In early November 2023, hundreds of developers from around the world filed into an auditorium in the heart of downtown San Francisco. It was OpenAI’s inaugural DevDay, an event where—à la WWDC—the company would preview its latest technology and discuss the future of AI. 

During the keynote address, Romain Huet, head of developer experience at OpenAI, took the stage to introduce the company’s latest tool: a voice-enabled intelligent assistant. To pull back the curtain on the assistant’s underlying function calls, Huet pulled up his terminal log. 

The terminal, an interface for using text commands to interact directly with a computer, is a 50-year-old staple of software engineering, beloved for its precision and efficiency while simultaneously hated for its extreme user-unfriendliness. Most people recognize the terminal from movies: a bare-bones user interface with a black screen and green text where hackers frantically type in commands. 

But the terminal that Huet displayed before the DevDay audience looked markedly different. It featured an IDE-like experience for editing commands, one-click sharing of terminal outputs with teammates, and an interface that allows users to interact with the terminal in natural language for coding and troubleshooting. As Huet demonstrated the inner workings of OpenAI’s new product, the terminal on display looked different from a traditional terminal because it was different: The command line interface powering OpenAI’s assistant was a terminal called Warp. 

Warp, founded by Zach Lloyd in June 2020, is a terminal reimagined for the modern developer. Today, as engineers race to build AI products, the terminal is proving essential due to its ability to navigate large volumes of data, automate batch processing and quickly execute scripts. Warp’s terminal checks these boxes while introducing new AI-based features that promote usability, collaboration and productivity. 

The notion, however, that Warp would be embraced by developers at OpenAI, among other companies including Atlassian, Cisco, Netflix and Salesforce, was never a given. A mere two years ago, the idea of reinventing something as sacrosanct in software engineering as the terminal flew in the face of accepted wisdom. 

In the summer of 2021, a post appeared on Hacker News titled “Warp: Fast, Rust-based terminal.” It announced the beta launch of the app and invited developers to sign up and try it.  Within minutes, the post began racking up comments—some intrigued, others skeptical. 

Criticisms boiled down to concerns over trust and security. Traditional computer terminals don’t require logins; if users entered passwords into Warp, would the company sell their data for profit? Would they make themselves liable to potential data breaches? There was also the ideological argument: Why should they pay for something new when its original version was free? And there was a sense of pride among those who understood the terminal’s eccentricities (and had memorized its arcane commands). To them, part of the appeal of the terminal was its difficulty of use; becoming skilled at using it was a rite of passage. 

“A lot of CEOs run away from this sort of experience because it’s unpleasant to have people telling you your idea is dumb,” says Andrew Reed, a partner at Sequoia. “One thing about Zach is that he’s a CEO who believes enough in the value of what he’s building that he is comfortable being yelled at on the internet.” 

Lloyd’s own experience had also taught him that skeptics of technology innovations aren’t always right. At Google, he’d been tasked with reimagining a software staple almost as foundational as the terminal—the spreadsheet—over fierce protests from Excel purists. The resulting product, Google Sheets, is now used by about a billion people per month. 

The internet could yell as much as it pleased. Lloyd wasn’t deterred. 

Lloyd’s interest in computer science began in college. He was drawn to software’s ability to make convoluted tasks faster and pulled his first all-nighter building a program that could find every word on a Boggle board. But Lloyd frequently encountered software for developers that was difficult to use, whether it had a steep learning curve or a nonintuitive design. “It drove me nuts,” he says. “It was like trying to get the shrink wrap off of the puzzle box before you can solve it. I wanted to actually solve the puzzle, not spend all my time trying to open the box.” 

Despite his interest in coding, following college graduation Lloyd let his curiosity guide him, testing the waters of a range of disciplines. He worked at NASA’s Jet Propulsion Lab, earned a master’s in philosophy, and even did a year of law school at Yale. “It took me longer than most people to figure out what I wanted to spend my life doing,” Lloyd says, “but I think there was value in being exposed to different fields and different ways of thinking.” It was in law school that Lloyd realized he missed building technology. He had an urge to create tools that were useful for people, and so he returned to engineering. 

