How Sourcegraph adopted a writing first culture and ditched the office
As a code intelligence platform built for software developers, both Sourcegraph’s employees and customers are well-versed in asynchronous communication. According to co-founder and Chief Technology Officer Beyang Liu, it’s always been more important to work with the best people than to have a workforce who could work from the same location. “Software development has always lent itself well to async workflows,” says Beyang.
Committing to a fully-remote team has allowed Sourcegraph to embrace its strength as a company: using writing to communicate, collaborate, plan, and grow. According to Beyang,
“Writing is the tool that has allowed us to scale with clarity and consistency over the years, especially having a geographically-distributed workforce.”
When Sourcegraph went fully remote in January 2020, the team was already used to communicating via writing first, and it was obvious that working under the same roof wasn’t doing much to enable knowledge-sharing. “I think there’s often a ‘water-cooler mythos’ to office culture, where you run into people in the hallway and have these serendipitous conversations, but that wasn’t the vibe in our office,” says Beyang.
It’s not that meeting up face-to-face was discounted altogether—Beyang found that meaningful in-person interactions occurred outside of the day-to-day rhythm of the company. “All the face-to-face bonds were formed when we were all at an event like a conference together, where there was a source of information and entertainment that gave people something common to talk about,” he says. This discovery led them to devote their rent money to a travel budget instead.
While most companies that ditched their offices in 2020 found that they had to work hard to close the communication chasm that the change exposed, this wasn’t true at Sourcegraph. “Going fully remote was not a big leap for us,” says Beyang. This was due in large part to the writing-oriented tools and process the company introduced over the years.
While Sourcegraph has always defaulted to documentation in order to share knowledge, the way that documentation was organized and catalogued became unsustainable as the company grew.
“We had a lot of disparate Google Docs floating around in Google Drive,” admits Beyang, which is why their documentation became unwieldy and hard to organize. To solve this, Sourcegraph took a page from GitLab’s book and created a company handbook. The handbook contains a great deal of documentation that lays out the operating system of the company.
“Any fact about how the company operates is committed to the handbook.”
The handbook is used heavily for onboarding, and as the source of truth for background or supplementary information. “Organizing all this helped to show how different policies and processes within the organization relate to one another,” says Beyang. Keeping it in one place ensures that everyone has access to important company information that’s up to date.
In an effort to prioritize transparency, the handbook is open to the public (and even has a section on why the company prefers asynchronous communication). “The open source world values transparency a lot,” says Beyang, so it was important to stay true to that when creating the handbook.
In Sourcegraph’s early days, the company had a developer-oriented culture, and issues were surfaced and worked out asynchronously via an issue tracker, code review, and over Slack, but at some point, "it made sense to start writing things out more formally in documents, which is how the RFC [request for comments] process came about.”
The RFC process looks like this: if an employee has a proposal for something they’d like to change, they can draft an RFC. After documenting the proposed change, the employee can then tag the colleagues they’d like to receive feedback and approvals from within the document, quite literally a “request for comments.”
Requiring the RFC as the first step in proposing work means that meetings are only called when they’re truly needed. “Generally, if we can get something done without a meeting, we will,” says Beyang. It’s only when there’s some disagreement or tension from different stakeholders that it really makes sense to talk it out over Zoom.
As a developer-oriented company, Sourcegraph communicates with their customers the same way they do internally. That means taking the company value of transparency and extending it further than most companies would. Sourcegraph invites customers to join Slack conversations, comment on public-facing documents, and weigh in on the company’s product roadmap, which is available for anyone who wants to see it. While this level of inclusion is rare in the tech world, Beyang wouldn’t have it any other way.
“There’s so much coordinated labor in the open source world that is driven by transparency. Keeping those channels of communication open ultimately lets us do better work.”
Beyang doubles down on the idea that writing is the mechanism that has allowed the company to find its current success. “Talking is a great way to establish emotional resonance, but specifics can get lost,” he explains. “Writing something down helps you clearly articulate what you mean in the same way to everyone. It becomes a source of truth and levels the playing field.”
As for the future, Beyang wants to see documentation used as a default mode of communication even more than it is now. “There are many reasons to prefer the written word over the spoken word, and being able to build that level of clarity into every level of the organization will help us as we continue to scale.”
Beyang Liu is the CTO and co-founder of Sourcegraph, the code intelligence platform helping engineering teams understand, fix, and automate changes across their entire codebase. Many of the world’s leading use Sourcegraph to understand, fix, and automate across their code base, including 3/5 FAANG, Uber, Plaid, and GE. Prior to Sourcegraph, Beyang was a software engineer at Palantir Technologies where he developed new data analysis software for Fortune 500 companies. Beyang studied Computer Science at Stanford, where he published research in probabilistic graphical models and computer vision at the Stanford AI Lab.