made it to the interview round. “We just weren’t able to break into the core of Silicon Valley’s circle,” he says. Time and again, Masad and Odeh were told Replit was not a sustainable business. 

So you can imagine Masad’s surprise when in winter 2017, Sam Altman, who was serving as YC’s President at the time, reached out to Masad on Twitter. “He told me that Paul Graham had found us on Hacker News, and as it turns out, had actually wanted to build something similar to Replit before YC took off,” Masad says. Over the next few months, Graham and Masad stayed in touch over email, exchanging ideas and communications. Graham encouraged Masad to apply again, and he begrudgingly obliged. 

The timing couldn’t have been tighter, however; applications were due the next morning. Already burnt from filling out the form three previous times, Masad didn’t have the time (or the patience) to upload a video demo. “So I uploaded Rick Astley’s “Never Going to Give You Up” music video,” he says. It almost cost them the application. But overall, the folks who reviewed Replit’s application were impressed with the idea and the passion of the co-founding team, and Replit was accepted into YC W18.

CULTIVATING GROWTH

Replit’s seed check could only carry them so far. By 2018, Masad and his co-founders were looking to raise again. Fresh out of Y Combinator, Masad had articulated his long-term vision for Replit in his pitch deck (loosely inspired by Elon Musk’s “master plan”). It was broken into a three-step process, and even foreshadowed the prospects of AI’s influence on coding, well before LLMs became mainstream: 

• Step 1: Create a collaborative cloud development environment
• Step 2: Train machine learning models to make coding easier and more accessible (based on this network and the data that they would collect).
• Step 3: Become a platform that democratizes access to “the incredible wealth creation engine of making software on the internet.” 

Step one was squared away with the original prototype of Replit and the nascent product-market fit the product had found with students and educators. To make it to the next step in his plan, Masad knew he needed to change gears and focus on a bigger market. That meant getting in front of professional developers. 

“The biggest question for our business to this day is: ‘Will Replit ever be a daily driver for developers?’ I define developers in this case as professionals making money and doing economically productive activity,” Masad says. In this case, developers who were charging others for software they made on Replit, or users building a startup on Replit. “These were the signals that stood out to me as the areas to go down if we wanted to build something that was more than just a toy.” 

This persona became the North Star for the company. 

We wanted to change the dogma around cloud development sucking, because at the time, a lot of it did suck. The challenge became, “How do we make programming more fun, accessible and engaging?” 

It was at this point he started to think more critically about growing Replit as a disruptive technology, and he cites Harvard Business School professor Clayton Christensen’s “Innovator’s Dilemma” as key source material for these thoughts. “Disruptive technology starts from this bottom aspect of the market. As it becomes more powerful through some kind of trend or catches tailwinds, it’s riding on an improvement in the product or the supply chain where it eventually will take over,” Masad says. 

“The classic example is how x86 microcomputer PC architecture killed the mainframe and became the standard way of doing computing. That was the lens I was taking,” he says. “I was trying to figure out where we were in that power curve, and if we were getting to the point where people felt empowered by using Replit and not limited.” 

Given this new mindset, Masad made three crucial decisions at this stage in Replit’s journey. Looking back, it ended up being what took Replit from a seed-stage startup that was spinning its wheels to a company with hockey-stick growth: 

1. Going broad with a big bet

At this time, companies were placing bets on which languages would be the future — with most going all in on JavaScript. “I remember in 2018 this behemoth came on the scene,” Masad says. “Joel Spolsky, who is a legend in the programming community, spun out a project called Glitch,” Masad says. “Glitch’s seed round was $30 million. They put a stake in the ground and declared Javascript as the language of the future.” 

Replit, on the other hand, zigged where others zagged. “Once I saw this, I decided to place a bet on becoming a general purpose platform,” he says. “I thought we shouldn’t be in the business of betting on certain languages, rather, we should support as many languages as we can.” 

This hunch proved to be the secret lever for Replit’s growth as Python started growing rapidly, overtaking Java and Lisp in CS101 courses and gaining popularity in web development and data science circles. By the time the pandemic arrived in 2020, Replit had 10 to 100 times more users and engagement than their nearest competitor. 

Everything seems rosier in retrospect. But as founders will know well, the decision to pivot away from the traditional product advice of sticking to a narrow scope doesn’t come easily. 

It was an agonizing decision to start building for multiple languages because the Silicon Valley dogma is that you should do one thing and do it right.

“There’s Peter Thiel’s Zero to One way of thinking about things, which I think is brilliant. But it’s also a misinterpretation of how Steve Jobs did product management and design,” Masad says. “People viewed Apple software as being very simple. But they forget about the fact that Apple built an entire stack in order to produce those pixels that seem very simple. Simplicity does not mean you shouldn’t have a powerful platform.” 

He uses one of Replit’s customers, ByteDance, to further expand on this point. 

“It might surprise people that ByteDance is a customer of Replit. They use it for a platform they built called LarkBase and LarkSuite,” he says. “You can describe it as a G Suite or a Workday, or a hundred other kinds of U.S. software. But what’s interesting is now you see TikTok adding shopping, and TikTok Live and all these other features. Chinese software tends to grow a lot faster and be a lot more ambitious. That’s a challenge to the Silicon Valley model of building narrowly. Maybe there’s a historical lesson here. Where in the early innings of a platform shift, you can be narrow and you can find a lot of product market fit, but as your product starts to mature like a web or cloud or what have you, you actually have to be a lot more ambitious.”  