In 2007, Lloyd was hired by Google to work on one of its newest products: Google Sheets. “I used to tell my friends I was working on Sheets, and none of them had heard of it,” Lloyd says. “When I would explain it, they would say, ‘Oh, Google has an Excel.’” 

The general opinion was that Excel—a decades-old, tried-and-true program—was irreplaceable. Why would anyone switch to a new spreadsheet software? But Excel had its shortcomings. The app was local, meaning if users wanted to collaborate on a spreadsheet, they had to export their work, email it to a teammate to edit, and hope that in the back-and-forth, versions weren’t lost and work wasn’t duplicated. Lloyd’s team saw an opportunity. What if they could offer a familiar experience but in the cloud, allowing for seamless, real-time collaboration?  

This undertaking came with immense technical challenges: recreating an on-prem program in the cloud, building real-time collaborative functions, and engineering the product in such a way that people could easily transfer their spreadsheets from Excel to Google Sheets and immediately continue their work. “We were trying to push forward innovation on document editing but had this everlasting tension where people have all these existing Microsoft Office files that we had to support,” Lloyd says, “and so transforming this technology in a way that would actually benefit people and not slow them down was very hard.” 

In his time at Google, Lloyd says he grew both as an engineer and a leader, learning to manage a team that grew in the following years from five people to more than 100. By the time Lloyd left in 2014, Sheets had hundreds of millions of users. “Everyone I knew, every start-up, was using Google Sheets,” he says. “By then, no one was referring to the product as Google’s version of Excel. People knew it as Google Sheets.” 

This was perhaps the most valuable insight Lloyd took from his time at Google: Some products are built as solutions to urgent, “hair-on-fire” problems. Others are built because someone is willing to question the status quo and decide it isn’t good enough. Some supposed “hard facts” are merely “hard problems” waiting to be solved. The experience of upending an accepted status quo was enticing, and Lloyd would go on to look for other supposed “hard facts” to turn on their heads.

In 2016, Lloyd cofounded a startup called SelfMade, which helps businesses succeed on social media. Leading his own company offered Lloyd a new set of lessons. At Google, he had been in a structured work setting, where there was always another rung of the ladder to climb and a clear path to get there. At a startup, there was no safety net and no definitive next step. Lloyd was happy to discover that he thrived in this strategic, competitive environment. But over time, something began to gnaw at him, a growing feeling that SelfMade wasn’t the right company for him. In 2019, he stepped down as its CTO.  

Lloyd spent the next few months reflecting on his experience at SelfMade, working to understand why it hadn’t been the right fit for him as a founder. “This might sound obvious, but I realized it’s really important to me to be working on a problem that I care deeply about,” Lloyd says. “SelfMade helped people with social media, and I don’t really use social media. It was a tech-enabled company, but the product was people’s labor. Building software is what I loved to do, so I wanted to work at a company where the software was the product.”

As Lloyd considered what he should do next, he thought back to the ethos that drove the development of Google Sheets: How do you build software that challenges a hard fact and makes peoples’ lives easier? More personal to Lloyd, how could you make developers’ lives easier? He considered the foundational tools of software development. And then it dawned on him that there was one that had never been improved: the terminal. 

While the average user relies on graphic user interfaces (GUIs) to communicate with computers, the terminal, which employs a command line interface (CLI), is an integral part of developers’ workflows and often the most efficient path to building software. While GUIs are great for certain tasks, there’s a class of developer tasks that are much better done in the command line. “A command line is more flexible than a GUI,” says Lloyd. “It allows developers to automate their own work. Most of the internet is actually just running command line apps. They are also the easiest to write—you just need to take text in and output text, no need to design anything with pixels.” Not only is working in the terminal faster, it is more memory efficient and offers higher precision and control over an operating system. 

Yet this critical tool, essentially unchanged since it first shipped in 1969, is also notoriously cumbersome. Engineers must be versed in the terminal’s library of complicated, esoteric commands and must use them precisely; one wrong character could brick your computer. What’s more, those working in the terminal can only access projects locally, making the development process a solo undertaking with little room for collaboration. “The terminal is truly archaic,”  Lloyd says. “I thought, What if I could take this old school basic interface of just text in and text out and really improve it?” This was the genesis of Warp. 

One of Lloyd’s takeaways from his time at SelfMade was the necessity of prioritizing product-market fit over growth. SelfMade had dedicated resources to expansion before it had a high level of user engagement and retention. Lloyd knew that with Warp, he would need to take incremental, intentional steps to build his business and make sure he had created something users would find invaluable. 

Lloyd’s first step was convincing his former colleague at Google, Shikhiu Ing, to come on board as a founding designer. They put together a prototype of Warp, which Lloyd knew would be critical in communicating his vision. “It’s just night and day when someone can actually see how your product will work. Once we showed it to people, they got it. That was huge for us.” Having Ing on board was also the social proof Lloyd needed that Warp was a good idea. “Once you get one smart person to sign on, it becomes marginally easier to get the next, and the next.” They hired two more talented engineers willing to take a risk on Warp, Michelle Lim and Aloke Desai, and got to work building the product.

It took six months, working tirelessly through the pandemic, for them to create a version of Warp they were happy with. Similar to Google Sheets taking on the incumbent Excel, Lloyd and his team had to balance the new with the familiar. “The real technical challenge with Warp was trying to innovate while making sure that the terminal that we’re building actually still works with people’s existing scripts and muscle memory,” Lloyd says. “We could have built something where we totally reimagined it and no one would have ever used it because none of their existing stuff worked.” At one point, struggling to match the terminal’s performance speed, the team decided to scrap Warp’s system entirely and rewrite it from scratch in Rust, a lower-level programming language popular with infrastructure engineers. The switch, while tedious, ensured Warp would offer terminal users the speed they were accustomed to. 

As for new elements, in Warp developers now have a modern IDE inside the familiar terminal. Warp allows teammates to share terminal commands and notebooks directly in their product by syncing them to the cloud. Warp also allows teammates to collaborate on the same terminal session in real time—similar to the multiplayer experience in Google Docs or Figma—so they can fight fires or debug issues. 

Lloyd’s next step was determining if there was an audience for Warp. Would the developer community even want it? To find out, in July of 2021, he published the article announcing the beta launch of Warp to Hacker News. 

Lloyd describes the critical feedback within the stream of comments on the Hacker News post as “painful—but it wasn’t going to kill the company.” As it turned out, those who were intrigued by the app far outnumbered the skeptics. Less than 24 hours after posting on Hacker News, over 10,000 developers had signed up to try it.

Over the next year, Warp iterated its product in beta with small groups of users. Lloyd solicited feedback, made adjustments, and in April 2022, the application—originally for MacOS—became available to the public. Within days, its user base grew to tens of thousands. In 2023, the company introduced Warp AI, an LLM-powered sidebar where users can ask questions if they need help remembering a command, fixing an error or tackling the next step in their workflow. In February 2024, Warp was officially released for Linux.

Lloyd’s hunch back in 2020, on the cusp of the LLM era, that an updated terminal would prove useful for developers, couldn’t have been more prescient. “Large language models are essentially command-line interfaces,” Lloyd told FastCompany, “and the terminal, the thing that we’re working on, is kind of the ideal way to interface with them.” As AI changes how code is written and software is consumed, the terminal is the most direct way to interface with AI—not just via chat but as a way of asking AI to do tasks for you. It’s a natural interface for launching AI agents (hence Warp’s June 2024 release of “Agent Mode,” which allows users to ask AI to directly accomplish tasks in the terminal).

Today, Warp has grown its user base to hundreds of thousands of developers. The company now has paying customers, both individuals and enterprises—proof of Lloyd’s vision that people will pay for AI development tools, even those that have traditionally been free utilities, when it helps ship better software more efficiently. His strategy is to strengthen Warp’s collaborative and intelligence features so that developers increasingly want to use Warp as a team—a tactic similar to Dropbox, Slack and Figma. “Monetization comes from companies wanting to pay to have their developers be more efficient and productive and working better together,” Lloyd says. “So our user growth is from the bottom up, but the revenue is like a B2B, enterprise or small team revenue model based on collaboration, intelligence and features that are for companies.” 

Lloyd’s long term goal is for Warp to help developers with a wider set of workflows, all available in one integrated app. Users will be able to log in to Warp and access not just a better terminal but improved versions of a range of tools, from code editing to DevOps and developer environment management. “Any place a developer is spending their time on a keyboard, Warp will be there,” Lloyd says.