2. Adopting AI early 

Masad’s bet on being a general-purpose platform was starting to generate big returns. By the end of 2019, their user base surged to six million users, growing by 122%, powered almost entirely on word-of-mouth growth. Masad was also able to close a Series A round in early 2020 with much more ease than previous fundraising attempts. Up until this point, Replit had remained a close-knit team of three co-founders (Masad, Odeh, and Masad’s brother, Faris.) They invested the money in scaling headcount on their team.

But Masad’s crucial (and prognostic) decision to also invest in AI during this time period truly cemented Replit’s pole position. Given his long-term vision for Replit, when GPT-1 was first introduced back in 2018, Masad kept a close eye on it. 

“It wasn’t until GPT-2 that we started actually writing code and experimenting with it,” Masad says. “But understanding the machine learning stack was so hard and the training was still inaccessible so, while we had some experiments that were exciting, we never shipped anything.” 

All of that changed on a dime in 2021, with the launch of GPT-3. Replit launched Explain Code in partnership with OpenAI, which gave developers step-by-step natural language explanations of their code. “It was one of the first applications of GPT which came into production,” Masad says. 

This spurred the production of Ghostwriter, Replit’s Copilot equivalent (now just called Replit AI). But they ran into a huge roadblock: latency. “At the time, GPT-3 was slow and expensive. But I had this vision of this programming environment as being AI-native, not an add-on.” With that vision in mind, the Replit team decided to build it from the ground up. 

3. Partnering with enterprises

As Replit continued to make progress building out its core REPL product, and doing internal R&D projects with its AI tools, perhaps its biggest growth lever came in the form of partnerships with much larger enterprise businesses. 

Their first breakthrough came in June 2022, when Replit released its first AI code complete product based on Salesforce’s CodeGen model. “They had trained a model that was good, but slow,” he says. “We did a lot of work internally, rewriting large aspects of the code base to make it faster. When we released it, we made a lot of noise, as it became our first open-source powered product release.” 

Ever since, Replit has been steadily rolling out new generative AI capabilities for its users. By October that same year, Replit made its own AI code completion tool, Ghostwriter, available to all users by integrating it into its core platform. It also scored its first big enterprise partnership by integrating Ghostwriter with Google Cloud.  Both of these steps helped Masad inch a little bit closer to his long term goal of democratizing making software on the internet. (Step 3 in his long-term vision that he articulated to investors six years earlier.) Alongside the Ghostwriter integration, Replit was also one of the first companies to open source a code-specific LLM known as “replit-code-v1.5-3b.” 

It was the first time someone challenged the idea that “copilot” is a thing that only OpenAI and Microsoft can build or that you need billions of dollars to do. I think we unlocked a lot of people’s imagination that this space could actually be accessible for startups.

Turbocharged growth followed closely behind, with Replit’s user base doubling from around 10 million users in December 2021 to 20 million by February 2023. 

LOOKING FORWARD

Replit’s AI capabilities have only continued to evolve. In October 2023, Replit re-relaunched Ghostwriter as part of a suite of generative AI tools known as Replit AI. Most recently, it unveiled an even more sophisticated version of its initial code-writing tool, Replit Agent, an AI system that can create and deploy applications from just a few sentences — not a single line of code required. With a single text prompt, users can create and design working applications from scratch.

But Agent is only one part of Replit’s current go-to-market strategy. In July 2024, just a few months before the Agent launch, it rolled out Replit Teams, its B2B package that already counts Coinbase, Asana and Google as customers. 

“We’re planning to grow this in a PLG-fashion. Since we’ve launched Agent, we’ve seen a tremendous acceleration in adopting Replit and then bringing it to work,” Masad says. “So we’ve expanded our customer profile to include what we call ‘citizen developers’ — product managers, designers, data scientists and operations professionals. ” 

With each step forward in tech, Masad quietly moves his pawns closer to a world where coding is universal. Zooming out, Masad reflects on the future of software development as an industry. “Technology can make things more accessible, but most of the time it creates or pushes a category that was already emerging to become mainstream. It typically doesn’t entirely disrupt the existing way of doing things,” he says. 

This line of thinking informs how he sees the role of a developer adapting in the coming years. “I think that there’s going to be a bifurcation of roles between software engineer and software creator,” Masad says. “With AI agents and all of that coming down the pipeline, software creators are really going to be superhuman in their ability to deliver customer value.”

Engineers are not going away. We need engineers, in the same way that Instagram did not obviate the need for professional photographers.

Instead, he believes this new kind of role, the software creator, will also enable a culture of tinkering and experimentation — something that he believes Silicon Valley has lost touch with over the last few years — that can then spark more innovation.

“This type of experimentation could generate a lot of non-obvious startups. My advice for founders entering the space: Think about all the cool projects to do and give yourself enough time — especially if you’re still in school — to pursue that. Don’t start with the intent of starting a company or commercializing. Instead, ask yourself: what is something really interesting and cool that I can do with this technology?”

This article is a lightly-edited summary of our podcast episode with Masad. If you prefer to listen, check it out here.